diff --git a/docs/knowledge-retrieval-audit.md b/docs/knowledge-retrieval-audit.md
new file mode 100644
index 00000000..0b5ecbd8
--- /dev/null
+++ b/docs/knowledge-retrieval-audit.md
@@ -0,0 +1,266 @@
+# Knowledge retrieval audit
+
+This audit covers the existing file-backed knowledge system and a bounded
+manifest optimization. The [accepted scope](https://github.com/InferMatrixLab/InferRSI/issues/190)
+owns the comparison protocol, revision pins and acceptance criteria. CPU
+retrieval correctness and latency do not establish agent task quality, model
+performance or an inference-service gain. The frozen 20-pair experiment
+observed a **40.01% paired-median reduction in hot relevance-collect latency**,
+with a **95% bootstrap interval of 38.38%–41.06%**, on the local CPU setup.
+The frozen 24-case quality comparison preserves
+every serialized output exactly; the quality figures below describe existing
+retrieval behavior, not an improvement introduced by this manifest patch.
+
+## Construction and ownership
+
+Complete procedures live in
+[`skill_library`](../src/inferrsi/data/skill_library/README.md): `SKILL.md`,
+sidecar metadata, owner-local references, cases, scripts and licenses. The
+registry maps stable names to categorized paths and backends.
+[`build_skill_index.py`](../tools/build_skill_index.py) derives packaged JSON
+abstracts and source-file hashes from those owners; it does not create another
+editable procedure authority. Hand-maintained repository knowledge records
+retain their own IDs, source revisions, applicability and verification limits.
+[`catalog.py`](../src/inferrsi/catalog.py) validates and reads contribution,
+dependency and peer records from their existing package-data directories.
+
+| Axis | Meaning | Boundary |
+| --- | --- | --- |
+| Kind | `skill`, `knowledge`, `case` | Generated packaged skill abstracts are `knowledge`; checkout procedures are `skill` |
+| Tier | `contribution`, `dependency`, `peer` | Repository relationship, independent of kind and empirical validity |
+| Authority | `package`, `checkout`, `store` | Qualified identity ownership; collisions require explicit authority |
+| Origin, state, applicability | Where guidance came from and where it may apply | Import or candidate state does not establish validation, permission or promotion |
+
+[`KnowledgeReader`](../src/inferrsi/knowledge_operations.py) normalizes these
+sources into complete Bundles. A package reader retains one catalog/library
+snapshot. Checkout reads reload their current bytes; the optional
+[`KnowledgeStore`](../src/inferrsi/knowledge_store.py) retains exact revisions in
+an explicitly selected local SQLite workspace. Store writes publish candidate
+bundles transactionally and do not edit the package or checkout.
+
+A Bundle revision covers normalized metadata, every included resource hash and
+resolved dependency revisions. Packaged records with `skill_documents` retain
+the entire shared library as a conservative dependency closure. Returning a
+small context therefore does not imply constructing a small Bundle. Reading
+content never executes its procedure or scripts.
+
+The editable architecture artifacts are retained with this campaign:
+[English SVG](../evidence/knowledge-retrieval-20261027/results/architecture/inferrsi-knowledge-architecture.en.svg),
+[Chinese SVG](../evidence/knowledge-retrieval-20261027/results/architecture/inferrsi-knowledge-architecture.zh-CN.svg),
+[Chinese PNG](../evidence/knowledge-retrieval-20261027/results/architecture/inferrsi-knowledge-architecture.zh-CN.png)
+and [source/rebuild notes](../evidence/knowledge-retrieval-20261027/results/architecture/SOURCE-NOTES.md).
+
+![Knowledge architecture, English preview](../evidence/knowledge-retrieval-20261027/results/architecture/inferrsi-knowledge-architecture.en.png)
+
+## Entrypoints and actual context delivery
+
+| Entrypoint | Matching and ordering | Returned material |
+| --- | --- | --- |
+| `knowledge search` / `search_knowledge()` | All query tokens must occur in serialized catalog JSON; occurrence-count score, then ID | Catalog records; no complete Bundle construction |
+| Default `collect()` | Whitespace-token AND substring matching over normalized metadata plus entrypoint; qualified identity order | Validated items; `material=True` adds the entrypoint body |
+| `collect(relevance=...)` | Eligible generated packaged skills; distinct lexical words weighted by title, summary, tags/applicability and core; score then qualified ID | Same validated items; explicit IDs preserve caller membership/order |
+| `collect(context_bytes=...)` | Mandatory complete core/references/readsets/licenses/notices, then optional complete owner resources ranked by cue or path | Separate exact-revision `contexts`, explicit omissions and UTF-8 body-byte accounting |
+| `read()` / `resource()` | Explicit owner and optional/required exact revision | Complete Bundle view or selected exact resource bytes |
+
+The merged [PR #129](https://github.com/InferMatrixLab/InferRSI/pull/129)
+implemented the optional selection and context path; it is present in the
+current source. The Python API returns context resources and omissions;
+[`cli.py`](../src/inferrsi/cli.py) exposes `--relevance` and `--context-bytes`,
+prints complete context bodies in text mode and returns the full envelope with
+`--json`. `KnowledgeStore.collect()` delegates those same options to its reader.
+An external agent must consume the returned material; these calls do not
+automatically activate skills or execute a workflow.
+
+For generated packaged skills, `material=True` still means `record.json`, not
+the canonical `SKILL.md`. The latter is supplied through the optional context
+view or exact resource access. Item limits and serialized `max_bytes` accounting
+remain separate from context-body bytes. An oversized mandatory context is
+omitted atomically; the selected item may still be returned.
+
+The entrypoints are not interchangeable. The Python collect API supports
+`repo`, `tier`, `backend` and `scenario`, but the current collect CLI parser and
+dispatch do not expose them. The generator's README example
+`inferrsi collect --repo-type inferrsi ...` consequently uses an unsupported
+flag. This is a separate usability correction; this patch does not modify the
+parser or regenerate catalog content.
+
+## Redundant manifest work and the bounded change
+
+Previously `Bundle.manifest()` hashed resource bytes during dependency
+validation and again for its returned resource inventory. Directory
+dependencies also rebuilt a sorted resource traversal. Required resource paths
+were reparsed after the same content keys had already passed path validation.
+
+The change creates one sorted resource digest map per manifest call, reuses it
+for direct and directory dependency checks, and returns that same map. Present
+string required paths reuse the earlier content validation; missing, unsafe
+and malformed requirements retain their validation path. Nothing is cached
+across calls: the frozen dataclass still contains mutable dictionaries, so each
+manifest must observe changed metadata, bytes, resources and dependencies.
+
+The remaining work is deliberate in this slice: each Bundle still validates
+its complete closure, and collection still constructs eligible Bundles before
+selection/output budgeting. Cross-Bundle caching, metadata-first indexes and
+selection changes require their own mutation/identity contracts and evidence.
+
+[`test_bundle_manifest.py`](../tests/test_bundle_manifest.py) compares the
+manifest and revision with the complete digest formula, checks one resource
+hash per manifest call, and exercises mutations, deletion, unsafe paths,
+required resources and dependency revisions. Existing reader/selection/store
+tests cover source drift, stale projections, checkout changes, budgets and
+revision rejection. The paired CPU report retains fresh/reused reader results, CLI costs,
+exact output parity and all measured regressions.
+
+## Constructed retrieval evaluation
+
+The [prospective packet](../evidence/knowledge-retrieval-20261027/README.md)
+contains 16 development and 8 withheld synthetic cases, covering natural
+phrasing, symbols, multiple terms, backend filters and negatives. An explicitly
+configured Codex `gpt-6.1-sol` / `xhigh` author labelled relevant identities from
+original shipped procedure content before inspecting rankings. These are
+partial, source-grounded AI judgements, not human gold labels, production logs,
+a statistically representative sample or an exhaustive relevance inventory.
+The withheld queries stayed hidden from the runtime author and orchestrator
+until the runtime candidate was fixed. No query, label or budget was tuned
+after the quality result.
+
+Both API entrypoints use `repo="inferrsi"`, the same case backend and `limit=5`.
+Search takes `query=cue`; relevance collect takes `query="", relevance=cue`.
+Default collect is a separate diagnostic with `query=cue`. The context/material
+mode adds a 1,048,576-byte item budget and 65,536-byte context-body budget.
+
+| API mode | Success@5, 20 positives | MRR@5, 20 positives | Empty items, 4 negatives |
+| --- | --- | --- | --- |
+| Catalog search | 4/20 (20%) | 0.2000 | 4/4 |
+| Default collect diagnostic | 4/20 (20%) | 0.2000 | 4/4 |
+| Relevance collect | 20/20 (100%) | 0.9625 | 2/4 |
+| Context/material item ranking | 20/20 (100%) | 0.9625 | 2/4 |
+
+Development search/default Success@5 was 3/14 and MRR@5 0.2143; withheld was
+1/6 and 0.1667. Relevance/context-item Success@5 was 14/14 and 6/6, with MRR@5
+0.9464 and 1.0000 respectively. Each split had two negatives: search/default
+returned empty items for both, relevance/context for one. Full per-case outputs
+and byte digests are retained in
+[`quality.json`](../evidence/knowledge-retrieval-20261027/results/quality.json).
+All **96 outputs per revision** had exact canonical-serialization parity,
+including identity order, revisions, resource hashes, omissions and byte
+accounting. Source and complete-corpus identities also matched their initial
+snapshots at the end. The manifest patch therefore preserved this measured
+behavior; it did not create the relevance path or its higher scores. That path
+was already delivered by #129.
+
+The item metrics do not establish delivery of the correct procedure body. A
+separate **post-hoc diagnostic**, not a changed frozen gate, found labelled-owner
+contexts delivered in **19/20** answerable cases. In `withheld06`, the correct
+`performance_ceiling_analysis` item is first, but its complete mandatory context
+requires **170,782 bytes**, more than the frozen **65,536-byte** budget.
+`context_limit` omits that owner's context atomically; 51,702 bytes of other
+candidates' contexts are still returned. Calling this 20/20 correct body
+delivery would hide the limitation.
+
+Three source-backed examples explain the measured boundaries:
+
+- `DefaultStager` (`dev02`) yields no search/default items because this symbol
+  is absent from their indexed record text. Relevance examines the canonical
+  core and finds `dcp_async_save_performance` first; context mode delivers its
+  two required resources, **12,636 bytes**. A metadata projection miss does not
+  mean the underlying procedure lacks the answer.
+- The negative `sourdough levain hydration recipe` (`dev16`) returns five
+  systems skills through relevance and two contexts totaling **38,037 bytes**.
+  The scorer accepts any positive substring overlap; `recipe` also occurs in
+  unrelated systems procedures as `recipe`/`recipes`. It has no calibrated
+  no-answer threshold or generic-term penalty.
+- For `dev14`, `claim_contract` is fourth: broader performance/evidence skills
+  precede the more specific validator-freeze answer. Broad word overlap can
+  therefore dilute first-position precision even when Success@5 is satisfied.
+
+## Prospective CPU latency result
+
+Twenty paired groups (ten AB, ten BA) measured four fixed hot-query types on the
+local Intel Core i7-8565U / Windows 10 / CPython 3.12.6 setup. The frozen paired
+median reduction was **40.01%**, with 95% bootstrap interval **38.38%–41.06%**,
+satisfying the prospective CPU retrieval gate. The measurement loop retained
+640 API calls and 160 CLI processes over 836.162 seconds. All 400 cross-version
+output pairs and 80 exact CLI stdout pairs matched; source/corpus identities
+remained unchanged. Independent read-only review reproduced the raw statistics
+and bootstrap exactly.
+
+Regressions are explicit: API search `dev01` p50 **+1.19%**; CLI search `dev05`
+p50 **+2.50%**; CLI relevance decode p95 **1.810 → 3.046 ms (+1.236 ms, +68.31%)**.
+No cause is established by these observations. Each per-query p95 uses only
+20 repeats, not additional query types or a service tail guarantee. OS cache,
+clock/power state and background load were uncontrolled; no service/model/agent
+benefit is inferred.
+
+[performance.md](../evidence/knowledge-retrieval-20261027/results/performance.md)
+retains the complete per-mode/query p50/p95 tables for all 48 revision-specific
+groups, methodology and limitations; [latency-summary.json](../evidence/knowledge-retrieval-20261027/results/latency-summary.json)
+preserves exact nanosecond summaries, and [benchmark.json](../evidence/knowledge-retrieval-20261027/results/benchmark.json)
+retains all raw observations and output digests. The baseline profile and #129
+trace remain separate diagnostics.
+
+## Local validation and remaining qualification
+
+The complete local pytest attempt reported **10 failed, 1,340 passed, 6 skipped,
+2 errors and 135 subtests passed**, in 953.11 seconds. Rechecking the eleven
+selected failing/erroring functions on both revisions gave 9 failed, 4 passed
+and 3 errors per side. The CRLF fixture initially failed, then errored during
+the polluted focused run; isolated from the large-parameter fixture, both
+revisions reproduced 1 failed/1 passed with `Skill metadata must remain
+unchanged`. All initial failures/errors were thus reproduced on the baseline.
+This remains a failed whole-suite result.
+
+The observed blockers involve Windows environment-variable length, GBK
+decoding, unavailable Bash, temporary-path differences, an asset-manifest
+snapshot and unavailable `os.getpgrp`, plus the CRLF fixture behavior above.
+All 20 new manifest regressions passed; the runtime author separately reported
+80 targeted passes and a selected baseline run reported 104 passes. Default
+isolated wheel/sdist builds succeeded after an earlier no-isolation attempt
+failed for missing setuptools. All 469 packaged data files and the packaged
+runtime source matched the source byte-for-byte. Exact qualifications and the
+wheel digest are in
+[`validation-summary.md`](../evidence/knowledge-retrieval-20261027/results/validation-summary.md).
+The evidence inventory tests subsequently passed **23 tests**; the read-only
+inventory check reported **no errors, 9 bundles and 120 verified hashes**.
+`ruff check src tests tools` passed. A wheel-only, site-disabled smoke read
+108 records and reproduced all four `DefaultStager` output hashes from the
+formal candidate quality result. Its first harness attempt stopped at an
+unresolved temporary short-path import-origin check before retrieval;
+normalizing the expected path fixed that harness check. No package defect
+was established by that retry. Remote CI is pending at PR creation. Rerun
+the full suite on a clean supported Linux environment; local baseline
+reproduction is not cross-platform qualification.
+
+Independent reviewers performed read-only checks; their model/effort settings
+are unknown and are not listed as material authors. The review record retains
+the pre-measurement validator corrections, quality/body-delivery distinction,
+raw benchmark recomputation and diagram check. These checks are separate
+from human/maintainer goal acceptance, merge and issue closure, which remain
+with the authorized owner.
+
+Follow-up work should remain separately scoped: first make canonical-body
+discovery and API/CLI filtering consistent; then evaluate generic-term weighting
+and no-answer behavior with independent labels; then distinguish item discovery
+from complete required-body delivery and make required bytes visible before
+choosing a budget. A later performance slice may address repeated full-library
+closure/hash work across Bundles, but only with explicit mutation and revision
+contracts. This patch changes none of those behaviors.
+
+## Quality limits and historical evidence
+
+Less hashing cannot repair retrieval behavior. Default collect still has AND
+substring matching and identity order. Optional relevance still uses literal
+word overlap, positive-score eligibility and fixed field weights: it does not
+resolve synonyms, infer task intent or provide calibrated no-answer confidence.
+Whole-resource context can still exceed a budget, and callers still need to
+follow omitted links through exact revisions. These are existing boundaries;
+any observed misses or false positives must be reported with the measured cases.
+No score, order, query, context budget or omission rule changes in this patch.
+
+The closed [issue #128 owner report](https://github.com/InferMatrixLab/InferRSI/issues/128)
+records accepted delivery of bounded whole-resource software and its comparison
+packet: 36 attempts, 34 task PASS and 2 INCONCLUSIVE. It reports that C used
+25.74% more input/output tokens than B, with no accepted reduced-cost/preserved-quality
+benefit. Those are historical owner-reported results, not a rerun or new
+independent verification by this audit. This work preserves that conclusion;
+retrieval metrics and CPU parity cannot upgrade it to agent efficacy.
diff --git a/docs/plans/2026-10-27-knowledge-retrieval.md b/docs/plans/2026-10-27-knowledge-retrieval.md
new file mode 100644
index 00000000..b1d6a61c
--- /dev/null
+++ b/docs/plans/2026-10-27-knowledge-retrieval.md
@@ -0,0 +1,46 @@
+# Knowledge retrieval implementation plan
+
+**Goal:** Reduce repeated work in `Bundle.manifest()` while retaining the same
+validated manifest, revision and retrieval results.
+
+**Architecture:** Reuse resource digests and validated content paths within one
+manifest call. `Bundle.metadata`, `Bundle.content` and `Bundle.dependencies`
+remain mutable, so no digest or manifest survives that call. Keep the existing
+package snapshot, checkout and store ownership boundaries.
+
+**Scope:** `src/inferrsi/knowledge_operations.py`, focused regression tests in
+`tests/test_bundle_manifest.py`, and the source-grounded audit in
+`docs/knowledge-retrieval-audit.md`. Catalog/source bytes, scoring, query
+semantics, ordering, budgets, omissions and dependency closure are unchanged.
+Performance gates, revision pins and frozen measurement inputs belong to the
+governing acceptance issue and its evidence packet, not reusable defaults.
+
+## Plan -> code -> test -> review
+
+1. Inspect the reader, selection helper, store and existing tests. Compare the
+   alternatives: local digest reuse is bounded; persistent Bundle caching is
+   unsafe with mutable inputs; changing selection is outside this slice.
+2. Add a small synthetic Bundle fixture with file and directory dependencies.
+   Verify the complete manifest and revision against the existing digest
+   formula; count hashing calls and verify each resource is hashed once per
+   call. Verify repeated calls detect changed bytes, removed resources, unsafe
+   paths, changed metadata/requirements and changed dependency revisions.
+3. In `Bundle.manifest()`, build one sorted resource digest map after content
+   validation and required-resource checks. Reuse it for dependency validation
+   and the returned resource inventory. Reuse validated required paths while
+   preserving missing/unsafe-path validation and existing error categories.
+4. Run `python -m pytest tests/test_bundle_manifest.py
+   tests/test_knowledge_operations.py tests/test_collect_selection.py
+   tests/test_knowledge_store.py` and `python -m ruff check
+   src/inferrsi/knowledge_operations.py tests/test_bundle_manifest.py`.
+   The accountable parent runs the full suite, installed-package checks and
+   paired performance/quality protocol after its independent fixture freeze.
+5. Review the diff for persistent caching, changed resource closure, digest
+   order, invalid-input behavior, private data and unrelated changes. Record
+   actual checks separately from pending measurements. Commit with the actual
+   configured identity, DCO sign-off and material-agent attribution; hand the
+   revision to the accountable parent for independent review and integration.
+
+The audit distinguishes the delivered optional whole-resource context from
+historical efficacy claims. It does not reinterpret the prior campaign or
+claim that less hashing improves selection quality or agent outcomes.
diff --git a/evidence/knowledge-retrieval-20261027/README.md b/evidence/knowledge-retrieval-20261027/README.md
new file mode 100644
index 00000000..2680a2c7
--- /dev/null
+++ b/evidence/knowledge-retrieval-20261027/README.md
@@ -0,0 +1,50 @@
+# CPU knowledge retrieval evidence
+
+Acceptance scope and ownership: [InferRSI #190](https://github.com/InferMatrixLab/InferRSI/issues/190).
+This packet tests a manifest implementation that preserves retrieval behavior.
+The frozen quality/parity and paired CPU performance gates passed. Campaign
+acceptance, merge and issue closure remain with the authorized owner.
+
+| Segment | Contents |
+| --- | --- |
+| `plan/` | Prospective protocol, current freeze, preserved first freeze and pre-measurement review amendment |
+| `validator/` | Standard-library runner, 16 development and 8 withheld synthetic cases with source-grounded partial relevance labels |
+| `results/` | Complete quality comparison, paired raw API/CLI measurements, baseline profile, #129 delivery trace, editable EN/ZH architecture artifacts and qualified local software/environment summaries |
+| `patch/` | Fixed runtime candidate and exact source identity |
+
+The current prospective freeze is
+`a465367319f3058398eb1f8579ee5d4174b8cb90457793e90788fb5796151057`.
+Independent review corrected completeness checks and removable assertions before
+formal candidate measurement; `plan/amendment-1.md` records the change. The
+withheld split was released only after the runtime candidate was fixed. Fixture
+text and labels did not change between freezes. Root `.gitattributes` preserves
+evidence bytes; copying or committing must retain the recorded hashes.
+
+The full quality result has **96 outputs per revision, exact parity and unchanged
+source/corpus identities**. For 20 labelled answerable cases, search/default
+collect score Success@5/MRR@5 = 0.20/0.20; relevance/context-item ranking scores
+1.00/0.9625. Four labelled negatives return empty items in 4/4 search/default
+cases and 2/4 relevance/context cases. These are unchanged existing #129
+behaviors, not a quality improvement from this runtime patch.
+
+Context-item ranking does not establish correct body delivery. A post-hoc
+inspection found the labelled owner's context delivered in **19/20** positive
+cases. `withheld06` finds its correct item first but requires 170,782 context
+bytes against a 65,536-byte budget, so that owner's context is omitted. This
+diagnostic does not change the frozen quality gate. Synthetic, AI-authored,
+partial source labels are not human gold judgements or complete Recall; no
+agent-task efficacy or production-distribution claim follows.
+
+The 20-pair CPU result observed **40.01%** hot-relevance paired-median latency
+reduction, 95% interval **[38.38%, 41.06%]**, with all 400 output pairs and
+80 CLI stdout pairs equal. [performance.md](results/performance.md) retains
+all per-mode/query p50/p95 rows, including search regressions and CLI relevance
+decode p95 **1.810 → 3.046 ms**. Independent read-only recomputation passed;
+reviewer model/effort is unknown and no human/maintainer acceptance is recorded.
+
+See the [audit](../../docs/knowledge-retrieval-audit.md) for source contracts,
+per-query examples, failure qualifications and follow-up priorities. Local
+software verification includes a failed whole-suite run whose initial failures
+were reproduced on baseline, successful focused manifest checks and successful
+isolated wheel/sdist construction; it must not be described as a green full
+suite. No model, GPU or inference-service benchmark ran.
diff --git a/evidence/knowledge-retrieval-20261027/patch/README.md b/evidence/knowledge-retrieval-20261027/patch/README.md
new file mode 100644
index 00000000..788fc65e
--- /dev/null
+++ b/evidence/knowledge-retrieval-20261027/patch/README.md
@@ -0,0 +1,11 @@
+# Runtime patch boundary
+
+The evaluation-material author changed no runtime source. `candidate.json`
+records the independently authored runtime commit and exact Python-source
+digest used by the formal comparison. The implementation reuses digest/path
+work within a manifest call; it introduces no cross-call mutable cache.
+
+Runtime integrity tests and independent review belong beside those revision
+records. Exact-output parity does not replace mutation or unsafe-path regression
+checks. Later evidence/documentation integration must preserve the tested
+runtime bytes; merge and issue closure remain with the authorized owner.
diff --git a/evidence/knowledge-retrieval-20261027/patch/candidate.json b/evidence/knowledge-retrieval-20261027/patch/candidate.json
new file mode 100644
index 00000000..0b34d094
--- /dev/null
+++ b/evidence/knowledge-retrieval-20261027/patch/candidate.json
@@ -0,0 +1,12 @@
+{
+  "schema_version": 1,
+  "governing_issue": "https://github.com/InferMatrixLab/InferRSI/issues/190",
+  "baseline_sha": "4817f92121df73fd9cd5fb4a319e3c56a68b4a5b",
+  "candidate_runtime_sha": "407bb244651bec1c68f4b3e94b0c697b233c4a09",
+  "runtime_author_commit": "864fd74e709a1411e4f8dfd3051cc9b737d8c333",
+  "candidate_python_sources_sha256": "d6edbaf975cf3bc00902d6aa7cf2bb955dc58b3e37c74f59bb226c5f0d2fa261",
+  "scope": "Reuse sorted resource digests and validated paths within each Bundle.manifest call. No cache survives a call and retrieval outputs must remain identical.",
+  "runtime_source": "src/inferrsi/knowledge_operations.py",
+  "runtime_regressions": "tests/test_bundle_manifest.py",
+  "integration_note": "Evidence, audit, diagrams and read-only inventory registration are integrated separately; compare their packaged runtime source to this fixed tested candidate. No merge or issue closure is implied."
+}
diff --git a/evidence/knowledge-retrieval-20261027/plan/README.md b/evidence/knowledge-retrieval-20261027/plan/README.md
new file mode 100644
index 00000000..e61d8bbc
--- /dev/null
+++ b/evidence/knowledge-retrieval-20261027/plan/README.md
@@ -0,0 +1,45 @@
+# Frozen evaluation protocol
+
+Governing acceptance: [InferRSI #190](https://github.com/InferMatrixLab/InferRSI/issues/190).
+The machine-readable protocol is `protocol.json`; the freeze manifest commits
+protocol, independently constructed inputs and validator bytes before candidate
+measurement. This packet is authored by Codex `gpt-6.1-sol` / `xhigh`.
+
+1. Read the shipped source content, construct 16 development and 8 withheld cases,
+   and label defensible relevant package identities by original resource paths.
+   Do not consult rankings while labelling. The partial labels are sufficient
+   for Success@5/MRR@5, not complete Recall. The queries are synthetic and have
+   no production-distribution claim.
+2. Validate fixture references and freeze all authored inputs and scripts. Keep
+   withheld queries/labels outside the runtime author's and orchestrator's view
+   until the candidate revision and source digest are fixed.
+3. Profile the read-only baseline separately. Run quality/parity and the frozen
+   20 balanced paired groups with one interpreter and an equal complete corpus.
+   Each process explicitly imports the selected checkout rather than a possibly
+   unrelated editable install. Keep all raw timings and exact-output digests.
+4. Report search and relevance retrieval separately at identical API scope
+   (`repo=inferrsi`, `limit=5`). Report default collect, context delivery and CLI
+   startup separately. The current collect CLI lacks repo/backend filters;
+   its wider-scope startup cost is not an equivalent retrieval-quality result.
+5. Apply frozen gates, self-check output integrity and submit artifacts for an
+   independent review. This evaluation does not establish model task efficacy,
+   full-library recall, service throughput or OS filesystem-cache-cold latency.
+
+The primary performance statistic is the median of 20 paired reductions, each
+formed from the median latency of four fixed hot relevance queries. The 95%
+percentile bootstrap resamples complete paired groups with seed `19020261027`
+for 10,000 iterations. The prospective issue bar requires at least 20% median
+reduction and a lower confidence bound above zero. No trimming or post-hoc
+choice of the fastest query, summary or run is allowed.
+
+Exact parity covers the complete canonical serialization, with readable
+projections for identity order, revisions, resource-hash inventories, omissions,
+collision metadata, material bytes and context bytes/budget. The API timer
+excludes the evaluator's serialization and hashing work. Full CLI timings
+include process creation, imports, command work and stdout capture; decoding is
+measured separately. Fresh-reader API calls include reader construction;
+reused-reader calls follow one untimed pass across all four hot queries.
+
+No candidate has been timed while this protocol is being authored. Missing or
+incomplete evidence remains INCONCLUSIVE. Existing #128/#129 evidence stays
+historical and is not relabelled as this campaign's measured result.
diff --git a/evidence/knowledge-retrieval-20261027/plan/amendment-1.md b/evidence/knowledge-retrieval-20261027/plan/amendment-1.md
new file mode 100644
index 00000000..79e2534f
--- /dev/null
+++ b/evidence/knowledge-retrieval-20261027/plan/amendment-1.md
@@ -0,0 +1,28 @@
+# Pre-measurement amendment 1
+
+Independent review found two fail-closed defects after the first packet freeze
+and before any formal candidate evaluation or timing:
+
+- Positional `zip` comparisons could silently ignore missing output rows. A
+  synthetic candidate with incomplete API/CLI measurements could satisfy the
+  timing gate without complete evidence.
+- Fixture and warmup checks used Python assertions, which optimized Python could
+  disable.
+
+The validator now requires the exact case-by-mode sets once per side, the full
+pair-ID range and each prescribed AB/BA order. Missing, extra or duplicate rows
+or pair IDs make the comparison INCONCLUSIVE and prevent a performance PASS.
+Quality comparison explicitly checks its complete expected set, including all
+96 rows per side for the full 24-case run. Keyed comparisons replace `zip`.
+Critical checks raise explicit exceptions; child processes set PYTHONOPTIMIZE=0
+and record optimization flags. Synthetic selfchecks cover incomplete/duplicate
+quality rows, missing benchmark rows and duplicate paired groups.
+
+The original `plan/freeze-v1.json` is preserved with content digest
+`2a90de774129b06597c3c99e5b06d05bf735026829deb84376d06a985cdcc276`.
+Before this supersession, only source-label/static validation and
+baseline-v-baseline mechanics had run, never formal candidate measurement.
+The revised `plan/freeze.json`
+supersedes it prospectively. Fixture text, relevance labels, scoring rules,
+query workloads, numeric gates and bootstrap method/seed are unchanged. The
+runtime candidate was already fixed before the withheld fixture was released.
diff --git a/evidence/knowledge-retrieval-20261027/plan/freeze-v1.json b/evidence/knowledge-retrieval-20261027/plan/freeze-v1.json
new file mode 100644
index 00000000..cde2e748
--- /dev/null
+++ b/evidence/knowledge-retrieval-20261027/plan/freeze-v1.json
@@ -0,0 +1,13 @@
+{
+  "schema_version": 1,
+  "frozen_at_utc": "2026-10-27T11:17:10.632059+00:00",
+  "files": {
+    "plan/protocol.json": "dc0fbbb4438a072fe1dbc530f5a60a4b1e1e779cc32697fd9adccc0eac3bc4fc",
+    "plan/README.md": "12e0bb007b6a4ce7a4177e1843ae46ab3327fdd7fcc6591c3332db5684f83c3d",
+    "validator/development.json": "d2f428c41787bfa6279e8a0b56532e76554482ab98dca553e7e9febe287e8dfb",
+    "validator/withheld.json": "5bf9b8e55871a986cf7848959bf83e401821939acd13ee067563dc83952182e9",
+    "validator/evaluate.py": "47d36dd7b0a0967cb74ae53a31646b8bf063958b3845769447c6e8af8f42e5e1",
+    "validator/README.md": "2ca7c11a3add1a277c7c60dc24820d8fe5077e239d5deaf4ea6b0fe1195abe7c"
+  },
+  "content_sha256": "2a90de774129b06597c3c99e5b06d05bf735026829deb84376d06a985cdcc276"
+}
diff --git a/evidence/knowledge-retrieval-20261027/plan/freeze.json b/evidence/knowledge-retrieval-20261027/plan/freeze.json
new file mode 100644
index 00000000..ac0c6262
--- /dev/null
+++ b/evidence/knowledge-retrieval-20261027/plan/freeze.json
@@ -0,0 +1,17 @@
+{
+  "schema_version": 1,
+  "revision": 2,
+  "supersedes_content_sha256": "2a90de774129b06597c3c99e5b06d05bf735026829deb84376d06a985cdcc276",
+  "frozen_at_utc": "2026-10-27T11:22:45.564236+00:00",
+  "files": {
+    "plan/protocol.json": "91f2966ddf7d37683cd2a79cd995ec1ad7b41993ea1e6c2cd008db48a3b99351",
+    "plan/README.md": "12e0bb007b6a4ce7a4177e1843ae46ab3327fdd7fcc6591c3332db5684f83c3d",
+    "validator/development.json": "d2f428c41787bfa6279e8a0b56532e76554482ab98dca553e7e9febe287e8dfb",
+    "validator/withheld.json": "5bf9b8e55871a986cf7848959bf83e401821939acd13ee067563dc83952182e9",
+    "validator/evaluate.py": "e5b8c99f6955420e67bb0d9fbcc4c08a6c8a95d857385a5d9f0a0f2f74e0f194",
+    "validator/README.md": "e9b5f55e2b74b184a6bf2d31e342074e120663fa48286d6b14496491bc39eea4",
+    "plan/amendment-1.md": "4feac787c0fff35bab49b04f1dc2fdd7648febc25a770a59d359f7d7ddb30c2c",
+    "plan/freeze-v1.json": "a1f727ea4914b194469e5183cc2df6c7ec2b42614cc294dc14c0ec0fbd8dcb05"
+  },
+  "content_sha256": "a465367319f3058398eb1f8579ee5d4174b8cb90457793e90788fb5796151057"
+}
diff --git a/evidence/knowledge-retrieval-20261027/plan/protocol.json b/evidence/knowledge-retrieval-20261027/plan/protocol.json
new file mode 100644
index 00000000..4dbc1827
--- /dev/null
+++ b/evidence/knowledge-retrieval-20261027/plan/protocol.json
@@ -0,0 +1,58 @@
+{
+  "schema_version": 1,
+  "campaign": "knowledge-retrieval-20261027",
+  "governing_issue": "https://github.com/InferMatrixLab/InferRSI/issues/190",
+  "baseline_sha": "4817f92121df73fd9cd5fb4a319e3c56a68b4a5b",
+  "author": {"tool": "Codex", "model_id": "gpt-6.1-sol", "effort": "xhigh", "role": "evaluation-material-author"},
+  "scope": "CPU-only offline retrieval software behavior and latency; no model, engine, service, production-distribution, complete recall or agent-efficacy claim",
+  "fixture": {
+    "total": 24,
+    "development": 16,
+    "withheld": 8,
+    "construction": "Independent source-content judgements made without retrieval rankings; synthetic questions, not private production logs or a statistically representative sample.",
+    "release": "Withheld text and labels must not be read by the runtime author or orchestrator before the candidate revision and runtime source digest are fixed. Freeze manifest commits their exact bytes."
+  },
+  "quality": {
+    "repo": "inferrsi",
+    "limit": 5,
+    "entries": {
+      "search": "catalog.search_knowledge(query=cue, repo='inferrsi', backend=case_backend, limit=5)",
+      "relevance": "KnowledgeReader.collect(query='', relevance=cue, repo='inferrsi', backend=case_backend, limit=5)",
+      "context_material": "Same relevance call with material=True, max_bytes=1048576, context_bytes=65536",
+      "default_diagnostic": "KnowledgeReader.collect(query=cue, repo='inferrsi', backend=case_backend, limit=5); report separately"
+    },
+    "answerable_metrics": ["Success@5 = mean(any labelled relevant identity in ordered top five))", "MRR@5 = mean(1 / first labelled relevant rank), or zero if absent"],
+    "denominator": "Answerable cases only; report entrypoint and development/withheld splits plus per-case judgements. Labels identify defensible relevant results, not an exhaustive relevance inventory. Do not call this Recall.",
+    "no_answer": "Separate count, returned identities and no_match behavior; do not include in Success/MRR and do not treat empty lists as proof that the corpus has no answer.",
+    "equivalence": "All four modes on every case must have exactly equal canonical JSON bytes between revisions, including order, omissions, resource hashes, revisions, material and context budgets. Retain byte count, SHA-256 and readable projection of each exact output.",
+    "serialization": "json.dumps(value, sort_keys=True, ensure_ascii=False, separators=(',', ':'), allow_nan=False).encode('utf-8')"
+  },
+  "performance": {
+    "pairs": 20,
+    "order": "Pair 0 AB, pair 1 BA, alternating for 10 groups of each order. A=baseline, B=candidate. Each side runs in a new isolated subprocess, sequentially. No concurrency.",
+    "hot_cases": ["dev01", "dev05", "dev09", "dev13"],
+    "warmup": "One untimed hot relevance collect per hot case on the same KnowledgeReader, followed by one default dev05 and one context_material dev09 warmup before measurements; record reader record/library cache population. Reuse this reader for hot measurements. Cold measurement creates a new reader inside the timed interval. OS filesystem caches are not evicted or claimed cold.",
+    "api_workloads": ["hot and fresh-reader relevance for all four hot cases", "search for all four hot cases", "hot and fresh-reader default collect using dev05", "hot and fresh-reader context_material using dev09"],
+    "timing": "perf_counter_ns around API call only; output serialization/hashing outside API timer. Warm measurements include normal output construction. Raw nanoseconds retained per query/pair/side/mode.",
+    "cli": "Separate full process subprocess.run([same_python, '-S', '-m', 'inferrsi', ...]), UTF-8 decode+JSON parse timing and raw stdout SHA-256/byte count. search uses --repo-type inferrsi; collect does not expose repo/backend filters and is explicitly an unfiltered CLI diagnostic. No CLI numbers pooled into the hot-API gate.",
+    "cli_cases": ["search dev05", "default collect dev05", "relevance collect dev05", "context/material collect dev09"],
+    "summaries": "Per mode/query/side p50 and p95 from all raw observations using linear interpolation at (n-1)*p. No trimming, outlier removal, retries or stopped-run pooling.",
+    "primary_estimator": "For each pair, take median of its four hot relevance query latencies for each side. Paired reduction r_i = 1 - median(candidate_queries_i) / median(baseline_queries_i). Report median(r_i). Each pair is one resampling unit, preserving its query set.",
+    "bootstrap": {"method": "percentile bootstrap of paired reductions; each draw samples 20 reductions with replacement, then computes their median", "iterations": 10000, "seed": 19020261027, "confidence": 0.95, "quantiles": [0.025, 0.975], "quantile_method": "linear interpolation at (n-1)*p"},
+    "gate": {"minimum_paired_median_reduction": 0.20, "lower_95_percent_bound_must_exceed": 0.0},
+    "limits": "Machine background load and filesystem-cache state remain observational limits. This is neither OS-cache-cold nor an engine/service benchmark. Cold/API/CLI regressions reported individually. Estimated bounded run length 5-10 minutes; actual elapsed and every completed sample retained. An incomplete run is INCONCLUSIVE.",
+    "completeness": "Before summary/gating require pair IDs exactly 0..19 once each with prescribed alternating AB/BA order and each side's exact16 API plus4 CLI case-by-mode set, no duplicates or extras. Quality requires the complete expected24x4 output set per side. Missing/duplicate/extra structure is INCONCLUSIVE and cannot satisfy the performance gate."
+  },
+  "source_control": {
+    "loading": "Explicit checkout/src overrides editable installs; -S excludes site initializers and editable hooks. Explicit exceptions verify inferrsi, catalog, knowledge_operations and CLI resolved inside requested source tree. Critical checks do not use removable Python assertions. Children set PYTHONOPTIMIZE=0 and record sys.flags.optimize; parent optimization is recorded. Same interpreter for both versions.",
+    "identity": "Record explicit git HEAD, dirty tracked-source status, all Python source relative-path SHA-256s and aggregate; all shipped repo_packs/**/*.json, knowledge_packs files and entire skill_library relative-path file SHA-256s, aggregate, counts and bytes. This covers contribution, dependency and peer records plus their package and library resources. Ignore incidental __pycache__/.pyc only. No personal absolute paths or host names in artifacts.",
+    "corpus": "Both variants must have exactly equal complete shipped corpus digest; do not use a reduced custom corpus. Corpus inventory covers all shipped records and library resources, not merely retrieved items.",
+    "freezing": "Freeze protocol, fixture and validator scripts in plan/freeze.json before candidate timing. Separate withheld fixture remains sealed until candidate freeze. A validator change requires a new freeze manifest and explanation; never silently amend frozen evidence."
+  },
+  "diagnostics": {
+    "profile": "Optional baseline-only cProfile with relative/sanitized source labels; keep counts, total and cumulative times. Profiling time is excluded from benchmark samples.",
+    "trace_129": "Execute distinct search, default, relevance and context/material modes on an independently authored development cue; record selection method/ranking flag, revisions, context resource roles and omissions. This tests merged #129 delivery code paths, not historical #128 task efficacy.",
+    "integrity": "Mutation/stale-index/dependency/path/checkout/revision regression tests belong to runtime patch verification; evaluator parity does not replace them."
+  },
+  "verdict": "FAIL for observed parity/metric/performance-gate failure; INCONCLUSIVE for missing required measurements or unequal/unverified corpus/source; PASS only for this bounded CPU retrieval campaign after all required evidence and independent review. Merge/issue closure remains with acceptance owner."
+}
diff --git a/evidence/knowledge-retrieval-20261027/results/README.md b/evidence/knowledge-retrieval-20261027/results/README.md
new file mode 100644
index 00000000..540c91af
--- /dev/null
+++ b/evidence/knowledge-retrieval-20261027/results/README.md
@@ -0,0 +1,37 @@
+# Completed bounded retrieval results
+
+The full 24-case quality comparison and 20-pair CPU benchmark passed their
+prospectively frozen gates. This is local offline retrieval evidence;
+human/maintainer acceptance, merge and Issue closure remain separate.
+
+- `quality.json`: 96 outputs per revision exactly equal; 20 partial-labelled
+  positives and four negative probes. Existing #129 relevance-item behavior
+  was preserved, not introduced or improved by this manifest patch.
+- `benchmark.json`: 20 balanced groups, 640 API calls and 160 CLI processes;
+  hot relevance paired-median reduction 40.0068267%, 95% interval
+  [38.3789227%, 41.0592071%]. All 400 output pairs and 80 stdout pairs equal;
+  final source/corpus identities unchanged.
+- `performance.md` and `latency-summary.json`: complete per-mode/query p50/p95
+  display, including regressions. CLI relevance decode p95 rose
+  1.810 to 3.046 ms; no cause or blanket tail improvement is inferred.
+- `context-delivery-diagnostic.json`: post-hoc API-response context presence
+  for 19/20 labelled owners, separate from the frozen item-ranking metric.
+- `baseline-profile.json` and `baseline-trace-129.json`: separate source/path
+  diagnostics, excluded from performance samples.
+- `validation-summary.md`, `environment.md`, `independent-review.md`: local
+  failed-suite/baseline reproduction, package checks, hardware/interpreter
+  limits and independent read-only review with unknown reviewer model/effort.
+- `architecture/`: editable EN/ZH SVGs, PNG previews, source, layout checks
+  and reproducible source notes; these diagrams establish software contracts.
+
+Files named `dry-run-*` are baseline-v-baseline mechanics only and explicitly
+INCONCLUSIVE. Their two cases/two paired groups are not pooled into formal
+results. Validator synthetic selfchecks include optimized Python and missing
+measurement cases; they are not retrieval performance samples.
+
+The whole Windows pytest attempt was not green. All initial failures/errors
+were reproduced on baseline; focused manifest/inventory checks, isolated
+package construction and wheel-only source/output parity checks passed.
+Remote CI is pending at PR creation. No model, GPU, service, production-query
+or agent-task experiment ran, and partial AI-authored source labels are not
+human gold or complete Recall.
diff --git a/evidence/knowledge-retrieval-20261027/results/architecture/SOURCE-NOTES.md b/evidence/knowledge-retrieval-20261027/results/architecture/SOURCE-NOTES.md
new file mode 100644
index 00000000..bc6a14de
--- /dev/null
+++ b/evidence/knowledge-retrieval-20261027/results/architecture/SOURCE-NOTES.md
@@ -0,0 +1,42 @@
+# InferRSI knowledge architecture diagrams
+
+Original SVG author configuration: **gpt-6.1-sol / xhigh**, explicitly configured for this delegated diagram task. The SVGs use editable text and vector shapes. No unknown-author draft or code was reused. The supplied reference image informed only the visual treatment: four horizontal panels, light backgrounds, green retrieval panel, blue contracts, solid data-flow arrows, dashed reference/review lines and an evidence strip.
+
+Evidence baseline: the explicitly selected baseline checkout at commit `4817f92121df73fd9cd5fb4a319e3c56a68b4a5b`. All diagram files and the rendering script stay in the selected output directory; the baseline is read-only.
+
+Content sources read:
+
+- `docs/knowledge-base.md`: canonical ownership, authority boundary, registry versus generated catalog, matrices, frozen evidence, explicit use.
+- `docs/knowledge-operations.md`: default Collect, independent relevance/context opt-ins, exact revisions, complete mandatory resources and byte accounting.
+- `docs/knowledge-store.md`: explicitly authorized local store, candidates only, immutable history, reviewed repository maintenance.
+- `src/inferrsi/catalog.py`: legacy catalog search tokenization, filtering, occurrence-count ordering and ID tie-break.
+- `src/inferrsi/knowledge_operations.py`: Bundle revision, package/checkout/store identity rules, record.json material, API filters, limits and omission behavior.
+- `src/inferrsi/_skill_selection.py`: generated packaged skill validation, deterministic lexical weights, complete-file context, licenses/notices and optional resource selection.
+- `tools/build_skill_index.py`: stable registry input, fixed committed source snapshot, source-file hashes and generated `kind=knowledge` catalog records.
+
+Interpretation limits:
+
+- Canonical skill content is organized under `hardware/`, `frameworks/`, `repositories/`, `methods/` and `workflows/`. These ownership categories are distinct from Skill/Knowledge/Case kinds, contribution/dependency/peer relationships, and backend compatibility qualification. Applicability is a metadata field.
+- This is an architecture/contract diagram, not a count of skills or proof of task performance, hardware compatibility, licensing clearance or empirical validation.
+- `knowledge search` and Collect remain separate retrieval modes. Collect's metadata filters are labeled as Reader API features; the figure does not imply that every API filter has a corresponding Collect CLI flag.
+- Generated packaged skill records retain `kind=knowledge`. Packaged `material` returns the record entrypoint (`record.json`); the separate context view supplies canonical SKILL.md and complete mandatory references within an explicit UTF-8 resource-body byte budget.
+- If mandatory context resources exceed the context byte budget, that owner's context is omitted. The selected item can still be returned in `items`; context omission is not item omission.
+- The projection is not an editable content authority, and the stable-ID registry is not the generated catalog. Source commits, local Bundle revisions and tested engine revisions have different meanings.
+- Local store writes do not mutate package/checkouts, evidence or matrices, and they do not promote candidates. The bottom dashed connections describe references/review, not automatic feedback or execution.
+- A performance implementation that reuses digest/path work inside one manifest does not add a new database, persistent cache or scoring method; it does not change this architecture.
+
+From this architecture-artifact directory, rebuild with the existing bundled Python runtime:
+
+```powershell
+& '<bundled-python>' './build_diagrams.py'
+```
+
+Pillow checks text widths and card heights; SVG XML parsing checks structure; CairoSVG renders 3200 × 2100 PNG previews from the 1600 × 1050 vector source. Fonts are Segoe UI (English) and Microsoft YaHei with Noto Sans SC fallback (Chinese). Text remains editable in SVG; preview rendering uses installed system fonts.
+
+Independent review clarified the English fourth-column budget line to
+“Required set over budget: omit context.” Codex `gpt-6.1-sol` / `xhigh` applied
+that one-line source correction, regenerated the assets/layout checks and
+visually inspected the English PNG. The Chinese text already expressed the
+required-set boundary and was unchanged. This states the context omission
+contract without implying that optional-resource overflow omits the whole
+context.
diff --git a/evidence/knowledge-retrieval-20261027/results/architecture/build_diagrams.py b/evidence/knowledge-retrieval-20261027/results/architecture/build_diagrams.py
new file mode 100644
index 00000000..27c1b601
--- /dev/null
+++ b/evidence/knowledge-retrieval-20261027/results/architecture/build_diagrams.py
@@ -0,0 +1,239 @@
+"""Original editable InferRSI knowledge diagrams, based on the pinned audit tree.
+
+Author configuration: gpt-6.1-sol / xhigh, explicitly configured for this delegated task.
+Run with the bundled Python runtime. Requires existing Pillow and CairoSVG.
+All output is confined to this script's directory.
+"""
+from __future__ import annotations
+
+from dataclasses import dataclass
+from html import escape
+from pathlib import Path
+import json
+import xml.etree.ElementTree as ET
+
+import cairosvg
+from PIL import Image, ImageFont
+
+OUT = Path(__file__).resolve().parent
+WIDTH, HEIGHT = 1600, 1050
+BASELINE = "4817f92121df73fd9cd5fb4a319e3c56a68b4a5b"
+PALETTE = {
+    "ink": "#1e293b", "body": "#526782", "muted": "#7c8da8",
+    "panel": "#f6f8fc", "panel_stroke": "#dce5f1",
+    "card": "#ffffff", "card_stroke": "#c5d3e8",
+    "blue": "#5482ff", "blue_fill": "#edf3ff", "blue_ink": "#2149bc",
+    "green": "#37ad70", "green_fill": "#effaf4", "green_ink": "#238051",
+    "purple": "#966bd7", "line": "#a9b8cc",
+}
+
+
+@dataclass
+class Card:
+    y: int
+    h: int
+    title: str
+    lines: list[str]
+    style: str = "plain"
+    line_height: int = 20
+
+
+EN = {
+    "title": "InferRSI Knowledge Architecture",
+    "subtitle": "Canonical sources → build & authorities → retrieval → exact reads & explicit use",
+    "top_note": "K0/K1 foundation · K2 operations",
+    "heads": ["1 · CANONICAL SOURCES", "2 · BUILD & AUTHORITIES", "3 · RETRIEVAL", "4 · READ & USE"],
+    "cards": [
+        [
+            Card(185, 125, "Complete skill library", ["SKILL.md: name + procedure", "metadata.json: state + applicability", "references/ + cases/: owner-local", "Licenses and source inventories"]),
+            Card(326, 120, "Imported catalog sources", ["repo_packs/ + knowledge_packs/", "Existing IDs, tiers, scope and sources", "Contribution · dependency · peer"]),
+            Card(462, 147, "Typed source identities", ["Git commit + inspected paths", "Paper version · document snapshot", "Internal run + artifact identity", "Source pin ≠ content revision", "Frozen evidence stays at its owner"]),
+            Card(625, 129, "Ownership and qualification", ["hardware/ · frameworks/", "repositories/ · methods/ · workflows/", "State ≠ validation ≠ permission", "Backend scope ≠ compatibility", "Unknown provenance/licenses retained"], "dashed", line_height=17),
+        ],
+        [
+            Card(185, 103, "Stable-ID registry", ["skill-index.json", "Skill IDs → paths + backend", "Separate from generated records"]),
+            Card(304, 172, "Source-pinned projection", ["tools/build_skill_index.py", "Fixed source commit + per-file SHA256", "inferrsi.<backend>.skills.*", "kind=knowledge · packaged-skill index", "Imported state belongs to the record", "Projection ≠ content authority"], "blue"),
+            Card(492, 145, "Explicit authorities", ["package: installed bundle, read-only", "checkout: selected root, read-only", "store: authorized local candidates", "Create / Update retain immutable history", "Adaptations need a distinct derived ID"]),
+            Card(653, 101, "Bundle identity contract", ["Qualified ref = <authority>:<id>", "Metadata + content + resources", "+ dependency revisions → SHA256"], "blue"),
+        ],
+        [
+            Card(185, 125, "Legacy: knowledge search", ["Catalog JSON discovery", "AND substring tokens", "Occurrence count; ties by ID", "Existing repo/backend/tier filters"]),
+            Card(326, 145, "Default: Collect", ["Query + metadata filters (reader API)", "Deterministic qualified-identity order", "IDs · revisions · limit · max_bytes", "No-match / omissions / collisions explicit", "Packaged material → record.json"], "green"),
+            Card(487, 267, "Two independent opt-ins", ["relevance: separate lexical cue", "Discovery: generated packaged skills", "4 title · 3 desc · 2 tags/applicability", "1 full core; score ↓, then qualified ref", "Ranking → item and byte limits", "Explicit IDs keep members + order", "context_bytes: explicit byte budget", "Whole-file core + required references", "Optional owner files selected by cue", "Stale projection → index_unavailable"], "green"),
+        ],
+        [
+            Card(185, 125, "Exact-revision access", ["read(ref, revision) / resource(...)", "Full dependency-closed Bundle", "Ambiguous identity is an error", "Unavailable old bytes never use latest"]),
+            Card(326, 187, "Complete context contract", ["Full canonical SKILL.md", "+ declared refs + provenance readsets", "+ required licenses and notices", "Exact UTF-8 resource-body bytes", "Required set over budget: omit context", "Mandatory instructions never truncated", "Other links still need exact reads"], "blue"),
+            Card(529, 101, "Inspectable returned material", ["Ref + revision + paths + SHA256", "Sources · attribution · limitations", "Whole resources + explicit omissions"]),
+            Card(646, 108, "Agent chooses explicit use", ["Goal + authorization + applicability", "Read before acting; honor limitations", "Reading never executes or promotes"], "dashed"),
+        ],
+    ],
+    "evidence_head": "5 · RETAINED EVIDENCE & REVIEWED MAINTENANCE",
+    "bottom": [
+        ("Frozen evidence", "Plans · results · patches · validators"),
+        ("Reference Matrix", "Sources · inspected scope · extraction"),
+        ("SearchSpace Matrix", "Resource × optimization-stage coverage"),
+        ("Reviewed repository changes", "Content + matrices + projection refresh"),
+    ],
+    "reference_note": "Frozen evidence: read-only references.",
+    "review_note": "Reviewed content + matrices; store candidates are not repository coverage.",
+    "legend": ["Contract", "Retrieval", "Data flow", "Reference / review"],
+    "footer": "InferRSI · source 4817f92121df · no automatic execution or promotion",
+}
+
+ZH = {
+    "title": "InferRSI 知识库架构",
+    "subtitle": "规范内容源 → 构建与权威边界 → 检索 → 精确修订读取与显式使用",
+    "top_note": "K0/K1 内容基础 · K2 读写操作",
+    "heads": ["1 · 规范内容源", "2 · 构建与权威边界", "3 · 检索", "4 · 读取与使用"],
+    "cards": [
+        [
+            Card(185, 125, "完整技能库", ["SKILL.md：名称、描述与完整步骤", "metadata.json：溯源、状态与适用性", "references/ 与 cases/：归属技能本地", "保留许可证与来源清单"]),
+            Card(326, 120, "导入的目录来源", ["repo_packs/ + knowledge_packs/", "保留既有 ID、层级、范围与来源", "贡献对象 · 依赖对象 · 同类参考"]),
+            Card(462, 147, "按类型保存来源身份", ["Git 提交与已检查路径", "论文版本、文档版本或快照", "内部运行与产物身份", "来源版本 ≠ 本地内容修订", "原始证据保留于冻结产物的所有者"]),
+            Card(625, 129, "内容归属与验证边界", ["硬件 / 框架 / 仓库 / 方法 / 工作流", "状态 ≠ 实测验证 ≠ 使用授权", "后端适用范围 ≠ 兼容性认证", "未知来源与许可证保持未知"], "dashed"),
+        ],
+        [
+            Card(185, 103, "稳定 ID 注册表", ["skill-index.json", "技能 ID → 分类路径与 backend", "与生成的目录记录是不同对象"]),
+            Card(304, 172, "固定来源的生成投影", ["tools/build_skill_index.py", "固定来源提交与每个文件的 SHA256", "inferrsi.<backend>.skills.*", "kind=knowledge；packaged-skill 索引", "imported 描述目录记录自身状态", "生成投影 ≠ 内容权威"], "blue"),
+            Card(492, 145, "显式选择三类权威", ["package：已安装内容，只读", "checkout：指定技能根目录，只读", "store：获授权的本地候选内容", "Create / Update 保留不可变历史", "改编须使用不同 ID 并记录派生来源"]),
+            Card(653, 101, "完整包身份契约", ["限定引用：<authority>:<id>", "元数据、正文、资源字节", "+ 依赖修订 → SHA256 内容修订"], "blue"),
+        ],
+        [
+            Card(185, 125, "既有入口：knowledge search", ["对目录 JSON 进行发现检索", "AND 子串匹配", "按出现次数排序，再按 ID 排序", "保留 repo/backend/tier 等筛选"]),
+            Card(326, 145, "默认入口：Collect", ["查询与元数据筛选（Reader API）", "默认按限定身份确定性排序", "ID、修订、条数与 max_bytes 上限", "显式报告无匹配、遗漏和身份冲突", "包目录的 material 返回 record.json"], "green"),
+            Card(487, 267, "两项相互独立的可选能力", ["relevance：独立的词法提示", "发现范围仅限生成的已打包技能", "标题 4、描述 3、标签/适用范围 2", "完整核心 1；分数降序后按限定引用", "排序后施加条数与字节上限", "显式 ID 保留成员与调用者顺序", "context_bytes：显式字节预算", "完整核心与强制引用按整文件返回", "可选本地资源按提示选择", "投影过期 → index_unavailable"], "green"),
+        ],
+        [
+            Card(185, 125, "读取精确内容修订", ["read(ref, revision) / resource(...)", "读取依赖闭合的完整 Bundle", "未限定的歧义身份直接报错", "旧字节不可用时不替换为最新内容"]),
+            Card(326, 187, "完整上下文契约", ["完整的规范 SKILL.md", "+ 已声明引用与来源检查材料", "+ 强制保留的许可证和通知", "按 UTF-8 资源正文字节计量", "必需资源超预算：仅省略上下文", "强制指令从不截断", "其他链接仍须按精确修订继续读取"], "blue"),
+            Card(529, 101, "可检查的返回材料", ["引用、修订、路径与 SHA256", "来源、署名和适用限制", "完整资源及显式遗漏原因"]),
+            Card(646, 108, "Agent 显式决定使用", ["结合目标、授权与适用条件", "先读取，再按适用限制行动", "读取不会执行内容或提升候选状态"], "dashed"),
+        ],
+    ],
+    "evidence_head": "5 · 冻结证据与经过审查的维护",
+    "bottom": [
+        ("冻结的原始证据", "计划、结果、补丁与验证器"),
+        ("Reference Matrix", "来源、已检查范围与提取内容"),
+        ("SearchSpace Matrix", "资源 × 优化阶段的覆盖情况"),
+        ("经过审查的仓库变更", "规范内容、两个矩阵与投影刷新"),
+    ],
+    "reference_note": "冻结证据只作引用；运行时不改写。",
+    "review_note": "规范内容与双矩阵经审查后变更；store 候选不等于仓库覆盖。",
+    "legend": ["契约", "检索", "数据流", "引用 / 审查"],
+    "footer": "InferRSI · 来源 4817f92121df · 不自动执行或提升候选状态",
+}
+
+
+def make_svg(locale: str, copy: dict) -> tuple[str, dict]:
+    chinese = locale == "zh-CN"
+    family = "Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" if chinese else "Segoe UI, Arial, sans-serif"
+    normal = ImageFont.truetype("C:/Windows/Fonts/msyh.ttc" if chinese else "C:/Windows/Fonts/segoeui.ttf", 15)
+    bold = ImageFont.truetype("C:/Windows/Fonts/msyhbd.ttc" if chinese else "C:/Windows/Fonts/segoeuib.ttf", 16)
+    pieces: list[str] = []
+    audit = {"locale": locale, "width": WIDTH, "height": HEIGHT, "baseline": BASELINE,
+             "author_model": "gpt-6.1-sol", "author_reasoning_effort": "xhigh", "text_checks": []}
+
+    def rect(x, y, w, h, fill, stroke, radius=9, dashed=False):
+        dash = ' stroke-dasharray="4 4"' if dashed else ''
+        pieces.append(f'<rect x="{x}" y="{y}" width="{w}" height="{h}" rx="{radius}" fill="{fill}" stroke="{stroke}" stroke-width="1.25"{dash}/>')
+
+    def text(x, y, value, size=15, color=None, weight=400, anchor="start", spacing=None):
+        ls = f' letter-spacing="{spacing}"' if spacing else ''
+        pieces.append(f'<text x="{x}" y="{y}" font-family="{family}" font-size="{size}" font-weight="{weight}" text-anchor="{anchor}" fill="{color or PALETTE["body"]}"{ls}>{escape(value)}</text>')
+
+    def path(d, color, dashed=False, marker=None, width=1.5):
+        dash = ' stroke-dasharray="5 5"' if dashed else ''
+        end = f' marker-end="url(#{marker})"' if marker else ''
+        pieces.append(f'<path d="{d}" fill="none" stroke="{color}" stroke-width="{width}"{dash}{end}/>')
+
+    def card(x, w, item: Card):
+        special = item.style
+        stroke = PALETTE["blue"] if special == "blue" else PALETTE["green"] if special == "green" else PALETTE["card_stroke"]
+        fill = PALETTE["blue_fill"] if special == "blue" else "#ffffff"
+        rect(x, item.y, w, item.h, fill, stroke, dashed=special == "dashed")
+        title_color = PALETTE["blue_ink"] if special == "blue" else PALETTE["green_ink"] if special == "green" else PALETTE["ink"]
+        text(x + 16, item.y + 27, item.title, 16, title_color, 600)
+        title_width = bold.getlength(item.title)
+        assert title_width <= w - 30, (locale, item.title, title_width, w)
+        for i, line in enumerate(item.lines):
+            measured = normal.getlength(line)
+            assert measured <= w - 30, (locale, item.title, line, measured, w)
+            baseline = item.y + 48 + item.line_height * i
+            assert baseline + 4 <= item.y + item.h - 9, (item.title, baseline, item.y + item.h)
+            text(x + 16, baseline, line)
+        audit["text_checks"].append({"title": item.title, "line_count": len(item.lines), "title_width": round(title_width, 2), "available_width": w - 30})
+
+    pieces.append(f'<svg xmlns="http://www.w3.org/2000/svg" width="{WIDTH}" height="{HEIGHT}" viewBox="0 0 {WIDTH} {HEIGHT}" role="img" aria-labelledby="title desc" xml:lang="{locale}">')
+    pieces.append(f'<title id="title">{escape(copy["title"])}</title>')
+    pieces.append('<desc id="desc">Canonical sources feed source-pinned generated projections and explicit read/write authorities. Legacy search, default Collect and optional lexical selection/context remain separate. Exact-revision reads return complete required resources; the agent decides authorized use. Frozen evidence and reviewed matrices retain their ownership.</desc>')
+    pieces.append('<defs><marker id="flow" markerWidth="9" markerHeight="9" refX="8" refY="4.5" orient="auto"><path d="M0,0 L9,4.5 L0,9 Z" fill="#5482ff"/></marker><marker id="reference" markerWidth="8" markerHeight="8" refX="7" refY="4" orient="auto"><path d="M0,0 L8,4 L0,8 Z" fill="#966bd7"/></marker><marker id="review" markerWidth="8" markerHeight="8" refX="7" refY="4" orient="auto"><path d="M0,0 L8,4 L0,8 Z" fill="#7c8da8"/></marker></defs>')
+    pieces.append('<rect width="1600" height="1050" fill="#ffffff"/>')
+    text(56, 61, copy["title"], 31, PALETTE["ink"], 650)
+    text(56, 94, copy["subtitle"], 16)
+    text(1544, 59, copy["top_note"], 14, PALETTE["muted"], anchor="end")
+    path("M56 113 H1544", PALETTE["panel_stroke"], width=1.2)
+
+    columns = [(56, 330), (414, 362), (804, 358), (1190, 354)]
+    for index, ((x, w), heading, cards) in enumerate(zip(columns, copy["heads"], copy["cards"])):
+        pieces.append(f'<g id="column-{index + 1}">')
+        rect(x, 134, w, 642, PALETTE["green_fill"] if index == 2 else PALETTE["panel"], "#b2e5c8" if index == 2 else PALETTE["panel_stroke"], radius=12)
+        text(x + 20, 164, heading, 14, PALETTE["green_ink"] if index == 2 else PALETTE["muted"], 600, spacing=None if chinese else 1.1)
+        for item in cards:
+            card(x + 20, w - 40, item)
+        pieces.append('</g>')
+
+    for start, end in [(386, 414), (776, 804), (1162, 1190)]:
+        path(f"M{start + 2} 423 H{end - 4}", PALETTE["blue"], marker="flow", width=1.7)
+
+    path("M232 884 V821 H216 V778", PALETTE["purple"], dashed=True, marker="reference")
+    text(257, 826, copy["reference_note"], 15, PALETTE["purple"])
+    path("M1366 884 V856 H594 V778", PALETTE["muted"], dashed=True, marker="review")
+    text(660, 848, copy["review_note"], 15, PALETTE["muted"])
+
+    pieces.append('<g id="evidence-and-maintenance">')
+    rect(56, 884, 1488, 118, "#fafbfe", PALETTE["panel_stroke"], radius=12)
+    text(76, 912, copy["evidence_head"], 14, PALETTE["muted"], 600, spacing=None if chinese else 1.1)
+    for x, w, (title, detail) in zip([76, 436, 796, 1156], [344, 344, 344, 368], copy["bottom"]):
+        rect(x, 928, w, 56, "#ffffff", PALETTE["card_stroke"], radius=8)
+        text(x + 15, 950, title, 16, PALETTE["ink"], 600)
+        text(x + 15, 972, detail, 14)
+    pieces.append('</g>')
+
+    rect(56, 1021, 14, 11, PALETTE["blue_fill"], PALETTE["blue"], radius=2)
+    text(80, 1032, copy["legend"][0], 14)
+    rect(184, 1021, 14, 11, PALETTE["green_fill"], PALETTE["green"], radius=2)
+    text(208, 1032, copy["legend"][1], 14)
+    path("M320 1027 H354", PALETTE["blue"], width=1.7)
+    text(365, 1032, copy["legend"][2], 14)
+    path("M482 1027 H516", PALETTE["purple"], dashed=True)
+    text(527, 1032, copy["legend"][3], 14)
+    text(1544, 1032, copy["footer"], 14, PALETTE["muted"], anchor="end")
+    pieces.append('</svg>')
+    svg = "\n".join(pieces) + "\n"
+    ET.fromstring(svg)
+    return svg, audit
+
+
+def main():
+    audits = []
+    for locale, copy in [("en", EN), ("zh-CN", ZH)]:
+        svg, audit = make_svg(locale, copy)
+        stem = OUT / f"inferrsi-knowledge-architecture.{locale}"
+        svg_path = stem.with_suffix(f".{locale}.svg")
+        png_path = stem.with_suffix(f".{locale}.png")
+        svg_path.write_text(svg, encoding="utf-8", newline="\n")
+        cairosvg.svg2png(bytestring=svg.encode(), write_to=str(png_path), output_width=3200, output_height=2100)
+        with Image.open(png_path) as rendered:
+            assert rendered.size == (3200, 2100)
+            assert rendered.getpixel((0, 0))[:3] == (255, 255, 255)
+        audit["svg"] = svg_path.name
+        audit["png"] = png_path.name
+        audit["render_size"] = [3200, 2100]
+        audits.append(audit)
+        print(f"Created {svg_path.name} and {png_path.name}")
+    (OUT / "layout-checks.json").write_text(json.dumps(audits, ensure_ascii=False, indent=2) + "\n", encoding="utf-8")
+
+
+if __name__ == "__main__":
+    main()
diff --git a/evidence/knowledge-retrieval-20261027/results/architecture/inferrsi-knowledge-architecture.en.png b/evidence/knowledge-retrieval-20261027/results/architecture/inferrsi-knowledge-architecture.en.png
new file mode 100644
index 00000000..2bfee9da
Binary files /dev/null and b/evidence/knowledge-retrieval-20261027/results/architecture/inferrsi-knowledge-architecture.en.png differ
diff --git a/evidence/knowledge-retrieval-20261027/results/architecture/inferrsi-knowledge-architecture.en.svg b/evidence/knowledge-retrieval-20261027/results/architecture/inferrsi-knowledge-architecture.en.svg
new file mode 100644
index 00000000..187773c7
--- /dev/null
+++ b/evidence/knowledge-retrieval-20261027/results/architecture/inferrsi-knowledge-architecture.en.svg
@@ -0,0 +1,158 @@
+<svg xmlns="http://www.w3.org/2000/svg" width="1600" height="1050" viewBox="0 0 1600 1050" role="img" aria-labelledby="title desc" xml:lang="en">
+<title id="title">InferRSI Knowledge Architecture</title>
+<desc id="desc">Canonical sources feed source-pinned generated projections and explicit read/write authorities. Legacy search, default Collect and optional lexical selection/context remain separate. Exact-revision reads return complete required resources; the agent decides authorized use. Frozen evidence and reviewed matrices retain their ownership.</desc>
+<defs><marker id="flow" markerWidth="9" markerHeight="9" refX="8" refY="4.5" orient="auto"><path d="M0,0 L9,4.5 L0,9 Z" fill="#5482ff"/></marker><marker id="reference" markerWidth="8" markerHeight="8" refX="7" refY="4" orient="auto"><path d="M0,0 L8,4 L0,8 Z" fill="#966bd7"/></marker><marker id="review" markerWidth="8" markerHeight="8" refX="7" refY="4" orient="auto"><path d="M0,0 L8,4 L0,8 Z" fill="#7c8da8"/></marker></defs>
+<rect width="1600" height="1050" fill="#ffffff"/>
+<text x="56" y="61" font-family="Segoe UI, Arial, sans-serif" font-size="31" font-weight="650" text-anchor="start" fill="#1e293b">InferRSI Knowledge Architecture</text>
+<text x="56" y="94" font-family="Segoe UI, Arial, sans-serif" font-size="16" font-weight="400" text-anchor="start" fill="#526782">Canonical sources → build &amp; authorities → retrieval → exact reads &amp; explicit use</text>
+<text x="1544" y="59" font-family="Segoe UI, Arial, sans-serif" font-size="14" font-weight="400" text-anchor="end" fill="#7c8da8">K0/K1 foundation · K2 operations</text>
+<path d="M56 113 H1544" fill="none" stroke="#dce5f1" stroke-width="1.2"/>
+<g id="column-1">
+<rect x="56" y="134" width="330" height="642" rx="12" fill="#f6f8fc" stroke="#dce5f1" stroke-width="1.25"/>
+<text x="76" y="164" font-family="Segoe UI, Arial, sans-serif" font-size="14" font-weight="600" text-anchor="start" fill="#7c8da8" letter-spacing="1.1">1 · CANONICAL SOURCES</text>
+<rect x="76" y="185" width="290" height="125" rx="9" fill="#ffffff" stroke="#c5d3e8" stroke-width="1.25"/>
+<text x="92" y="212" font-family="Segoe UI, Arial, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#1e293b">Complete skill library</text>
+<text x="92" y="233" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">SKILL.md: name + procedure</text>
+<text x="92" y="253" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">metadata.json: state + applicability</text>
+<text x="92" y="273" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">references/ + cases/: owner-local</text>
+<text x="92" y="293" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Licenses and source inventories</text>
+<rect x="76" y="326" width="290" height="120" rx="9" fill="#ffffff" stroke="#c5d3e8" stroke-width="1.25"/>
+<text x="92" y="353" font-family="Segoe UI, Arial, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#1e293b">Imported catalog sources</text>
+<text x="92" y="374" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">repo_packs/ + knowledge_packs/</text>
+<text x="92" y="394" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Existing IDs, tiers, scope and sources</text>
+<text x="92" y="414" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Contribution · dependency · peer</text>
+<rect x="76" y="462" width="290" height="147" rx="9" fill="#ffffff" stroke="#c5d3e8" stroke-width="1.25"/>
+<text x="92" y="489" font-family="Segoe UI, Arial, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#1e293b">Typed source identities</text>
+<text x="92" y="510" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Git commit + inspected paths</text>
+<text x="92" y="530" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Paper version · document snapshot</text>
+<text x="92" y="550" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Internal run + artifact identity</text>
+<text x="92" y="570" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Source pin ≠ content revision</text>
+<text x="92" y="590" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Frozen evidence stays at its owner</text>
+<rect x="76" y="625" width="290" height="129" rx="9" fill="#ffffff" stroke="#c5d3e8" stroke-width="1.25" stroke-dasharray="4 4"/>
+<text x="92" y="652" font-family="Segoe UI, Arial, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#1e293b">Ownership and qualification</text>
+<text x="92" y="673" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">hardware/ · frameworks/</text>
+<text x="92" y="690" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">repositories/ · methods/ · workflows/</text>
+<text x="92" y="707" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">State ≠ validation ≠ permission</text>
+<text x="92" y="724" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Backend scope ≠ compatibility</text>
+<text x="92" y="741" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Unknown provenance/licenses retained</text>
+</g>
+<g id="column-2">
+<rect x="414" y="134" width="362" height="642" rx="12" fill="#f6f8fc" stroke="#dce5f1" stroke-width="1.25"/>
+<text x="434" y="164" font-family="Segoe UI, Arial, sans-serif" font-size="14" font-weight="600" text-anchor="start" fill="#7c8da8" letter-spacing="1.1">2 · BUILD &amp; AUTHORITIES</text>
+<rect x="434" y="185" width="322" height="103" rx="9" fill="#ffffff" stroke="#c5d3e8" stroke-width="1.25"/>
+<text x="450" y="212" font-family="Segoe UI, Arial, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#1e293b">Stable-ID registry</text>
+<text x="450" y="233" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">skill-index.json</text>
+<text x="450" y="253" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Skill IDs → paths + backend</text>
+<text x="450" y="273" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Separate from generated records</text>
+<rect x="434" y="304" width="322" height="172" rx="9" fill="#edf3ff" stroke="#5482ff" stroke-width="1.25"/>
+<text x="450" y="331" font-family="Segoe UI, Arial, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#2149bc">Source-pinned projection</text>
+<text x="450" y="352" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">tools/build_skill_index.py</text>
+<text x="450" y="372" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Fixed source commit + per-file SHA256</text>
+<text x="450" y="392" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">inferrsi.&lt;backend&gt;.skills.*</text>
+<text x="450" y="412" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">kind=knowledge · packaged-skill index</text>
+<text x="450" y="432" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Imported state belongs to the record</text>
+<text x="450" y="452" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Projection ≠ content authority</text>
+<rect x="434" y="492" width="322" height="145" rx="9" fill="#ffffff" stroke="#c5d3e8" stroke-width="1.25"/>
+<text x="450" y="519" font-family="Segoe UI, Arial, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#1e293b">Explicit authorities</text>
+<text x="450" y="540" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">package: installed bundle, read-only</text>
+<text x="450" y="560" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">checkout: selected root, read-only</text>
+<text x="450" y="580" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">store: authorized local candidates</text>
+<text x="450" y="600" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Create / Update retain immutable history</text>
+<text x="450" y="620" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Adaptations need a distinct derived ID</text>
+<rect x="434" y="653" width="322" height="101" rx="9" fill="#edf3ff" stroke="#5482ff" stroke-width="1.25"/>
+<text x="450" y="680" font-family="Segoe UI, Arial, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#2149bc">Bundle identity contract</text>
+<text x="450" y="701" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Qualified ref = &lt;authority&gt;:&lt;id&gt;</text>
+<text x="450" y="721" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Metadata + content + resources</text>
+<text x="450" y="741" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">+ dependency revisions → SHA256</text>
+</g>
+<g id="column-3">
+<rect x="804" y="134" width="358" height="642" rx="12" fill="#effaf4" stroke="#b2e5c8" stroke-width="1.25"/>
+<text x="824" y="164" font-family="Segoe UI, Arial, sans-serif" font-size="14" font-weight="600" text-anchor="start" fill="#238051" letter-spacing="1.1">3 · RETRIEVAL</text>
+<rect x="824" y="185" width="318" height="125" rx="9" fill="#ffffff" stroke="#c5d3e8" stroke-width="1.25"/>
+<text x="840" y="212" font-family="Segoe UI, Arial, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#1e293b">Legacy: knowledge search</text>
+<text x="840" y="233" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Catalog JSON discovery</text>
+<text x="840" y="253" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">AND substring tokens</text>
+<text x="840" y="273" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Occurrence count; ties by ID</text>
+<text x="840" y="293" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Existing repo/backend/tier filters</text>
+<rect x="824" y="326" width="318" height="145" rx="9" fill="#ffffff" stroke="#37ad70" stroke-width="1.25"/>
+<text x="840" y="353" font-family="Segoe UI, Arial, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#238051">Default: Collect</text>
+<text x="840" y="374" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Query + metadata filters (reader API)</text>
+<text x="840" y="394" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Deterministic qualified-identity order</text>
+<text x="840" y="414" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">IDs · revisions · limit · max_bytes</text>
+<text x="840" y="434" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">No-match / omissions / collisions explicit</text>
+<text x="840" y="454" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Packaged material → record.json</text>
+<rect x="824" y="487" width="318" height="267" rx="9" fill="#ffffff" stroke="#37ad70" stroke-width="1.25"/>
+<text x="840" y="514" font-family="Segoe UI, Arial, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#238051">Two independent opt-ins</text>
+<text x="840" y="535" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">relevance: separate lexical cue</text>
+<text x="840" y="555" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Discovery: generated packaged skills</text>
+<text x="840" y="575" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">4 title · 3 desc · 2 tags/applicability</text>
+<text x="840" y="595" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">1 full core; score ↓, then qualified ref</text>
+<text x="840" y="615" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Ranking → item and byte limits</text>
+<text x="840" y="635" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Explicit IDs keep members + order</text>
+<text x="840" y="655" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">context_bytes: explicit byte budget</text>
+<text x="840" y="675" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Whole-file core + required references</text>
+<text x="840" y="695" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Optional owner files selected by cue</text>
+<text x="840" y="715" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Stale projection → index_unavailable</text>
+</g>
+<g id="column-4">
+<rect x="1190" y="134" width="354" height="642" rx="12" fill="#f6f8fc" stroke="#dce5f1" stroke-width="1.25"/>
+<text x="1210" y="164" font-family="Segoe UI, Arial, sans-serif" font-size="14" font-weight="600" text-anchor="start" fill="#7c8da8" letter-spacing="1.1">4 · READ &amp; USE</text>
+<rect x="1210" y="185" width="314" height="125" rx="9" fill="#ffffff" stroke="#c5d3e8" stroke-width="1.25"/>
+<text x="1226" y="212" font-family="Segoe UI, Arial, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#1e293b">Exact-revision access</text>
+<text x="1226" y="233" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">read(ref, revision) / resource(...)</text>
+<text x="1226" y="253" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Full dependency-closed Bundle</text>
+<text x="1226" y="273" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Ambiguous identity is an error</text>
+<text x="1226" y="293" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Unavailable old bytes never use latest</text>
+<rect x="1210" y="326" width="314" height="187" rx="9" fill="#edf3ff" stroke="#5482ff" stroke-width="1.25"/>
+<text x="1226" y="353" font-family="Segoe UI, Arial, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#2149bc">Complete context contract</text>
+<text x="1226" y="374" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Full canonical SKILL.md</text>
+<text x="1226" y="394" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">+ declared refs + provenance readsets</text>
+<text x="1226" y="414" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">+ required licenses and notices</text>
+<text x="1226" y="434" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Exact UTF-8 resource-body bytes</text>
+<text x="1226" y="454" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Required set over budget: omit context</text>
+<text x="1226" y="474" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Mandatory instructions never truncated</text>
+<text x="1226" y="494" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Other links still need exact reads</text>
+<rect x="1210" y="529" width="314" height="101" rx="9" fill="#ffffff" stroke="#c5d3e8" stroke-width="1.25"/>
+<text x="1226" y="556" font-family="Segoe UI, Arial, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#1e293b">Inspectable returned material</text>
+<text x="1226" y="577" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Ref + revision + paths + SHA256</text>
+<text x="1226" y="597" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Sources · attribution · limitations</text>
+<text x="1226" y="617" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Whole resources + explicit omissions</text>
+<rect x="1210" y="646" width="314" height="108" rx="9" fill="#ffffff" stroke="#c5d3e8" stroke-width="1.25" stroke-dasharray="4 4"/>
+<text x="1226" y="673" font-family="Segoe UI, Arial, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#1e293b">Agent chooses explicit use</text>
+<text x="1226" y="694" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Goal + authorization + applicability</text>
+<text x="1226" y="714" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Read before acting; honor limitations</text>
+<text x="1226" y="734" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Reading never executes or promotes</text>
+</g>
+<path d="M388 423 H410" fill="none" stroke="#5482ff" stroke-width="1.7" marker-end="url(#flow)"/>
+<path d="M778 423 H800" fill="none" stroke="#5482ff" stroke-width="1.7" marker-end="url(#flow)"/>
+<path d="M1164 423 H1186" fill="none" stroke="#5482ff" stroke-width="1.7" marker-end="url(#flow)"/>
+<path d="M232 884 V821 H216 V778" fill="none" stroke="#966bd7" stroke-width="1.5" stroke-dasharray="5 5" marker-end="url(#reference)"/>
+<text x="257" y="826" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#966bd7">Frozen evidence: read-only references.</text>
+<path d="M1366 884 V856 H594 V778" fill="none" stroke="#7c8da8" stroke-width="1.5" stroke-dasharray="5 5" marker-end="url(#review)"/>
+<text x="660" y="848" font-family="Segoe UI, Arial, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#7c8da8">Reviewed content + matrices; store candidates are not repository coverage.</text>
+<g id="evidence-and-maintenance">
+<rect x="56" y="884" width="1488" height="118" rx="12" fill="#fafbfe" stroke="#dce5f1" stroke-width="1.25"/>
+<text x="76" y="912" font-family="Segoe UI, Arial, sans-serif" font-size="14" font-weight="600" text-anchor="start" fill="#7c8da8" letter-spacing="1.1">5 · RETAINED EVIDENCE &amp; REVIEWED MAINTENANCE</text>
+<rect x="76" y="928" width="344" height="56" rx="8" fill="#ffffff" stroke="#c5d3e8" stroke-width="1.25"/>
+<text x="91" y="950" font-family="Segoe UI, Arial, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#1e293b">Frozen evidence</text>
+<text x="91" y="972" font-family="Segoe UI, Arial, sans-serif" font-size="14" font-weight="400" text-anchor="start" fill="#526782">Plans · results · patches · validators</text>
+<rect x="436" y="928" width="344" height="56" rx="8" fill="#ffffff" stroke="#c5d3e8" stroke-width="1.25"/>
+<text x="451" y="950" font-family="Segoe UI, Arial, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#1e293b">Reference Matrix</text>
+<text x="451" y="972" font-family="Segoe UI, Arial, sans-serif" font-size="14" font-weight="400" text-anchor="start" fill="#526782">Sources · inspected scope · extraction</text>
+<rect x="796" y="928" width="344" height="56" rx="8" fill="#ffffff" stroke="#c5d3e8" stroke-width="1.25"/>
+<text x="811" y="950" font-family="Segoe UI, Arial, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#1e293b">SearchSpace Matrix</text>
+<text x="811" y="972" font-family="Segoe UI, Arial, sans-serif" font-size="14" font-weight="400" text-anchor="start" fill="#526782">Resource × optimization-stage coverage</text>
+<rect x="1156" y="928" width="368" height="56" rx="8" fill="#ffffff" stroke="#c5d3e8" stroke-width="1.25"/>
+<text x="1171" y="950" font-family="Segoe UI, Arial, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#1e293b">Reviewed repository changes</text>
+<text x="1171" y="972" font-family="Segoe UI, Arial, sans-serif" font-size="14" font-weight="400" text-anchor="start" fill="#526782">Content + matrices + projection refresh</text>
+</g>
+<rect x="56" y="1021" width="14" height="11" rx="2" fill="#edf3ff" stroke="#5482ff" stroke-width="1.25"/>
+<text x="80" y="1032" font-family="Segoe UI, Arial, sans-serif" font-size="14" font-weight="400" text-anchor="start" fill="#526782">Contract</text>
+<rect x="184" y="1021" width="14" height="11" rx="2" fill="#effaf4" stroke="#37ad70" stroke-width="1.25"/>
+<text x="208" y="1032" font-family="Segoe UI, Arial, sans-serif" font-size="14" font-weight="400" text-anchor="start" fill="#526782">Retrieval</text>
+<path d="M320 1027 H354" fill="none" stroke="#5482ff" stroke-width="1.7"/>
+<text x="365" y="1032" font-family="Segoe UI, Arial, sans-serif" font-size="14" font-weight="400" text-anchor="start" fill="#526782">Data flow</text>
+<path d="M482 1027 H516" fill="none" stroke="#966bd7" stroke-width="1.5" stroke-dasharray="5 5"/>
+<text x="527" y="1032" font-family="Segoe UI, Arial, sans-serif" font-size="14" font-weight="400" text-anchor="start" fill="#526782">Reference / review</text>
+<text x="1544" y="1032" font-family="Segoe UI, Arial, sans-serif" font-size="14" font-weight="400" text-anchor="end" fill="#7c8da8">InferRSI · source 4817f92121df · no automatic execution or promotion</text>
+</svg>
diff --git a/evidence/knowledge-retrieval-20261027/results/architecture/inferrsi-knowledge-architecture.zh-CN.png b/evidence/knowledge-retrieval-20261027/results/architecture/inferrsi-knowledge-architecture.zh-CN.png
new file mode 100644
index 00000000..362352b8
Binary files /dev/null and b/evidence/knowledge-retrieval-20261027/results/architecture/inferrsi-knowledge-architecture.zh-CN.png differ
diff --git a/evidence/knowledge-retrieval-20261027/results/architecture/inferrsi-knowledge-architecture.zh-CN.svg b/evidence/knowledge-retrieval-20261027/results/architecture/inferrsi-knowledge-architecture.zh-CN.svg
new file mode 100644
index 00000000..3182564e
--- /dev/null
+++ b/evidence/knowledge-retrieval-20261027/results/architecture/inferrsi-knowledge-architecture.zh-CN.svg
@@ -0,0 +1,157 @@
+<svg xmlns="http://www.w3.org/2000/svg" width="1600" height="1050" viewBox="0 0 1600 1050" role="img" aria-labelledby="title desc" xml:lang="zh-CN">
+<title id="title">InferRSI 知识库架构</title>
+<desc id="desc">Canonical sources feed source-pinned generated projections and explicit read/write authorities. Legacy search, default Collect and optional lexical selection/context remain separate. Exact-revision reads return complete required resources; the agent decides authorized use. Frozen evidence and reviewed matrices retain their ownership.</desc>
+<defs><marker id="flow" markerWidth="9" markerHeight="9" refX="8" refY="4.5" orient="auto"><path d="M0,0 L9,4.5 L0,9 Z" fill="#5482ff"/></marker><marker id="reference" markerWidth="8" markerHeight="8" refX="7" refY="4" orient="auto"><path d="M0,0 L8,4 L0,8 Z" fill="#966bd7"/></marker><marker id="review" markerWidth="8" markerHeight="8" refX="7" refY="4" orient="auto"><path d="M0,0 L8,4 L0,8 Z" fill="#7c8da8"/></marker></defs>
+<rect width="1600" height="1050" fill="#ffffff"/>
+<text x="56" y="61" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="31" font-weight="650" text-anchor="start" fill="#1e293b">InferRSI 知识库架构</text>
+<text x="56" y="94" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="16" font-weight="400" text-anchor="start" fill="#526782">规范内容源 → 构建与权威边界 → 检索 → 精确修订读取与显式使用</text>
+<text x="1544" y="59" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="14" font-weight="400" text-anchor="end" fill="#7c8da8">K0/K1 内容基础 · K2 读写操作</text>
+<path d="M56 113 H1544" fill="none" stroke="#dce5f1" stroke-width="1.2"/>
+<g id="column-1">
+<rect x="56" y="134" width="330" height="642" rx="12" fill="#f6f8fc" stroke="#dce5f1" stroke-width="1.25"/>
+<text x="76" y="164" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="14" font-weight="600" text-anchor="start" fill="#7c8da8">1 · 规范内容源</text>
+<rect x="76" y="185" width="290" height="125" rx="9" fill="#ffffff" stroke="#c5d3e8" stroke-width="1.25"/>
+<text x="92" y="212" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#1e293b">完整技能库</text>
+<text x="92" y="233" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">SKILL.md：名称、描述与完整步骤</text>
+<text x="92" y="253" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">metadata.json：溯源、状态与适用性</text>
+<text x="92" y="273" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">references/ 与 cases/：归属技能本地</text>
+<text x="92" y="293" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">保留许可证与来源清单</text>
+<rect x="76" y="326" width="290" height="120" rx="9" fill="#ffffff" stroke="#c5d3e8" stroke-width="1.25"/>
+<text x="92" y="353" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#1e293b">导入的目录来源</text>
+<text x="92" y="374" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">repo_packs/ + knowledge_packs/</text>
+<text x="92" y="394" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">保留既有 ID、层级、范围与来源</text>
+<text x="92" y="414" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">贡献对象 · 依赖对象 · 同类参考</text>
+<rect x="76" y="462" width="290" height="147" rx="9" fill="#ffffff" stroke="#c5d3e8" stroke-width="1.25"/>
+<text x="92" y="489" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#1e293b">按类型保存来源身份</text>
+<text x="92" y="510" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Git 提交与已检查路径</text>
+<text x="92" y="530" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">论文版本、文档版本或快照</text>
+<text x="92" y="550" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">内部运行与产物身份</text>
+<text x="92" y="570" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">来源版本 ≠ 本地内容修订</text>
+<text x="92" y="590" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">原始证据保留于冻结产物的所有者</text>
+<rect x="76" y="625" width="290" height="129" rx="9" fill="#ffffff" stroke="#c5d3e8" stroke-width="1.25" stroke-dasharray="4 4"/>
+<text x="92" y="652" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#1e293b">内容归属与验证边界</text>
+<text x="92" y="673" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">硬件 / 框架 / 仓库 / 方法 / 工作流</text>
+<text x="92" y="693" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">状态 ≠ 实测验证 ≠ 使用授权</text>
+<text x="92" y="713" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">后端适用范围 ≠ 兼容性认证</text>
+<text x="92" y="733" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">未知来源与许可证保持未知</text>
+</g>
+<g id="column-2">
+<rect x="414" y="134" width="362" height="642" rx="12" fill="#f6f8fc" stroke="#dce5f1" stroke-width="1.25"/>
+<text x="434" y="164" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="14" font-weight="600" text-anchor="start" fill="#7c8da8">2 · 构建与权威边界</text>
+<rect x="434" y="185" width="322" height="103" rx="9" fill="#ffffff" stroke="#c5d3e8" stroke-width="1.25"/>
+<text x="450" y="212" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#1e293b">稳定 ID 注册表</text>
+<text x="450" y="233" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">skill-index.json</text>
+<text x="450" y="253" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">技能 ID → 分类路径与 backend</text>
+<text x="450" y="273" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">与生成的目录记录是不同对象</text>
+<rect x="434" y="304" width="322" height="172" rx="9" fill="#edf3ff" stroke="#5482ff" stroke-width="1.25"/>
+<text x="450" y="331" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#2149bc">固定来源的生成投影</text>
+<text x="450" y="352" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">tools/build_skill_index.py</text>
+<text x="450" y="372" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">固定来源提交与每个文件的 SHA256</text>
+<text x="450" y="392" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">inferrsi.&lt;backend&gt;.skills.*</text>
+<text x="450" y="412" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">kind=knowledge；packaged-skill 索引</text>
+<text x="450" y="432" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">imported 描述目录记录自身状态</text>
+<text x="450" y="452" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">生成投影 ≠ 内容权威</text>
+<rect x="434" y="492" width="322" height="145" rx="9" fill="#ffffff" stroke="#c5d3e8" stroke-width="1.25"/>
+<text x="450" y="519" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#1e293b">显式选择三类权威</text>
+<text x="450" y="540" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">package：已安装内容，只读</text>
+<text x="450" y="560" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">checkout：指定技能根目录，只读</text>
+<text x="450" y="580" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">store：获授权的本地候选内容</text>
+<text x="450" y="600" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">Create / Update 保留不可变历史</text>
+<text x="450" y="620" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">改编须使用不同 ID 并记录派生来源</text>
+<rect x="434" y="653" width="322" height="101" rx="9" fill="#edf3ff" stroke="#5482ff" stroke-width="1.25"/>
+<text x="450" y="680" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#2149bc">完整包身份契约</text>
+<text x="450" y="701" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">限定引用：&lt;authority&gt;:&lt;id&gt;</text>
+<text x="450" y="721" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">元数据、正文、资源字节</text>
+<text x="450" y="741" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">+ 依赖修订 → SHA256 内容修订</text>
+</g>
+<g id="column-3">
+<rect x="804" y="134" width="358" height="642" rx="12" fill="#effaf4" stroke="#b2e5c8" stroke-width="1.25"/>
+<text x="824" y="164" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="14" font-weight="600" text-anchor="start" fill="#238051">3 · 检索</text>
+<rect x="824" y="185" width="318" height="125" rx="9" fill="#ffffff" stroke="#c5d3e8" stroke-width="1.25"/>
+<text x="840" y="212" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#1e293b">既有入口：knowledge search</text>
+<text x="840" y="233" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">对目录 JSON 进行发现检索</text>
+<text x="840" y="253" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">AND 子串匹配</text>
+<text x="840" y="273" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">按出现次数排序，再按 ID 排序</text>
+<text x="840" y="293" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">保留 repo/backend/tier 等筛选</text>
+<rect x="824" y="326" width="318" height="145" rx="9" fill="#ffffff" stroke="#37ad70" stroke-width="1.25"/>
+<text x="840" y="353" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#238051">默认入口：Collect</text>
+<text x="840" y="374" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">查询与元数据筛选（Reader API）</text>
+<text x="840" y="394" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">默认按限定身份确定性排序</text>
+<text x="840" y="414" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">ID、修订、条数与 max_bytes 上限</text>
+<text x="840" y="434" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">显式报告无匹配、遗漏和身份冲突</text>
+<text x="840" y="454" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">包目录的 material 返回 record.json</text>
+<rect x="824" y="487" width="318" height="267" rx="9" fill="#ffffff" stroke="#37ad70" stroke-width="1.25"/>
+<text x="840" y="514" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#238051">两项相互独立的可选能力</text>
+<text x="840" y="535" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">relevance：独立的词法提示</text>
+<text x="840" y="555" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">发现范围仅限生成的已打包技能</text>
+<text x="840" y="575" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">标题 4、描述 3、标签/适用范围 2</text>
+<text x="840" y="595" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">完整核心 1；分数降序后按限定引用</text>
+<text x="840" y="615" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">排序后施加条数与字节上限</text>
+<text x="840" y="635" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">显式 ID 保留成员与调用者顺序</text>
+<text x="840" y="655" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">context_bytes：显式字节预算</text>
+<text x="840" y="675" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">完整核心与强制引用按整文件返回</text>
+<text x="840" y="695" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">可选本地资源按提示选择</text>
+<text x="840" y="715" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">投影过期 → index_unavailable</text>
+</g>
+<g id="column-4">
+<rect x="1190" y="134" width="354" height="642" rx="12" fill="#f6f8fc" stroke="#dce5f1" stroke-width="1.25"/>
+<text x="1210" y="164" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="14" font-weight="600" text-anchor="start" fill="#7c8da8">4 · 读取与使用</text>
+<rect x="1210" y="185" width="314" height="125" rx="9" fill="#ffffff" stroke="#c5d3e8" stroke-width="1.25"/>
+<text x="1226" y="212" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#1e293b">读取精确内容修订</text>
+<text x="1226" y="233" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">read(ref, revision) / resource(...)</text>
+<text x="1226" y="253" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">读取依赖闭合的完整 Bundle</text>
+<text x="1226" y="273" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">未限定的歧义身份直接报错</text>
+<text x="1226" y="293" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">旧字节不可用时不替换为最新内容</text>
+<rect x="1210" y="326" width="314" height="187" rx="9" fill="#edf3ff" stroke="#5482ff" stroke-width="1.25"/>
+<text x="1226" y="353" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#2149bc">完整上下文契约</text>
+<text x="1226" y="374" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">完整的规范 SKILL.md</text>
+<text x="1226" y="394" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">+ 已声明引用与来源检查材料</text>
+<text x="1226" y="414" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">+ 强制保留的许可证和通知</text>
+<text x="1226" y="434" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">按 UTF-8 资源正文字节计量</text>
+<text x="1226" y="454" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">必需资源超预算：仅省略上下文</text>
+<text x="1226" y="474" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">强制指令从不截断</text>
+<text x="1226" y="494" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">其他链接仍须按精确修订继续读取</text>
+<rect x="1210" y="529" width="314" height="101" rx="9" fill="#ffffff" stroke="#c5d3e8" stroke-width="1.25"/>
+<text x="1226" y="556" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#1e293b">可检查的返回材料</text>
+<text x="1226" y="577" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">引用、修订、路径与 SHA256</text>
+<text x="1226" y="597" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">来源、署名和适用限制</text>
+<text x="1226" y="617" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">完整资源及显式遗漏原因</text>
+<rect x="1210" y="646" width="314" height="108" rx="9" fill="#ffffff" stroke="#c5d3e8" stroke-width="1.25" stroke-dasharray="4 4"/>
+<text x="1226" y="673" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#1e293b">Agent 显式决定使用</text>
+<text x="1226" y="694" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">结合目标、授权与适用条件</text>
+<text x="1226" y="714" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">先读取，再按适用限制行动</text>
+<text x="1226" y="734" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#526782">读取不会执行内容或提升候选状态</text>
+</g>
+<path d="M388 423 H410" fill="none" stroke="#5482ff" stroke-width="1.7" marker-end="url(#flow)"/>
+<path d="M778 423 H800" fill="none" stroke="#5482ff" stroke-width="1.7" marker-end="url(#flow)"/>
+<path d="M1164 423 H1186" fill="none" stroke="#5482ff" stroke-width="1.7" marker-end="url(#flow)"/>
+<path d="M232 884 V821 H216 V778" fill="none" stroke="#966bd7" stroke-width="1.5" stroke-dasharray="5 5" marker-end="url(#reference)"/>
+<text x="257" y="826" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#966bd7">冻结证据只作引用；运行时不改写。</text>
+<path d="M1366 884 V856 H594 V778" fill="none" stroke="#7c8da8" stroke-width="1.5" stroke-dasharray="5 5" marker-end="url(#review)"/>
+<text x="660" y="848" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="15" font-weight="400" text-anchor="start" fill="#7c8da8">规范内容与双矩阵经审查后变更；store 候选不等于仓库覆盖。</text>
+<g id="evidence-and-maintenance">
+<rect x="56" y="884" width="1488" height="118" rx="12" fill="#fafbfe" stroke="#dce5f1" stroke-width="1.25"/>
+<text x="76" y="912" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="14" font-weight="600" text-anchor="start" fill="#7c8da8">5 · 冻结证据与经过审查的维护</text>
+<rect x="76" y="928" width="344" height="56" rx="8" fill="#ffffff" stroke="#c5d3e8" stroke-width="1.25"/>
+<text x="91" y="950" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#1e293b">冻结的原始证据</text>
+<text x="91" y="972" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="14" font-weight="400" text-anchor="start" fill="#526782">计划、结果、补丁与验证器</text>
+<rect x="436" y="928" width="344" height="56" rx="8" fill="#ffffff" stroke="#c5d3e8" stroke-width="1.25"/>
+<text x="451" y="950" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#1e293b">Reference Matrix</text>
+<text x="451" y="972" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="14" font-weight="400" text-anchor="start" fill="#526782">来源、已检查范围与提取内容</text>
+<rect x="796" y="928" width="344" height="56" rx="8" fill="#ffffff" stroke="#c5d3e8" stroke-width="1.25"/>
+<text x="811" y="950" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#1e293b">SearchSpace Matrix</text>
+<text x="811" y="972" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="14" font-weight="400" text-anchor="start" fill="#526782">资源 × 优化阶段的覆盖情况</text>
+<rect x="1156" y="928" width="368" height="56" rx="8" fill="#ffffff" stroke="#c5d3e8" stroke-width="1.25"/>
+<text x="1171" y="950" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="16" font-weight="600" text-anchor="start" fill="#1e293b">经过审查的仓库变更</text>
+<text x="1171" y="972" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="14" font-weight="400" text-anchor="start" fill="#526782">规范内容、两个矩阵与投影刷新</text>
+</g>
+<rect x="56" y="1021" width="14" height="11" rx="2" fill="#edf3ff" stroke="#5482ff" stroke-width="1.25"/>
+<text x="80" y="1032" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="14" font-weight="400" text-anchor="start" fill="#526782">契约</text>
+<rect x="184" y="1021" width="14" height="11" rx="2" fill="#effaf4" stroke="#37ad70" stroke-width="1.25"/>
+<text x="208" y="1032" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="14" font-weight="400" text-anchor="start" fill="#526782">检索</text>
+<path d="M320 1027 H354" fill="none" stroke="#5482ff" stroke-width="1.7"/>
+<text x="365" y="1032" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="14" font-weight="400" text-anchor="start" fill="#526782">数据流</text>
+<path d="M482 1027 H516" fill="none" stroke="#966bd7" stroke-width="1.5" stroke-dasharray="5 5"/>
+<text x="527" y="1032" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="14" font-weight="400" text-anchor="start" fill="#526782">引用 / 审查</text>
+<text x="1544" y="1032" font-family="Microsoft YaHei, Noto Sans SC, Segoe UI, sans-serif" font-size="14" font-weight="400" text-anchor="end" fill="#7c8da8">InferRSI · 来源 4817f92121df · 不自动执行或提升候选状态</text>
+</svg>
diff --git a/evidence/knowledge-retrieval-20261027/results/architecture/layout-checks.json b/evidence/knowledge-retrieval-20261027/results/architecture/layout-checks.json
new file mode 100644
index 00000000..de329871
--- /dev/null
+++ b/evidence/knowledge-retrieval-20261027/results/architecture/layout-checks.json
@@ -0,0 +1,214 @@
+[
+  {
+    "locale": "en",
+    "width": 1600,
+    "height": 1050,
+    "baseline": "4817f92121df73fd9cd5fb4a319e3c56a68b4a5b",
+    "author_model": "gpt-6.1-sol",
+    "author_reasoning_effort": "xhigh",
+    "text_checks": [
+      {
+        "title": "Complete skill library",
+        "line_count": 4,
+        "title_width": 163.0,
+        "available_width": 260
+      },
+      {
+        "title": "Imported catalog sources",
+        "line_count": 3,
+        "title_width": 193.0,
+        "available_width": 260
+      },
+      {
+        "title": "Typed source identities",
+        "line_count": 5,
+        "title_width": 177.0,
+        "available_width": 260
+      },
+      {
+        "title": "Ownership and qualification",
+        "line_count": 5,
+        "title_width": 217.0,
+        "available_width": 260
+      },
+      {
+        "title": "Stable-ID registry",
+        "line_count": 3,
+        "title_width": 133.0,
+        "available_width": 292
+      },
+      {
+        "title": "Source-pinned projection",
+        "line_count": 6,
+        "title_width": 195.0,
+        "available_width": 292
+      },
+      {
+        "title": "Explicit authorities",
+        "line_count": 5,
+        "title_width": 144.0,
+        "available_width": 292
+      },
+      {
+        "title": "Bundle identity contract",
+        "line_count": 3,
+        "title_width": 185.0,
+        "available_width": 292
+      },
+      {
+        "title": "Legacy: knowledge search",
+        "line_count": 4,
+        "title_width": 199.0,
+        "available_width": 288
+      },
+      {
+        "title": "Default: Collect",
+        "line_count": 5,
+        "title_width": 118.0,
+        "available_width": 288
+      },
+      {
+        "title": "Two independent opt-ins",
+        "line_count": 10,
+        "title_width": 192.0,
+        "available_width": 288
+      },
+      {
+        "title": "Exact-revision access",
+        "line_count": 4,
+        "title_width": 160.0,
+        "available_width": 284
+      },
+      {
+        "title": "Complete context contract",
+        "line_count": 7,
+        "title_width": 203.0,
+        "available_width": 284
+      },
+      {
+        "title": "Inspectable returned material",
+        "line_count": 3,
+        "title_width": 226.0,
+        "available_width": 284
+      },
+      {
+        "title": "Agent chooses explicit use",
+        "line_count": 3,
+        "title_width": 202.0,
+        "available_width": 284
+      }
+    ],
+    "svg": "inferrsi-knowledge-architecture.en.svg",
+    "png": "inferrsi-knowledge-architecture.en.png",
+    "render_size": [
+      3200,
+      2100
+    ]
+  },
+  {
+    "locale": "zh-CN",
+    "width": 1600,
+    "height": 1050,
+    "baseline": "4817f92121df73fd9cd5fb4a319e3c56a68b4a5b",
+    "author_model": "gpt-6.1-sol",
+    "author_reasoning_effort": "xhigh",
+    "text_checks": [
+      {
+        "title": "完整技能库",
+        "line_count": 4,
+        "title_width": 80.0,
+        "available_width": 260
+      },
+      {
+        "title": "导入的目录来源",
+        "line_count": 3,
+        "title_width": 112.0,
+        "available_width": 260
+      },
+      {
+        "title": "按类型保存来源身份",
+        "line_count": 5,
+        "title_width": 144.0,
+        "available_width": 260
+      },
+      {
+        "title": "内容归属与验证边界",
+        "line_count": 4,
+        "title_width": 144.0,
+        "available_width": 260
+      },
+      {
+        "title": "稳定 ID 注册表",
+        "line_count": 3,
+        "title_width": 108.0,
+        "available_width": 292
+      },
+      {
+        "title": "固定来源的生成投影",
+        "line_count": 6,
+        "title_width": 144.0,
+        "available_width": 292
+      },
+      {
+        "title": "显式选择三类权威",
+        "line_count": 5,
+        "title_width": 128.0,
+        "available_width": 292
+      },
+      {
+        "title": "完整包身份契约",
+        "line_count": 3,
+        "title_width": 112.0,
+        "available_width": 292
+      },
+      {
+        "title": "既有入口：knowledge search",
+        "line_count": 4,
+        "title_width": 226.0,
+        "available_width": 288
+      },
+      {
+        "title": "默认入口：Collect",
+        "line_count": 5,
+        "title_width": 136.0,
+        "available_width": 288
+      },
+      {
+        "title": "两项相互独立的可选能力",
+        "line_count": 10,
+        "title_width": 176.0,
+        "available_width": 288
+      },
+      {
+        "title": "读取精确内容修订",
+        "line_count": 4,
+        "title_width": 128.0,
+        "available_width": 284
+      },
+      {
+        "title": "完整上下文契约",
+        "line_count": 7,
+        "title_width": 112.0,
+        "available_width": 284
+      },
+      {
+        "title": "可检查的返回材料",
+        "line_count": 3,
+        "title_width": 128.0,
+        "available_width": 284
+      },
+      {
+        "title": "Agent 显式决定使用",
+        "line_count": 3,
+        "title_width": 150.0,
+        "available_width": 284
+      }
+    ],
+    "svg": "inferrsi-knowledge-architecture.zh-CN.svg",
+    "png": "inferrsi-knowledge-architecture.zh-CN.png",
+    "render_size": [
+      3200,
+      2100
+    ]
+  }
+]
diff --git a/evidence/knowledge-retrieval-20261027/results/baseline-profile.json b/evidence/knowledge-retrieval-20261027/results/baseline-profile.json
new file mode 100644
index 00000000..a62c9d8d
--- /dev/null
+++ b/evidence/knowledge-retrieval-20261027/results/baseline-profile.json
@@ -0,0 +1,744 @@
+{
+  "case": "dev05",
+  "mode": "fresh-reader relevance, cProfile diagnostic only",
+  "modules": {
+    "inferrsi": {
+      "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+      "source": "src/inferrsi/__init__.py"
+    },
+    "inferrsi.catalog": {
+      "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+      "source": "src/inferrsi/catalog.py"
+    },
+    "inferrsi.cli": {
+      "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+      "source": "src/inferrsi/cli.py"
+    },
+    "inferrsi.knowledge_operations": {
+      "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+      "source": "src/inferrsi/knowledge_operations.py"
+    }
+  },
+  "profile": [
+    {
+      "calls": 1,
+      "cumulative_seconds": 1.4654439000000001,
+      "function": "invoke",
+      "line": 189,
+      "primitive_calls": 1,
+      "source": "<external>/evaluate.py",
+      "total_seconds": 0.00011070000000000001
+    },
+    {
+      "calls": 1,
+      "cumulative_seconds": 1.465332,
+      "function": "collect",
+      "line": 415,
+      "primitive_calls": 1,
+      "source": "src/inferrsi/knowledge_operations.py",
+      "total_seconds": 0.0031245
+    },
+    {
+      "calls": 60,
+      "cumulative_seconds": 1.1570676,
+      "function": "view",
+      "line": 150,
+      "primitive_calls": 60,
+      "source": "src/inferrsi/knowledge_operations.py",
+      "total_seconds": 0.0007606000000000001
+    },
+    {
+      "calls": 60,
+      "cumulative_seconds": 1.1085909,
+      "function": "manifest",
+      "line": 90,
+      "primitive_calls": 60,
+      "source": "src/inferrsi/knowledge_operations.py",
+      "total_seconds": 0.047438100000000004
+    },
+    {
+      "calls": 41563,
+      "cumulative_seconds": 0.6908788,
+      "function": "_hash",
+      "line": 45,
+      "primitive_calls": 41563,
+      "source": "src/inferrsi/knowledge_operations.py",
+      "total_seconds": 0.0270916
+    },
+    {
+      "calls": 41563,
+      "cumulative_seconds": 0.6259869,
+      "function": "<built-in method _hashlib.openssl_sha256>",
+      "line": 0,
+      "primitive_calls": 41563,
+      "source": "<external>/~",
+      "total_seconds": 0.6259869
+    },
+    {
+      "calls": 41818,
+      "cumulative_seconds": 0.35470650000000004,
+      "function": "safe_path",
+      "line": 28,
+      "primitive_calls": 41818,
+      "source": "src/inferrsi/knowledge_operations.py",
+      "total_seconds": 0.08225220000000001
+    },
+    {
+      "calls": 60,
+      "cumulative_seconds": 0.1535705,
+      "function": "_read",
+      "line": 264,
+      "primitive_calls": 60,
+      "source": "src/inferrsi/knowledge_operations.py",
+      "total_seconds": 0.0001132
+    },
+    {
+      "calls": 60,
+      "cumulative_seconds": 0.1534327,
+      "function": "_knowledge",
+      "line": 315,
+      "primitive_calls": 60,
+      "source": "src/inferrsi/knowledge_operations.py",
+      "total_seconds": 0.0009647000000000001
+    },
+    {
+      "calls": 60,
+      "cumulative_seconds": 0.1432243,
+      "function": "_skill_library",
+      "line": 196,
+      "primitive_calls": 60,
+      "source": "src/inferrsi/knowledge_operations.py",
+      "total_seconds": 0.0002001
+    },
+    {
+      "calls": 197,
+      "cumulative_seconds": 0.1371058,
+      "function": "_tree",
+      "line": 166,
+      "primitive_calls": 1,
+      "source": "src/inferrsi/knowledge_operations.py",
+      "total_seconds": 0.0038594000000000002
+    },
+    {
+      "calls": 41988,
+      "cumulative_seconds": 0.1159894,
+      "function": "<built-in method builtins.any>",
+      "line": 0,
+      "primitive_calls": 41988,
+      "source": "<external>/~",
+      "total_seconds": 0.058914100000000004
+    },
+    {
+      "calls": 1213,
+      "cumulative_seconds": 0.1028087,
+      "function": "dumps",
+      "line": 183,
+      "primitive_calls": 1213,
+      "source": "<external>/__init__.py",
+      "total_seconds": 0.0025973000000000003
+    },
+    {
+      "calls": 1213,
+      "cumulative_seconds": 0.09975060000000001,
+      "function": "encode",
+      "line": 183,
+      "primitive_calls": 1213,
+      "source": "<external>/encoder.py",
+      "total_seconds": 0.0017369000000000002
+    },
+    {
+      "calls": 253,
+      "cumulative_seconds": 0.08739540000000001,
+      "function": "_json",
+      "line": 41,
+      "primitive_calls": 253,
+      "source": "src/inferrsi/knowledge_operations.py",
+      "total_seconds": 0.0007021
+    },
+    {
+      "calls": 493,
+      "cumulative_seconds": 0.083722,
+      "function": "iterencode",
+      "line": 205,
+      "primitive_calls": 493,
+      "source": "<external>/encoder.py",
+      "total_seconds": 0.083722
+    },
+    {
+      "calls": 42551,
+      "cumulative_seconds": 0.08183270000000001,
+      "function": "__init__",
+      "line": 358,
+      "primitive_calls": 42551,
+      "source": "<external>/pathlib.py",
+      "total_seconds": 0.0488065
+    },
+    {
+      "calls": 342,
+      "cumulative_seconds": 0.07582190000000001,
+      "function": "read_bytes",
+      "line": 1015,
+      "primitive_calls": 342,
+      "source": "<external>/pathlib.py",
+      "total_seconds": 0.0010411
+    },
+    {
+      "calls": 349,
+      "cumulative_seconds": 0.058220600000000004,
+      "function": "<method 'read' of '_io.BufferedReader' objects>",
+      "line": 0,
+      "primitive_calls": 349,
+      "source": "<external>/~",
+      "total_seconds": 0.058220600000000004
+    },
+    {
+      "calls": 312067,
+      "cumulative_seconds": 0.055727200000000005,
+      "function": "<genexpr>",
+      "line": 35,
+      "primitive_calls": 312067,
+      "source": "src/inferrsi/knowledge_operations.py",
+      "total_seconds": 0.055727200000000005
+    },
+    {
+      "calls": 2,
+      "cumulative_seconds": 0.0447465,
+      "function": "owners",
+      "line": 204,
+      "primitive_calls": 2,
+      "source": "src/inferrsi/knowledge_operations.py",
+      "total_seconds": 0.0001465
+    },
+    {
+      "calls": 338,
+      "cumulative_seconds": 0.0446084,
+      "function": "records",
+      "line": 190,
+      "primitive_calls": 338,
+      "source": "src/inferrsi/knowledge_operations.py",
+      "total_seconds": 0.0001292
+    },
+    {
+      "calls": 1,
+      "cumulative_seconds": 0.0444792,
+      "function": "knowledge_records",
+      "line": 88,
+      "primitive_calls": 1,
+      "source": "src/inferrsi/catalog.py",
+      "total_seconds": 0.0009307
+    },
+    {
+      "calls": 60,
+      "cumulative_seconds": 0.043403000000000004,
+      "function": "indexed_skill",
+      "line": 24,
+      "primitive_calls": 60,
+      "source": "src/inferrsi/_skill_selection.py",
+      "total_seconds": 0.012047200000000001
+    },
+    {
+      "calls": 1298,
+      "cumulative_seconds": 0.0423665,
+      "function": "<built-in method nt.stat>",
+      "line": 0,
+      "primitive_calls": 1298,
+      "source": "<external>/~",
+      "total_seconds": 0.0402517
+    },
+    {
+      "calls": 1265,
+      "cumulative_seconds": 0.0415885,
+      "function": "stat",
+      "line": 835,
+      "primitive_calls": 1265,
+      "source": "<external>/pathlib.py",
+      "total_seconds": 0.0006975000000000001
+    },
+    {
+      "calls": 41563,
+      "cumulative_seconds": 0.0378003,
+      "function": "<method 'hexdigest' of '_hashlib.HASH' objects>",
+      "line": 0,
+      "primitive_calls": 41563,
+      "source": "<external>/~",
+      "total_seconds": 0.0378003
+    },
+    {
+      "calls": 60,
+      "cumulative_seconds": 0.0359603,
+      "function": "skill_score",
+      "line": 85,
+      "primitive_calls": 60,
+      "source": "src/inferrsi/_skill_selection.py",
+      "total_seconds": 0.0003035
+    },
+    {
+      "calls": 60,
+      "cumulative_seconds": 0.0351922,
+      "function": "lexical_score",
+      "line": 16,
+      "primitive_calls": 60,
+      "source": "src/inferrsi/_skill_selection.py",
+      "total_seconds": 0.00014030000000000002
+    },
+    {
+      "calls": 300,
+      "cumulative_seconds": 0.0343604,
+      "function": "<built-in method builtins.sum>",
+      "line": 0,
+      "primitive_calls": 60,
+      "source": "<external>/~",
+      "total_seconds": 0.0005647
+    },
+    {
+      "calls": 300,
+      "cumulative_seconds": 0.034226,
+      "function": "<genexpr>",
+      "line": 18,
+      "primitive_calls": 300,
+      "source": "src/inferrsi/_skill_selection.py",
+      "total_seconds": 0.0002928
+    },
+    {
+      "calls": 1200,
+      "cumulative_seconds": 0.0335029,
+      "function": "<genexpr>",
+      "line": 19,
+      "primitive_calls": 1200,
+      "source": "src/inferrsi/_skill_selection.py",
+      "total_seconds": 0.0029366
+    },
+    {
+      "calls": 41818,
+      "cumulative_seconds": 0.0325714,
+      "function": "is_absolute",
+      "line": 748,
+      "primitive_calls": 41818,
+      "source": "<external>/pathlib.py",
+      "total_seconds": 0.021395800000000003
+    },
+    {
+      "calls": 633,
+      "cumulative_seconds": 0.0324894,
+      "function": "<built-in method builtins.sorted>",
+      "line": 0,
+      "primitive_calls": 633,
+      "source": "<external>/~",
+      "total_seconds": 0.0075525
+    },
+    {
+      "calls": 125712,
+      "cumulative_seconds": 0.0319652,
+      "function": "<method 'startswith' of 'str' objects>",
+      "line": 0,
+      "primitive_calls": 125712,
+      "source": "<external>/~",
+      "total_seconds": 0.0319652
+    },
+    {
+      "calls": 41818,
+      "cumulative_seconds": 0.026291500000000002,
+      "function": "__new__",
+      "line": 343,
+      "primitive_calls": 41818,
+      "source": "<external>/pathlib.py",
+      "total_seconds": 0.015218500000000001
+    },
+    {
+      "calls": 727,
+      "cumulative_seconds": 0.022727400000000002,
+      "function": "is_dir",
+      "line": 870,
+      "primitive_calls": 727,
+      "source": "<external>/pathlib.py",
+      "total_seconds": 0.0009384
+    },
+    {
+      "calls": 538,
+      "cumulative_seconds": 0.021223500000000003,
+      "function": "is_symlink",
+      "line": 909,
+      "primitive_calls": 538,
+      "source": "<external>/pathlib.py",
+      "total_seconds": 0.0007321000000000001
+    },
+    {
+      "calls": 113574,
+      "cumulative_seconds": 0.0206064,
+      "function": "<built-in method builtins.isinstance>",
+      "line": 0,
+      "primitive_calls": 113574,
+      "source": "<external>/~",
+      "total_seconds": 0.0206064
+    },
+    {
+      "calls": 538,
+      "cumulative_seconds": 0.0203108,
+      "function": "lstat",
+      "line": 842,
+      "primitive_calls": 538,
+      "source": "<external>/pathlib.py",
+      "total_seconds": 0.0002974
+    },
+    {
+      "calls": 42031,
+      "cumulative_seconds": 0.0195109,
+      "function": "<method 'split' of 'str' objects>",
+      "line": 0,
+      "primitive_calls": 42031,
+      "source": "<external>/~",
+      "total_seconds": 0.0195109
+    },
+    {
+      "calls": 450,
+      "cumulative_seconds": 0.018421200000000002,
+      "function": "open",
+      "line": 1005,
+      "primitive_calls": 450,
+      "source": "<external>/pathlib.py",
+      "total_seconds": 0.0004108
+    },
+    {
+      "calls": 450,
+      "cumulative_seconds": 0.0179943,
+      "function": "<built-in method _io.open>",
+      "line": 0,
+      "primitive_calls": 450,
+      "source": "<external>/~",
+      "total_seconds": 0.017426900000000002
+    },
+    {
+      "calls": 1141,
+      "cumulative_seconds": 0.0162028,
+      "function": "<method 'casefold' of 'str' objects>",
+      "line": 0,
+      "primitive_calls": 1141,
+      "source": "<external>/~",
+      "total_seconds": 0.0162028
+    },
+    {
+      "calls": 973,
+      "cumulative_seconds": 0.0159764,
+      "function": "iterdir",
+      "line": 1050,
+      "primitive_calls": 973,
+      "source": "<external>/pathlib.py",
+      "total_seconds": 0.0006183
+    },
+    {
+      "calls": 720,
+      "cumulative_seconds": 0.0135604,
+      "function": "<built-in method _json.encode_basestring>",
+      "line": 0,
+      "primitive_calls": 720,
+      "source": "<external>/~",
+      "total_seconds": 0.0135604
+    },
+    {
+      "calls": 42568,
+      "cumulative_seconds": 0.011472000000000001,
+      "function": "<built-in method __new__ of type object at 0x00007FFB556C58F0>",
+      "line": 0,
+      "primitive_calls": 42568,
+      "source": "<external>/~",
+      "total_seconds": 0.011450100000000001
+    },
+    {
+      "calls": 2,
+      "cumulative_seconds": 0.0114146,
+      "function": "wrapper",
+      "line": 33,
+      "primitive_calls": 2,
+      "source": "<external>/_common.py",
+      "total_seconds": 2.5e-06
+    },
+    {
+      "calls": 2,
+      "cumulative_seconds": 0.011412100000000001,
+      "function": "files",
+      "line": 51,
+      "primitive_calls": 2,
+      "source": "<external>/_common.py",
+      "total_seconds": 5.4e-06
+    },
+    {
+      "calls": 2,
+      "cumulative_seconds": 0.0113577,
+      "function": "from_package",
+      "line": 107,
+      "primitive_calls": 2,
+      "source": "<external>/_common.py",
+      "total_seconds": 8.900000000000001e-06
+    },
+    {
+      "calls": 2,
+      "cumulative_seconds": 0.0113384,
+      "function": "get_resource_reader",
+      "line": 28,
+      "primitive_calls": 2,
+      "source": "<external>/_adapters.py",
+      "total_seconds": 5.1e-06
+    },
+    {
+      "calls": 2,
+      "cumulative_seconds": 0.011332700000000001,
+      "function": "_native",
+      "line": 149,
+      "primitive_calls": 2,
+      "source": "<external>/_adapters.py",
+      "total_seconds": 4.2000000000000004e-06
+    },
+    {
+      "calls": 2,
+      "cumulative_seconds": 0.011327100000000001,
+      "function": "_reader",
+      "line": 144,
+      "primitive_calls": 2,
+      "source": "<external>/_adapters.py",
+      "total_seconds": 1.29e-05
+    },
+    {
+      "calls": 9,
+      "cumulative_seconds": 0.0113124,
+      "function": "_check_name_wrapper",
+      "line": 643,
+      "primitive_calls": 2,
+      "source": "<external>/<frozen importlib._bootstrap_external>",
+      "total_seconds": 1.6400000000000002e-05
+    },
+    {
+      "calls": 2,
+      "cumulative_seconds": 0.0113089,
+      "function": "get_resource_reader",
+      "line": 1192,
+      "primitive_calls": 2,
+      "source": "<external>/<frozen importlib._bootstrap_external>",
+      "total_seconds": 7.7e-06
+    },
+    {
+      "calls": 10,
+      "cumulative_seconds": 0.011211800000000001,
+      "function": "_find_and_load",
+      "line": 1349,
+      "primitive_calls": 3,
+      "source": "<external>/<frozen importlib._bootstrap>",
+      "total_seconds": 6.620000000000001e-05
+    },
+    {
+      "calls": 8,
+      "cumulative_seconds": 0.0111423,
+      "function": "_find_and_load_unlocked",
+      "line": 1304,
+      "primitive_calls": 1,
+      "source": "<external>/<frozen importlib._bootstrap>",
+      "total_seconds": 4.2900000000000006e-05
+    },
+    {
+      "calls": 402,
+      "cumulative_seconds": 0.010942,
+      "function": "_walk",
+      "line": 33,
+      "primitive_calls": 134,
+      "source": "src/inferrsi/catalog.py",
+      "total_seconds": 0.00046320000000000004
+    },
+    {
+      "calls": 8,
+      "cumulative_seconds": 0.01094,
+      "function": "_load_unlocked",
+      "line": 911,
+      "primitive_calls": 1,
+      "source": "<external>/<frozen importlib._bootstrap>",
+      "total_seconds": 3.15e-05
+    },
+    {
+      "calls": 7,
+      "cumulative_seconds": 0.0108844,
+      "function": "exec_module",
+      "line": 989,
+      "primitive_calls": 1,
+      "source": "<external>/<frozen importlib._bootstrap_external>",
+      "total_seconds": 2.2800000000000002e-05
+    },
+    {
+      "calls": 108,
+      "cumulative_seconds": 0.010346600000000001,
+      "function": "read_text",
+      "line": 1022,
+      "primitive_calls": 108,
+      "source": "<external>/pathlib.py",
+      "total_seconds": 0.0002852
+    },
+    {
+      "calls": 16,
+      "cumulative_seconds": 0.0102004,
+      "function": "_call_with_frames_removed",
+      "line": 480,
+      "primitive_calls": 2,
+      "source": "<external>/<frozen importlib._bootstrap>",
+      "total_seconds": 1.27e-05
+    },
+    {
+      "calls": 8,
+      "cumulative_seconds": 0.010197300000000001,
+      "function": "<built-in method builtins.exec>",
+      "line": 0,
+      "primitive_calls": 1,
+      "source": "<external>/~",
+      "total_seconds": 2.57e-05
+    },
+    {
+      "calls": 1,
+      "cumulative_seconds": 0.0101954,
+      "function": "<module>",
+      "line": 1,
+      "primitive_calls": 1,
+      "source": "<external>/readers.py",
+      "total_seconds": 6e-06
+    },
+    {
+      "calls": 44564,
+      "cumulative_seconds": 0.010047,
+      "function": "<method 'append' of 'list' objects>",
+      "line": 0,
+      "primitive_calls": 44564,
+      "source": "<external>/~",
+      "total_seconds": 0.010047
+    },
+    {
+      "calls": 271,
+      "cumulative_seconds": 0.0092724,
+      "function": "<built-in method nt.listdir>",
+      "line": 0,
+      "primitive_calls": 271,
+      "source": "<external>/~",
+      "total_seconds": 0.0090214
+    },
+    {
+      "calls": 1,
+      "cumulative_seconds": 0.0090048,
+      "function": "<module>",
+      "line": 1,
+      "primitive_calls": 1,
+      "source": "<external>/readers.py",
+      "total_seconds": 3.16e-05
+    },
+    {
+      "calls": 44266,
+      "cumulative_seconds": 0.0086978,
+      "function": "<built-in method nt.fspath>",
+      "line": 0,
+      "primitive_calls": 44266,
+      "source": "<external>/~",
+      "total_seconds": 0.0086978
+    },
+    {
+      "calls": 168,
+      "cumulative_seconds": 0.0086245,
+      "function": "loads",
+      "line": 299,
+      "primitive_calls": 168,
+      "source": "<external>/__init__.py",
+      "total_seconds": 0.0004487
+    },
+    {
+      "calls": 154,
+      "cumulative_seconds": 0.008316,
+      "function": "<genexpr>",
+      "line": 78,
+      "primitive_calls": 154,
+      "source": "src/inferrsi/_skill_selection.py",
+      "total_seconds": 0.003953900000000001
+    },
+    {
+      "calls": 12,
+      "cumulative_seconds": 0.007492100000000001,
+      "function": "select",
+      "line": 475,
+      "primitive_calls": 12,
+      "source": "src/inferrsi/knowledge_operations.py",
+      "total_seconds": 2.5200000000000003e-05
+    },
+    {
+      "calls": 168,
+      "cumulative_seconds": 0.0074681,
+      "function": "decode",
+      "line": 332,
+      "primitive_calls": 168,
+      "source": "<external>/decoder.py",
+      "total_seconds": 0.0005573
+    },
+    {
+      "calls": 7,
+      "cumulative_seconds": 0.007377300000000001,
+      "function": "get_code",
+      "line": 1062,
+      "primitive_calls": 7,
+      "source": "<external>/<frozen importlib._bootstrap_external>",
+      "total_seconds": 8.72e-05
+    },
+    {
+      "calls": 168,
+      "cumulative_seconds": 0.0062762,
+      "function": "raw_decode",
+      "line": 343,
+      "primitive_calls": 168,
+      "source": "<external>/decoder.py",
+      "total_seconds": 0.0062762
+    },
+    {
+      "calls": 704,
+      "cumulative_seconds": 0.0062142000000000005,
+      "function": "_make_child_relpath",
+      "line": 1065,
+      "primitive_calls": 704,
+      "source": "<external>/pathlib.py",
+      "total_seconds": 0.0018295000000000002
+    },
+    {
+      "calls": 7,
+      "cumulative_seconds": 0.0057663,
+      "function": "get_data",
+      "line": 1183,
+      "primitive_calls": 7,
+      "source": "<external>/<frozen importlib._bootstrap_external>",
+      "total_seconds": 3.3600000000000004e-05
+    },
+    {
+      "calls": 7,
+      "cumulative_seconds": 0.0050149,
+      "function": "<built-in method _io.open_code>",
+      "line": 0,
+      "primitive_calls": 7,
+      "source": "<external>/~",
+      "total_seconds": 0.0050149
+    },
+    {
+      "calls": 457,
+      "cumulative_seconds": 0.0048805,
+      "function": "<method '__exit__' of '_io._IOBase' objects>",
+      "line": 0,
+      "primitive_calls": 457,
+      "source": "<external>/~",
+      "total_seconds": 0.0048805
+    },
+    {
+      "calls": 1,
+      "cumulative_seconds": 0.004521000000000001,
+      "function": "<module>",
+      "line": 1,
+      "primitive_calls": 1,
+      "source": "<external>/__init__.py",
+      "total_seconds": 4.47e-05
+    },
+    {
+      "calls": 731,
+      "cumulative_seconds": 0.0043057,
+      "function": "with_segments",
+      "line": 380,
+      "primitive_calls": 731,
+      "source": "<external>/pathlib.py",
+      "total_seconds": 0.0009497
+    }
+  ]
+}
diff --git a/evidence/knowledge-retrieval-20261027/results/baseline-trace-129.json b/evidence/knowledge-retrieval-20261027/results/baseline-trace-129.json
new file mode 100644
index 00000000..d1813029
--- /dev/null
+++ b/evidence/knowledge-retrieval-20261027/results/baseline-trace-129.json
@@ -0,0 +1,713 @@
+{
+  "modules": {
+    "inferrsi": {
+      "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+      "source": "src/inferrsi/__init__.py"
+    },
+    "inferrsi.catalog": {
+      "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+      "source": "src/inferrsi/catalog.py"
+    },
+    "inferrsi.cli": {
+      "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+      "source": "src/inferrsi/cli.py"
+    },
+    "inferrsi.knowledge_operations": {
+      "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+      "source": "src/inferrsi/knowledge_operations.py"
+    }
+  },
+  "outputs": [
+    {
+      "answerable": true,
+      "case": "dev09",
+      "mode": "search",
+      "output": {
+        "ordered_items": [],
+        "serialized_bytes": 2,
+        "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+      },
+      "reciprocal_rank_at_5": 0,
+      "success_at_5": 0
+    },
+    {
+      "answerable": true,
+      "case": "dev09",
+      "mode": "relevance",
+      "output": {
+        "collisions": [],
+        "material_bytes": 432011,
+        "no_match": false,
+        "omissions": [
+          {
+            "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.weight_loading_optimization",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.assess_serving_preemption",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.comfy_ui",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.host_memory_management",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.host_overhead_optimization",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.inference_benchmark_design",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.kernel_writing_guide",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.npu_skill_factory",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.optimization_search_policy",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.plan_activation_residency",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.rsi_evidence",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.storage_io_measurement",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.training_input_pipeline",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.training_resource_budget",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+            "reason": "output_limit"
+          }
+        ],
+        "ordered_items": [
+          {
+            "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+            "id": "inferrsi.common.skills.kv_tier_management",
+            "ref": "package:inferrsi.common.skills.kv_tier_management",
+            "resource_count": 343,
+            "resource_hashes_sha256": "ee42b55fa8a404149ecd1dccc32a892566ee8867710531bc0af553409758a019",
+            "revision": "sha256:5816c2e526a71152486af2230ace24d27b54baeb373bf5d27e353ccd44f640aa"
+          },
+          {
+            "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+            "id": "inferrsi.common.skills.bound_windowed_attention_replay",
+            "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+            "resource_count": 343,
+            "resource_hashes_sha256": "fb70e2e746f7ba833a696431c6a4deda830188acf5c4eafce6b0008818f8190a",
+            "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334"
+          },
+          {
+            "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+            "id": "inferrsi.common.skills.audit_sd_cpp_residency",
+            "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+            "resource_count": 343,
+            "resource_hashes_sha256": "c83d7a76dffa5a3d97e932a3eac4602ad82e20babb09d9942504fff6b4db53d6",
+            "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde"
+          },
+          {
+            "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+            "id": "inferrsi.common.skills.diffusers_invocation_state",
+            "ref": "package:inferrsi.common.skills.diffusers_invocation_state",
+            "resource_count": 343,
+            "resource_hashes_sha256": "64b10a60e44276bc4a20b3441f65afd929341e8558fde32491ba27a20fde4020",
+            "revision": "sha256:a8529f2c6bf5f84ce48ba0d6426f1bf1bc6fce172cd0a599b6a149878e3b4be4"
+          },
+          {
+            "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+            "id": "inferrsi.common.skills.generation_cache_lifecycle",
+            "ref": "package:inferrsi.common.skills.generation_cache_lifecycle",
+            "resource_count": 343,
+            "resource_hashes_sha256": "052e200da2a160661cf44647685c70e52feec4b72fc0f2e7b34a10ce54710e03",
+            "revision": "sha256:22411ef79b30ddd4f8b01a7ed5d00fc4fc75a3c9f1e4d31f56f2e440d9b09d2f"
+          }
+        ],
+        "selection": {
+          "cue": "LMCache skip_l1 retain",
+          "method": "weighted-lexical-v1",
+          "ranking_applied": true,
+          "scope": "generated-packaged-skills"
+        },
+        "serialized_bytes": 435629,
+        "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+      },
+      "reciprocal_rank_at_5": 1.0,
+      "success_at_5": 1
+    },
+    {
+      "answerable": true,
+      "case": "dev09",
+      "mode": "context_material",
+      "output": {
+        "collisions": [],
+        "context_budget": {
+          "accounting": "utf8-resource-content",
+          "excludes": "provenance, omission and control metadata",
+          "max_bytes": 65536
+        },
+        "context_bytes": 61835,
+        "context_omissions": [
+          {
+            "available_bytes": 3701,
+            "reason": "context_limit",
+            "ref": "package:inferrsi.common.skills.diffusers_invocation_state",
+            "required_bytes": 10129
+          },
+          {
+            "available_bytes": 3701,
+            "reason": "context_limit",
+            "ref": "package:inferrsi.common.skills.generation_cache_lifecycle",
+            "required_bytes": 10146
+          }
+        ],
+        "contexts": [
+          {
+            "attribution": [
+              {
+                "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                "content_sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27",
+                "paths": [
+                  "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md",
+                  "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/metadata.json",
+                  "src/inferrsi/data/skill_library/skill-index.json"
+                ],
+                "type": "git",
+                "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md"
+              }
+            ],
+            "content_bytes": 10812,
+            "limitations": [
+              "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+              "The imported status describes this index, not the original skill; its declared status is candidate.",
+              "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+              "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+              "Source-derived candidate; no local efficacy validation.",
+              "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+              "Imported guidance; reading does not validate or execute it.",
+              "Historical revisions require the matching installed snapshot or a retained store.",
+              "Selected resources are complete files; this view is not a standalone Bundle.",
+              "Follow omitted or other linked resources through the exact owner revision."
+            ],
+            "omissions": [
+              {
+                "path": "data/skill_library/repositories/lmcache/kv-tier-management/metadata.json",
+                "reason": "not_context_resource"
+              }
+            ],
+            "ref": "package:inferrsi.common.skills.kv_tier_management",
+            "resources": [
+              {
+                "locator": "whole-resource",
+                "path": "data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md",
+                "role": "core",
+                "sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27"
+              }
+            ],
+            "revision": "sha256:5816c2e526a71152486af2230ace24d27b54baeb373bf5d27e353ccd44f640aa",
+            "sources": [
+              {
+                "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                "content_sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27",
+                "paths": [
+                  "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md",
+                  "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/metadata.json",
+                  "src/inferrsi/data/skill_library/skill-index.json"
+                ],
+                "revision_precision": "full",
+                "type": "git",
+                "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md"
+              }
+            ]
+          },
+          {
+            "attribution": [
+              {
+                "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                "content_sha256": "4703beb6e5194ef2ff2c0b5ec69f64e35417a4bc0dfb50a10c80a920b77ef3b5",
+                "paths": [
+                  "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md",
+                  "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/metadata.json",
+                  "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/source-notes.md",
+                  "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/validation.md",
+                  "src/inferrsi/data/skill_library/skill-index.json"
+                ],
+                "type": "git",
+                "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md"
+              }
+            ],
+            "content_bytes": 29732,
+            "limitations": [
+              "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+              "The imported status describes this index, not the original skill; its declared status is candidate.",
+              "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+              "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+              "Source-derived candidate; no equivalence run was performed.",
+              "Applies to architectures with a demonstrated boundary dependency, not to arbitrary sliding-window stacks.",
+              "Window sizes, boundary-layer index and ring lengths are model-config specific and are not transferable.",
+              "Query-row compaction does not imply dropping historical keys; the two decisions are separate.",
+              "Prompt logprobs are unrecoverable for the clipped region; the reference forces them off with a warning.",
+              "The truncation path is off by default in the reference source.",
+              "Imported guidance; reading does not validate or execute it.",
+              "Historical revisions require the matching installed snapshot or a retained store.",
+              "Selected resources are complete files; this view is not a standalone Bundle.",
+              "Follow omitted or other linked resources through the exact owner revision."
+            ],
+            "omissions": [
+              {
+                "path": "data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/metadata.json",
+                "reason": "not_context_resource"
+              }
+            ],
+            "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+            "resources": [
+              {
+                "locator": "whole-resource",
+                "path": "data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md",
+                "role": "core",
+                "sha256": "4703beb6e5194ef2ff2c0b5ec69f64e35417a4bc0dfb50a10c80a920b77ef3b5"
+              },
+              {
+                "locator": "whole-resource",
+                "path": "data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/source-notes.md",
+                "role": "reference",
+                "sha256": "fd917bd5ac80bee9c9f4525f8e0300f6d67dc144e2a16632af42a67f3d53d098"
+              },
+              {
+                "locator": "whole-resource",
+                "path": "data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/validation.md",
+                "role": "reference",
+                "sha256": "60e71e70c5519bcdf6c171eec2be2f342e8801bedd746d6005f11adc646dfaff"
+              }
+            ],
+            "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334",
+            "sources": [
+              {
+                "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                "content_sha256": "4703beb6e5194ef2ff2c0b5ec69f64e35417a4bc0dfb50a10c80a920b77ef3b5",
+                "paths": [
+                  "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md",
+                  "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/metadata.json",
+                  "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/source-notes.md",
+                  "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/validation.md",
+                  "src/inferrsi/data/skill_library/skill-index.json"
+                ],
+                "revision_precision": "full",
+                "type": "git",
+                "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md"
+              }
+            ]
+          },
+          {
+            "attribution": [
+              {
+                "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                "content_sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6",
+                "paths": [
+                  "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE",
+                  "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md",
+                  "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json",
+                  "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md",
+                  "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json",
+                  "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md",
+                  "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md",
+                  "src/inferrsi/data/skill_library/skill-index.json"
+                ],
+                "type": "git",
+                "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md"
+              }
+            ],
+            "content_bytes": 21291,
+            "limitations": [
+              "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+              "The imported status describes this index, not the original skill; its declared status is candidate.",
+              "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+              "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+              "Source-derived candidate; native inference, allocator behavior, image quality and performance were not run.",
+              "Eighteen bounded readset paths plus one separately identity-checked README match the pinned Git commit/tree blob IDs and sizes (19 blobs total); header coverage is limited to selected owner/state declarations, and this does not authenticate every archive member or submodule contents.",
+              "GGML and other submodules are not expanded; matching private GGML headers are required even for a system library.",
+              "CPU bypasses the inspected accelerator capacity gate and segmentation eligibility, so CPU checks cannot qualify accelerator memory budgets or prefetch.",
+              "Video/audio pipelines, detailed model kernels, approximate denoising caches and full multi-device execution remain outside this readset.",
+              "Imported guidance; reading does not validate or execute it.",
+              "Historical revisions require the matching installed snapshot or a retained store.",
+              "Selected resources are complete files; this view is not a standalone Bundle.",
+              "Follow omitted or other linked resources through the exact owner revision."
+            ],
+            "omissions": [
+              {
+                "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json",
+                "reason": "not_context_resource"
+              },
+              {
+                "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json",
+                "reason": "not_relevant"
+              },
+              {
+                "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md",
+                "reason": "not_relevant"
+              }
+            ],
+            "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+            "resources": [
+              {
+                "locator": "whole-resource",
+                "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md",
+                "role": "core",
+                "sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6"
+              },
+              {
+                "locator": "whole-resource",
+                "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md",
+                "role": "reference",
+                "sha256": "b67affbd7d237d82b7f51be49f1537e9820fe56ec70a71ee76809f9e599ca693"
+              },
+              {
+                "locator": "whole-resource",
+                "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE",
+                "role": "reference",
+                "sha256": "b53fa08f515cb5a6fff7b9fd8fcd0961a4b80df29d74372d25e1f9171aa042ee"
+              },
+              {
+                "locator": "whole-resource",
+                "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md",
+                "role": "reference",
+                "sha256": "7b537cb972347dd68e303960f142892346d8b19b884e4d08016ea36a23bdaa7a"
+              }
+            ],
+            "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde",
+            "sources": [
+              {
+                "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                "content_sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6",
+                "paths": [
+                  "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE",
+                  "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md",
+                  "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json",
+                  "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md",
+                  "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json",
+                  "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md",
+                  "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md",
+                  "src/inferrsi/data/skill_library/skill-index.json"
+                ],
+                "revision_precision": "full",
+                "type": "git",
+                "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md"
+              }
+            ]
+          }
+        ],
+        "material_bytes": 462705,
+        "no_match": false,
+        "omissions": [
+          {
+            "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.weight_loading_optimization",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.assess_serving_preemption",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.comfy_ui",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.host_memory_management",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.host_overhead_optimization",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.inference_benchmark_design",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.kernel_writing_guide",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.npu_skill_factory",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.optimization_search_policy",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.plan_activation_residency",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.rsi_evidence",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.storage_io_measurement",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.training_input_pipeline",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.common.skills.training_resource_budget",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+            "reason": "output_limit"
+          },
+          {
+            "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+            "reason": "output_limit"
+          }
+        ],
+        "ordered_items": [
+          {
+            "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+            "id": "inferrsi.common.skills.kv_tier_management",
+            "ref": "package:inferrsi.common.skills.kv_tier_management",
+            "resource_count": 343,
+            "resource_hashes_sha256": "ee42b55fa8a404149ecd1dccc32a892566ee8867710531bc0af553409758a019",
+            "revision": "sha256:5816c2e526a71152486af2230ace24d27b54baeb373bf5d27e353ccd44f640aa"
+          },
+          {
+            "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+            "id": "inferrsi.common.skills.bound_windowed_attention_replay",
+            "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+            "resource_count": 343,
+            "resource_hashes_sha256": "fb70e2e746f7ba833a696431c6a4deda830188acf5c4eafce6b0008818f8190a",
+            "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334"
+          },
+          {
+            "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+            "id": "inferrsi.common.skills.audit_sd_cpp_residency",
+            "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+            "resource_count": 343,
+            "resource_hashes_sha256": "c83d7a76dffa5a3d97e932a3eac4602ad82e20babb09d9942504fff6b4db53d6",
+            "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde"
+          },
+          {
+            "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+            "id": "inferrsi.common.skills.diffusers_invocation_state",
+            "ref": "package:inferrsi.common.skills.diffusers_invocation_state",
+            "resource_count": 343,
+            "resource_hashes_sha256": "64b10a60e44276bc4a20b3441f65afd929341e8558fde32491ba27a20fde4020",
+            "revision": "sha256:a8529f2c6bf5f84ce48ba0d6426f1bf1bc6fce172cd0a599b6a149878e3b4be4"
+          },
+          {
+            "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+            "id": "inferrsi.common.skills.generation_cache_lifecycle",
+            "ref": "package:inferrsi.common.skills.generation_cache_lifecycle",
+            "resource_count": 343,
+            "resource_hashes_sha256": "052e200da2a160661cf44647685c70e52feec4b72fc0f2e7b34a10ce54710e03",
+            "revision": "sha256:22411ef79b30ddd4f8b01a7ed5d00fc4fc75a3c9f1e4d31f56f2e440d9b09d2f"
+          }
+        ],
+        "selection": {
+          "cue": "LMCache skip_l1 retain",
+          "method": "weighted-lexical-v1",
+          "ranking_applied": true,
+          "scope": "generated-packaged-skills"
+        },
+        "serialized_bytes": 541917,
+        "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+      },
+      "reciprocal_rank_at_5": 1.0,
+      "success_at_5": 1
+    },
+    {
+      "answerable": true,
+      "case": "dev09",
+      "mode": "default_diagnostic",
+      "output": {
+        "collisions": [],
+        "material_bytes": 0,
+        "no_match": true,
+        "omissions": [],
+        "ordered_items": [],
+        "serialized_bytes": 312,
+        "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+      },
+      "reciprocal_rank_at_5": 0,
+      "success_at_5": 0
+    }
+  ]
+}
diff --git a/evidence/knowledge-retrieval-20261027/results/benchmark.json b/evidence/knowledge-retrieval-20261027/results/benchmark.json
new file mode 100644
index 00000000..5861c039
--- /dev/null
+++ b/evidence/knowledge-retrieval-20261027/results/benchmark.json
@@ -0,0 +1,13545 @@
+{
+  "started_at_utc": "2026-10-27T11:30:36.319056+00:00",
+  "dry_run": false,
+  "entry_python_optimization": 0,
+  "verdict": "PASS",
+  "freeze_content_sha256": "a465367319f3058398eb1f8579ee5d4174b8cb90457793e90788fb5796151057",
+  "identities": {
+    "A": {
+      "corpus": {
+        "bytes": 2883003,
+        "file_count": 458,
+        "files": {
+          "src/inferrsi/data/knowledge_packs/README.md": {
+            "bytes": 14172,
+            "sha256": "b877e67d9cccf0dcbe57419eeb3ff9374e5bd03ff8eec63adef1025c446140e9"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/diffusers/cuda/diffusion/group-offload-transfer-overlap.json": {
+            "bytes": 3277,
+            "sha256": "9bec3d912192efa2f2c853d3fb70bdae21a75875630a02294fcd084b18c30414"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/flashattention/cuda/attention/generation-specific-dispatch.json": {
+            "bytes": 3085,
+            "sha256": "90290d7bc80b1626863e0520c91bb196d93ebfe08190a450e9de59c6505ae8ad"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/flashinfer/cuda/llm/paged-decode-dispatch.json": {
+            "bytes": 3745,
+            "sha256": "0f2986494fcc9b751df71e598fd61caa68e28bc4b0ac401e4cc763bc3f75283b"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/cross-stream-dependency.json": {
+            "bytes": 3809,
+            "sha256": "b829315e55adf8e11f0bd581d5f45dd130d7eb3f81ce7de532f63c7b9c27cbb5"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/evaluation-completion.json": {
+            "bytes": 3841,
+            "sha256": "19d2e506ac4f50895776a83a062059aa7a052175c8e4b21dd0d46e1a5a6716a1"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/post-detachment-buffer-lifetime.json": {
+            "bytes": 4663,
+            "sha256": "18d78c0e542dacb42ca33e5264d92aef80fa06d163fd16d4feafd5ed0827ab99"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/pytorch/cuda/compiler/cudagraph-shapes-and-lifetimes.json": {
+            "bytes": 3623,
+            "sha256": "7fef4122ff794967e61f55d02a4a6b558d7e498d9513607f5ca86db6386928a8"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/transformers/common/llm/kv-cache-strategy-contracts.json": {
+            "bytes": 3072,
+            "sha256": "a2df26052112a9f3fc6242b2bd636ab13991442db57e56646168d6c52032d38d"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/triton/common/kernel/bandwidth-fusion-resource-limits.json": {
+            "bytes": 2941,
+            "sha256": "6e9673d1a600a6ad34820e4325f2a6417f44f0f96d8ed0bcf21466d998f19ee4"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/cache_dit/common/diffusion/residual-cache-calibration.json": {
+            "bytes": 3841,
+            "sha256": "bb9509ad0225a91b52bc6787477581d05f0ecd5e44fa59f08f718eea36a2ac45"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/bounded-windowed-replay.json": {
+            "bytes": 10959,
+            "sha256": "43613225e9ccdbd34c0f7e3e0767887b9bbfd0f38b2c21684677d99e31a432f0"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/speculative-verification-budget.json": {
+            "bytes": 10177,
+            "sha256": "a729e1fc00be44e1480a5dc65bbcdd58df8599681df74dbc651a5a49f27e9028"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/zero-overhead-deferred-scheduling.json": {
+            "bytes": 8191,
+            "sha256": "ed5426af145f066f6183bd316c4073c46f324f5ee83aa9bd616387c0d0e7149c"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/lmdeploy/cuda/llm/kv-quantization-capacity-cost.json": {
+            "bytes": 3366,
+            "sha256": "873997b38981ccc03ff73e6fb84e96022f293a019843efd8c4aa50f27878ac9a"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/codec-constants-and-state.json": {
+            "bytes": 4250,
+            "sha256": "1cecd7905fef5af3ef8952a4d803900a7df91d662ebae34db47cc0dbf352d4e7"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/fixed-customvoice-prompt.json": {
+            "bytes": 3660,
+            "sha256": "ea5e8fdf54cb731b255b73aa0cf47ef53fac103baa2bd5246be45f95d54cb593"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/graph-batching-state-ownership.json": {
+            "bytes": 3950,
+            "sha256": "130741a9f7ac9d366784d6c4dd42860262235caff0c277de1e8a8f8ec933b37f"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/sglang/cuda/llm/hierarchical-prefix-cache.json": {
+            "bytes": 3641,
+            "sha256": "b50ca58df28d4343545d07adf0da23f8f181297b30aba7870e5690630e7d7315"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/tensorrt_llm/cuda/llm/prefill-decode-disaggregation.json": {
+            "bytes": 2967,
+            "sha256": "349856dd98628f94987e014889cd1b15d0e3ae6ab6c576cdde917f28fbafb102"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/xdit/cuda/diffusion/parallelism-freshness-and-topology.json": {
+            "bytes": 3236,
+            "sha256": "3674b5916cb7b384951bf3e449b6796b161878daabcc1d7e7f6d6f8ff861cc17"
+          },
+          "src/inferrsi/data/repo_packs/afd_plugin/pack.json": {
+            "bytes": 502,
+            "sha256": "00042abeabdc9c0e24f1a4edaba47e233df398aef0ad4e181fc70191d383da5f"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-aclgraph-capture-replay.json": {
+            "bytes": 4311,
+            "sha256": "327414443ea9537ae320aa405912c338351bc06cefdc0b99352e1eea36f61dc3"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-device-health.json": {
+            "bytes": 4115,
+            "sha256": "633fa5fdcea1e1e9edda8a218d6a3fde0d12637848348e1e2fe7d1b0bd744a16"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-distributed-parallel.json": {
+            "bytes": 5841,
+            "sha256": "709ca0d061e4d6917fe8fa049fe93165e71dfecf96c7bf264ced25346381f603"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-graph-mode.json": {
+            "bytes": 4135,
+            "sha256": "bb82ab3d628b91be957e557545810ef5c20790be7d43296096f20a21f45ec391"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-model-adaptation.json": {
+            "bytes": 4137,
+            "sha256": "7ea1ab0f0044afff935a91071b36ac09bc46b5ef044fe2a68ab0608256ddd853"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-profiling.json": {
+            "bytes": 4186,
+            "sha256": "0d27058fd8d0ef8b524b7a60d1cb76950622746c94b5a03915e64e6edb4ab3d4"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-tuning.json": {
+            "bytes": 4439,
+            "sha256": "fcb49d71f59d8b39c1eaa03d24e95e2cbfe5f9d7f6e0f22df03ee0d60f6a568a"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-optimization-casebook.json": {
+            "bytes": 4371,
+            "sha256": "05783630bfaf3c7ef99b848865f4bbab50d0b52e80ad04fcd028c624d7638e9b"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-performance-analysis.json": {
+            "bytes": 7200,
+            "sha256": "b6eea897c12d0c6ac9d172089535c82c25eb71c09d7a4972e64812cc4682b5c9"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascendc-tiling-design.json": {
+            "bytes": 4505,
+            "sha256": "e69b6e89dd342ba55ae54b5fac14aa1b91ee2a361da40cb669c287ce519f4517"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/cann-attribution-overlap.json": {
+            "bytes": 4915,
+            "sha256": "8ac01b60f5cd37fd594d6cb5119e4e6525199d017296fc8779dc68a576f5525c"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/torchnpu-device-memory-lifecycle.json": {
+            "bytes": 4349,
+            "sha256": "7a79d0eb2c50381d40086e6db72d4bdd2cec5d2261bfb00cf73a60701a94da2a"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/torchnpu-pinned-host-memory-lifecycle.json": {
+            "bytes": 4431,
+            "sha256": "8bb34dcfbebb48667a94434b036a1a408b970af9890b047a3de325f364ed5f7c"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/assess-serving-preemption.json": {
+            "bytes": 4198,
+            "sha256": "e3d2a1a72d0d980f31fc85deee27db26bc0daa0c709e8ae99e42deb814a4a7f4"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/assess-split-backward-schedule.json": {
+            "bytes": 3915,
+            "sha256": "a8561794786eb0a897d831f00c92b4fdd3b0492316c1b1d8ad13c1f8f75ac815"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-h3-runtime-residency.json": {
+            "bytes": 12169,
+            "sha256": "602dd625874bd4a89861d5bb25881cc63e4b5a7fdedc2867089f0952943f6eff"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-llama-runtime-state.json": {
+            "bytes": 5816,
+            "sha256": "64f6ab8a95172378350323cead08bc6c201c624b711e6242938a6a21a8e83ad7"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-sd-cpp-residency.json": {
+            "bytes": 8349,
+            "sha256": "95ed97b717255be3b817e5c69debd6d41197559dbf21771507d5da79134a72e1"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-sglang-diffusion-lifetimes.json": {
+            "bytes": 7276,
+            "sha256": "5e5889223878c6cf42a49c6f09c60e9d9add88cb7e7d0df050c4e75d88a05aa6"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-vllm-kv-lifetimes.json": {
+            "bytes": 4275,
+            "sha256": "00424005a5eeb8b15f239edd2cb2c71047787e3ecf55924860cb7a78d137b586"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/bound-windowed-attention-replay.json": {
+            "bytes": 7877,
+            "sha256": "98b2291147a796ef59cf20a075ff662998ee6a529f85b49e9c6fb62fe41fad9f"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/budget-speculative-verification.json": {
+            "bytes": 7490,
+            "sha256": "66448f59460c9dffac46e74160388fab650cf8bbbd5bdc95c1465b70daf0c2dc"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/calibrate-serving-simulation.json": {
+            "bytes": 3992,
+            "sha256": "055b70e0b86f65156775b519e4b8cbf06a956a8993b06a66175167cedb78f2cb"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/cgroup-resource-controls.json": {
+            "bytes": 4801,
+            "sha256": "ad406b6dd5d9415bd28d448e64269a388b50e11f9eb7f5a9adb6b0a6246fd303"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/checkpoint-recovery-consistency.json": {
+            "bytes": 4740,
+            "sha256": "ccb85314a831bde58f8cc515580d001b174c11dca0bbcbc6221303acf629046d"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/choose-parallel-training-plan.json": {
+            "bytes": 3901,
+            "sha256": "40afcb41c59988fc1af8fdd00d2340800c54e96847dd0de16ecddf7ac221d1b4"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/cite-check-downstream-pr.json": {
+            "bytes": 2533,
+            "sha256": "dabff14684d5ae9a2b05a9743ff5b76d00a90fdb29319588aba67f5087aad706"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/claim-contract.json": {
+            "bytes": 2469,
+            "sha256": "78f15f6533434fc29adf1df651f7a15eb8ba81733f8b02c710d90102541d707b"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/comfy-ui.json": {
+            "bytes": 6919,
+            "sha256": "1496807d958df055f720ca93e6dcf7f7e8f506f3f12c94c9a2166f789b00058d"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/competitor-scan.json": {
+            "bytes": 2449,
+            "sha256": "c2e97e49e5185710cdfaa08ac1c2b1c19b958e68ebfca885d5aee0ed8c9f8c8b"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dcp-async-save-performance.json": {
+            "bytes": 4538,
+            "sha256": "359a5acf8304a78fc074688676f9a0be62f56f463d0960a544e409f49843bdaf"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dense-performance-evidence.json": {
+            "bytes": 4227,
+            "sha256": "d87f93a9dc9ddc583e3929baaac85f7a425666dd6dd71665933f18344ccec834"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/diagnose-vllm-admission-pressure.json": {
+            "bytes": 9789,
+            "sha256": "a1de40def775b2e840c27330b67e6f37c3c244db1731fab2bd3e8a787143210e"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/diffusers-invocation-state.json": {
+            "bytes": 6109,
+            "sha256": "914d2ffe9776ba931bed780a2a270c1525e3c6cbaf0735b8531064d3180f3233"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/draft-acceptance-issue.json": {
+            "bytes": 2549,
+            "sha256": "6aaa63dbfe6cda982ccd3a4e368527b80fd2873672c000c66d06d28e40a2bf14"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/evaluate-elastic-serving-placement.json": {
+            "bytes": 4334,
+            "sha256": "fab1fa814e789ec2d86587fcf7de7c38174464b9e9ba0cf53ef3fb470053e7c4"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/generation-cache-lifecycle.json": {
+            "bytes": 5972,
+            "sha256": "e5679cdecb722bbd17280913913da8473052ad09e9e7a13b3a9b95463a139ef6"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/host-memory-management.json": {
+            "bytes": 6530,
+            "sha256": "a36aea21169281866b2468860d3ffeb912e61edd571308b02f82939f17f694a8"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/host-overhead-optimization.json": {
+            "bytes": 4243,
+            "sha256": "05fd7f054966b96892980d585352bc1c3fcd47f439cb34268fdf07feee58a998"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/inference-benchmark-design.json": {
+            "bytes": 10735,
+            "sha256": "091bb086c6f0123dbb5db78968c13f846cad135e8969982efe557644b7c7613b"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/inference-optimization-casebook.json": {
+            "bytes": 6531,
+            "sha256": "c512025d7a59934c91665e24752d9393ca9b6a9dfb624c72789294defae93c0e"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/kernel-writing-guide.json": {
+            "bytes": 4545,
+            "sha256": "f1f0ca1d11d3550c9b8c5fdbed8f6cd09b877ac14341f973d58dca02b5d8995d"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/kv-tier-management.json": {
+            "bytes": 4609,
+            "sha256": "6141f61fb6b0d147b704ab7a2627e9babdb53a90e15b659152a3d565967afddf"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/lazy-evaluation-stream-lifetimes.json": {
+            "bytes": 5934,
+            "sha256": "34f47009a8a33f8907bd78f110b95fcd62409bef1b38e1055acb46b2eab285ff"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/npu-skill-factory.json": {
+            "bytes": 11868,
+            "sha256": "24c1d3bff288e86b99ea674e3252ec6f54d9ba944274aef6a7666197afd6788a"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/optimization-search-policy.json": {
+            "bytes": 7101,
+            "sha256": "d7bf71d5491765f62eec02fe5013dc1a5ef56e1516892da48593c6165a86366d"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/os-process-placement.json": {
+            "bytes": 6304,
+            "sha256": "e6e8fe5d7d66f3e98f05762d94441726dfa757ce7fbffd2b35402b3f327f8fb3"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/performance-ceiling-analysis.json": {
+            "bytes": 13695,
+            "sha256": "47c034d96026e1c2a02848d9c9c3caf4024a5033664bfd089ad30d9f23a80deb"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/plan-activation-residency.json": {
+            "bytes": 3871,
+            "sha256": "436e364020cc854fd52a10ee76e4593c7519bb0546c8aa5238fcfd89158e8ab0"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/repository-evolution-analysis.json": {
+            "bytes": 5865,
+            "sha256": "e0af402c3320ebd9bd36c774cb52c0708721a86ca446c05f6e94306888838e30"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/rsi-evidence.json": {
+            "bytes": 2475,
+            "sha256": "c8b4bb5a8c88009ac5b1526e82639715381c940d3c3eb295f0333cae6c2d3ca9"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/skill-evolution-playbook.json": {
+            "bytes": 4091,
+            "sha256": "0e063c2e7686de0c27c061e1aac0eb2d8c4fa83246e32ddeeb0e28126c176425"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/storage-io-measurement.json": {
+            "bytes": 4783,
+            "sha256": "80e589e3a0b67d3a045e03feeadc45e720ceb277e6124c917eec04863eb67419"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/training-input-pipeline.json": {
+            "bytes": 4296,
+            "sha256": "ae696bea8d2b7bfdaa29b6e5bc0bd861abcc0e42e7d9bec6ae20c6e7d6f27669"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/training-resource-budget.json": {
+            "bytes": 6944,
+            "sha256": "ea5e5e864e3aa34bd00be67c8e741877ae01480f039652073b65b71dfcd54c97"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/weight-loading-optimization.json": {
+            "bytes": 8139,
+            "sha256": "7f2d1fbc6b1cfd80c3d6ed8df107d4ca9b97655626ee078c345445a303b1b367"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/audit-afd-p2p-exchanges.json": {
+            "bytes": 6640,
+            "sha256": "9ac45aa784b430ad1ca9ffddecce8dbae9740cfe9c98c10b330e01d350015a38"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/diagnose-attention-bottlenecks.json": {
+            "bytes": 7602,
+            "sha256": "e525d75578de1cc164e6553d01efbd0a014e74646200580f59b1161c030413ad"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/host-overhead-profiling.json": {
+            "bytes": 3965,
+            "sha256": "e9e6ae8c5b9986e83cade3bd029baae112859e01aa6c96f741a264f3bee44adc"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/reason-hopper-kernels.json": {
+            "bytes": 15086,
+            "sha256": "758a649e651d8f9635ac2df45d18043e156c8d54b8f8dbc95f8e68ebcb866a28"
+          },
+          "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/parallelism-and-extension.json": {
+            "bytes": 3399,
+            "sha256": "1253f5f01bcf1498703c7f34960f0c50580df5ba8f140af557a903e296ad81d3"
+          },
+          "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/runtime-radix-hicache.json": {
+            "bytes": 3317,
+            "sha256": "5a0b2aed51649af214cdc1034f57650b7ce195c934a17452ee67a35131f80cdd"
+          },
+          "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/tuning-and-validation.json": {
+            "bytes": 3385,
+            "sha256": "ef4d4f7182da4433d59d72dda0d21b4aeafadd73260be11ab8d4cb36c20fafc6"
+          },
+          "src/inferrsi/data/repo_packs/sglang/pack.json": {
+            "bytes": 620,
+            "sha256": "bce2e0c42796a9ef499d459c37b56d1d56f4b19344841359c534fca29c0359ce"
+          },
+          "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/audio/serving-and-streaming-quality.json": {
+            "bytes": 3440,
+            "sha256": "97fc1d5a11a39a0634dae9e9602a4ec47a4d2348e121076c2d5c836cf0692114"
+          },
+          "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/omni/extension-lifecycle-validation.json": {
+            "bytes": 3482,
+            "sha256": "7da8dc5d2655099de2bcd5f0fa922d9e1bcb0bfd2f571f026a004a7f402a75b6"
+          },
+          "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/omni/pipeline-state-and-transport.json": {
+            "bytes": 3404,
+            "sha256": "0c57692772019504264e39bf0444a69e20cff86f6e112b072c53cf7a4299cc82"
+          },
+          "src/inferrsi/data/repo_packs/sglang_omni/pack.json": {
+            "bytes": 630,
+            "sha256": "8e70336bc2aac0c74732a026337813652b5d5c6e1fc60f5571e9442345077b96"
+          },
+          "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/architecture-search-space.json": {
+            "bytes": 3294,
+            "sha256": "2c5ad72d5bab4f7082fc3be1532c347444cde38f5a84639fe16b22134b68685e"
+          },
+          "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/controlled-execution-tuning.json": {
+            "bytes": 3255,
+            "sha256": "6a2bf611a432213919e1c237687507ceb0ab753cccad0e934dc77f86604805e9"
+          },
+          "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/engine-validation-contract.json": {
+            "bytes": 3295,
+            "sha256": "ef89775bc82965fe87a832c7ec8a5cf55cb9eba4ecfc6279caea9ce3b0410c41"
+          },
+          "src/inferrsi/data/repo_packs/strata/pack.json": {
+            "bytes": 617,
+            "sha256": "643b32defb8db9aa230f624856f4fb4b595b5c23db58d9453091f5420cff71c3"
+          },
+          "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/bounded-serving-measurement.json": {
+            "bytes": 3239,
+            "sha256": "fb00ddac14ef6b06f48da9ba335529bd55d5d83c2a2fe331f0aa1ca16afceb37"
+          },
+          "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/extension-correctness-gates.json": {
+            "bytes": 3343,
+            "sha256": "4ea7083d277df205ea4c361bee87a8207951596e9ef6cfa136cc7d2aeab02f46"
+          },
+          "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/runtime-and-kv-ownership.json": {
+            "bytes": 3280,
+            "sha256": "e6142c1065754716a5f26f42119722908dc8fff63e667c986d57f1227b8610bd"
+          },
+          "src/inferrsi/data/repo_packs/vllm/pack.json": {
+            "bytes": 508,
+            "sha256": "1705d96710249c4514c15d8d011b43c1a1ae7c15fbc9ddf11ef2d0e6b2ab7d07"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/minimax-h3-reference-kv-reuse.json": {
+            "bytes": 4047,
+            "sha256": "80b853fc594d387ab48bed04dca245e0a22b8cdaf7c9b853551ab3bafa302e84"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/minimax-h3-sp-head-buckets.json": {
+            "bytes": 3742,
+            "sha256": "c4d29d1bceb5e2cf86151bf53c9dafb3d83428fb1e9d2a848d852bb8824bf857"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/qwen-image-21-eager-t2i.json": {
+            "bytes": 5268,
+            "sha256": "ef107d43cc448219eda4d64ab961b61cda6350e20d30d94de1ff7c4ffff7e775"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/audio/streaming-serving-measurement.json": {
+            "bytes": 3389,
+            "sha256": "2680f32f28907f6abc7a7e1e3ecab4f46ee412f40f4898ed2f371ffadf47ae4d"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/measurement-and-profiling.json": {
+            "bytes": 4472,
+            "sha256": "c8b25583cd160024c7472a20debea34cb6d4bc79714b4bad2953555adc9f794e"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/production-integration-gates.json": {
+            "bytes": 6112,
+            "sha256": "649b49ff64d3e9e88bd620027dabf9d126ca69bbe91b448f28e845ecbe1f8e2f"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/step-state-and-output-validation.json": {
+            "bytes": 5090,
+            "sha256": "e1ffe28c4fa54af739cdef2fedf4755bba18e078f8380d4fde8138ea712e5778"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/upstream-review-gates.json": {
+            "bytes": 7614,
+            "sha256": "5127709bb85ce31008d19468e1b4fcb5454a71776f69f991841cd9ef7558f9c4"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/omni/stage-ownership-and-search.json": {
+            "bytes": 3427,
+            "sha256": "e75d6c2829ae22881893f5da2ad7edba099fd64ccfb0fc93d30473e06e9d4d80"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/audio/concurrency-scaling-state-and-overlap.json": {
+            "bytes": 25324,
+            "sha256": "cb47fabb232db81ed525cbb54371becc7e4a28b2bae9d8492f60dfb6f31563e8"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-adaln-schedule-cache.json": {
+            "bytes": 4253,
+            "sha256": "ddbba3bb72b5789938595a82d3280e861f75f6ed54f05a1d2b7dcb1089d30f63"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-cuda-mxfp8.json": {
+            "bytes": 4468,
+            "sha256": "e4c538f03f4ab90a4acc8e3d8721bbd226aec1ee9bde45757a7c27d3d970821d"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-fp32-modulation.json": {
+            "bytes": 3625,
+            "sha256": "abb8d94a1b025f0e2c8e085275e96ee85d68aeb27cb12d0918e4647f8255b693"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-qk-norm-rope.json": {
+            "bytes": 3706,
+            "sha256": "4c1c4acfb9cc0baa90bc28fed1903020ea1ea38755b84315711795db45e48bb6"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-ref2va-teacache.json": {
+            "bytes": 3835,
+            "sha256": "8aeb3f99ac4227cf8b3a5cece13d938b19082e265a557eddae27b7e495bcc22e"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-vae-stacked-tiling.json": {
+            "bytes": 3508,
+            "sha256": "8cc18eacbbfd8fc4d3a6b4152c4b584d559fc9b15377da56ef072e542d807c93"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/pack.json": {
+            "bytes": 566,
+            "sha256": "58c971e9dcf6ac15752629ed4d9426aeb5b5d857fc37b5bcf921aa2e81c0cdd5"
+          },
+          "src/inferrsi/data/repo_packs/vllm_rlt/pack.json": {
+            "bytes": 649,
+            "sha256": "6097041b8d026740060e527f49b1b4e288315b1c3227f7583805b86d8fa415f7"
+          },
+          "src/inferrsi/data/skill_library/README.md": {
+            "bytes": 18819,
+            "sha256": "ae583f3578425df1dc91ba32b8b6543c0469c9a817ebea873a77051d6b13e1ce"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/SKILL.md": {
+            "bytes": 5494,
+            "sha256": "7581b56336d07213a87d01477793ae1a8acf0b7cf876fce36c9a761ee32c1ff4"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/metadata.json": {
+            "bytes": 1429,
+            "sha256": "68ac7f64b7ab947fc2db664bc3990f102c1cc5539b42350aa5818d0bbf17d86c"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/ownership.md": {
+            "bytes": 2081,
+            "sha256": "40688cc44183027efbf2fca8d3548c299cdf2e33fef54f76ab3549353eb9bd1b"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/source-map.md": {
+            "bytes": 4635,
+            "sha256": "10562a75683bd9bd07ec2e37a1cbd0f93f2534d6f85b55a566adb2d97485215f"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/validation.md": {
+            "bytes": 1477,
+            "sha256": "5a0b854b87d091d7d505b0bdd5c03330052507a851ab6edf35b93a6675f2d4f7"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/SKILL.md": {
+            "bytes": 4010,
+            "sha256": "a95569bcc0da6e715fd74783a6a1e0489e92b680c582f15dafbf98df589d505a"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/metadata.json": {
+            "bytes": 1519,
+            "sha256": "9c2293c8956508de987c886c7224d673d22912e7e915b026b05abdd287e97401"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/execution-lifetime.md": {
+            "bytes": 2147,
+            "sha256": "c4bc41e08aca68f661325f61e6f0db1760a079dc362f295c847bdc374fee3bf6"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/source-map.md": {
+            "bytes": 5918,
+            "sha256": "798b9cb89ed6b885eb47464ca9f137d2802a54571b9327e631d7742d3e5abd5d"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/validation.md": {
+            "bytes": 1608,
+            "sha256": "ebfc00bf7d6ab0a1dc415953ae84d56ccdc713d5614aff85e433c4a843793bff"
+          },
+          "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md": {
+            "bytes": 7322,
+            "sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845"
+          },
+          "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json": {
+            "bytes": 587,
+            "sha256": "db7ad368394ccbfdeff6a16e3423c300d51305b1f95ff5c3936df91738a8cc35"
+          },
+          "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md": {
+            "bytes": 5314,
+            "sha256": "22afe4f4d77f441096cc657d55d763473bfcbea8fb531c8e493fe49ab84955f7"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/SKILL.md": {
+            "bytes": 6745,
+            "sha256": "d494fc7e539cbb7388189bd7b3f04bb5c4983fdd7a82484348aad1cee7f741f2"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/metadata.json": {
+            "bytes": 558,
+            "sha256": "71c68431663ed2336cccc985a354adab6b13b4cfd6123a18a5d8d53c28268509"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/references/source-notes.md": {
+            "bytes": 13004,
+            "sha256": "6ffbfb68cc7e04923b4fbca2a73a7555b7c3b45a6f40819bb5f3213ef05cf88d"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/SKILL.md": {
+            "bytes": 5977,
+            "sha256": "03829039860ece35519f2e4c339a48b5014ec25e07a2c150a3efb6a3a4e35b4a"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/metadata.json": {
+            "bytes": 559,
+            "sha256": "80d80c15514919826fa72d6f6cf26b76db8accd85d21187fc239232ca454e7bf"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/SKILL.md": {
+            "bytes": 14703,
+            "sha256": "4ce9dcadd1ae3c7a8cc633336760d0a3c382f354f1b7028e3a10d28fa0d0c9fa"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/metadata.json": {
+            "bytes": 1403,
+            "sha256": "b6cb25879b1bca5f5af08b48776f2e664cf5dcd11a0efb1c4f56879e2afa9006"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/references/awesome-ascend-communication-tests.md": {
+            "bytes": 9388,
+            "sha256": "0cb27a3d24409171bb414d7ebf0c49ad62be5f46482f29624695496ba9eb3e63"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/SKILL.md": {
+            "bytes": 5839,
+            "sha256": "9203cc5a14a1d65df5fbf275734b93be023e06527459b2f79864f28a338cb44a"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/metadata.json": {
+            "bytes": 536,
+            "sha256": "72b41e7845915d34b09219624e82976386f3125816293540c2c38143fe6b477f"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/SKILL.md": {
+            "bytes": 6440,
+            "sha256": "8fdf46fb2743b67a66bc069d22a45305a222debfb606777bc451b4b8f31d9cc6"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/metadata.json": {
+            "bytes": 536,
+            "sha256": "be681a3206e5e183b4c9670c9d7784bcffce18c891449e5b15f94ca5f1324b9a"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/SKILL.md": {
+            "bytes": 12988,
+            "sha256": "401981cf7828edcab0c9ea882f7ecd83badfa3c2a95006e912cf99d67b3d6a12"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/metadata.json": {
+            "bytes": 760,
+            "sha256": "efd37bb23acb500dad4ec8ebe7ddfe5c3f196374052825f923ac85da8445f06b"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/SKILL.md": {
+            "bytes": 12236,
+            "sha256": "38b151bc6b4e9734c6fa41efb7a30343559400c0c75bbdd26bd8a968b8ea4592"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/metadata.json": {
+            "bytes": 557,
+            "sha256": "04d2002fc8a0a7becf86b0b73daaf2da8716ab466d7f988da43d41a10f26680e"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/SKILL.md": {
+            "bytes": 9172,
+            "sha256": "647f5e9a288f2c4e1617ba53df59d9d827df749eb4fe86182d21a9c41175caf8"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/metadata.json": {
+            "bytes": 536,
+            "sha256": "f425f7c1abbd616387273e8932c6bd02f4b1da012f099286ad4171e0f6037b22"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/SKILL.md": {
+            "bytes": 9880,
+            "sha256": "a057b7f86a24f34a8f84abbbb4910d58d4e50ff47060e1c404600017e3febe89"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/metadata.json": {
+            "bytes": 2297,
+            "sha256": "9e705625c8308013c287576b0247da6f6116d817a4a535cbecff80abd6ea35ee"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-CANN-2.0-LICENSE.md": {
+            "bytes": 5262,
+            "sha256": "89b1c114df62cfd183004f38d18d07f186436c3b37203dd69c3e84becf9d2282"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-stage-evidence.md": {
+            "bytes": 11789,
+            "sha256": "f56e4125325c9c1b9235a1d64fd732861c7686444492422daff44ba8dd648be4"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md": {
+            "bytes": 11860,
+            "sha256": "93aa1b775ced304cebfe6c4dec0503ddb2260ecb3cb01edc17ce329446c5ce2b"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/metadata.json": {
+            "bytes": 552,
+            "sha256": "f6d94f104cb70969024b2f4542078a240e200712c66279eca8b94f60bee1dd71"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md": {
+            "bytes": 8972,
+            "sha256": "0e6f094dc554abea12e07def50f20b233b47127779a8aefad595b4ab1efd24c2"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/metadata.json": {
+            "bytes": 569,
+            "sha256": "acccc8eb86de0df6caf78a9b3c6db76a2a686adc4a5ba0029a4d62d50ff9a8f8"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md": {
+            "bytes": 5989,
+            "sha256": "8428df7157af024fa17cfe0b3818ed54bd9f468306dda46eb8d4f0edf4f3f230"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/metadata.json": {
+            "bytes": 579,
+            "sha256": "c650dce88c07b636d68e063dbda4d319f6f7f3ecbcdd5e47f393cbfc3af7c71d"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/references/source-notes.md": {
+            "bytes": 11470,
+            "sha256": "80cba4d3a884e7ee2618326cf74b9f9c3a34e4d9cf443707f6bbe673f2d5ba65"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/SKILL.md": {
+            "bytes": 6159,
+            "sha256": "23e93ea7c98f99743509fdc04139d78ed2e7123c63e80b366c6beaba7eb70d0d"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/metadata.json": {
+            "bytes": 581,
+            "sha256": "97836a38465106d816a9ace894a7b277d6d330f9accade8f8493ce6f0fb57ef9"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/references/source-notes.md": {
+            "bytes": 13493,
+            "sha256": "435193590b4a2ba32e2e4c6fa9d7e8f889528d412cbbb48e35ec5433f7dab384"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/SKILL.md": {
+            "bytes": 7137,
+            "sha256": "a25738dbb3b9daa15c5827cb536a20fe13f7c8b9dd14aa3fb7cfa7e132879431"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/metadata.json": {
+            "bytes": 1007,
+            "sha256": "3abe626a6e4c73c943429f609ce2c62aa77aae7e1726475ebbbdfbe1794f6f7d"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/mechanisms.md": {
+            "bytes": 17376,
+            "sha256": "fe08b2292fa587be3874774ddf41063b84c579a0085df422ec6710d8aa90b8b7"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 4024,
+            "sha256": "371b01ee201ef51fecbfd6da892d517aaba295930b9395025f03f34e21c030a1"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/NEXT_GATE_LEADS.md": {
+            "bytes": 1623,
+            "sha256": "35b8b58ae5db02b2df51ccbe5cdf870a7383d17f345f9ee5eb7b0a4146dfc74b"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/README.md": {
+            "bytes": 2888,
+            "sha256": "e0b1ec1b39c10067f5ddce58320851ad089568c4b2296a24777389097cb5b388"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/io-reduction-can-plateau.md": {
+            "bytes": 2671,
+            "sha256": "5cc35f514224ac6e6efd2da6db221bba468b2f4db66807295e67b240bcdfdba5"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/overlap-and-fp8-counter-signals.md": {
+            "bytes": 3692,
+            "sha256": "80c8a60a6df55f3d84fab23e03567d8b38e23bcdbd63cc09881e46db1e2888ab"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/partitioning-after-fusion.md": {
+            "bytes": 3132,
+            "sha256": "4d4cabe0991eab554dcfbd4fa9c8199e61fa4767709b0c4cd646e2588bb398a2"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/knowledge/attention-cost-model-boundaries.md": {
+            "bytes": 5223,
+            "sha256": "5ed8f888443e38ccb72545b6025447aab7b92936f78c151fed680e582126239b"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/sources.md": {
+            "bytes": 9069,
+            "sha256": "53436e0de4e6dda2d29426591bf61f8526378cdc6f9ec7967b88b1bcd4c61e6f"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/host-overhead-profiling/SKILL.md": {
+            "bytes": 11199,
+            "sha256": "326e566fbe30d3ba04e4b50c3e87aad9df037f0ac4ad6fe129165ae97dc5f91d"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/host-overhead-profiling/metadata.json": {
+            "bytes": 527,
+            "sha256": "76e34c1c64aba473a9e126b032a776417caa0f913f6ac3cf4dba3a5d123b1ee0"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/SKILL.md": {
+            "bytes": 4905,
+            "sha256": "cec4e3aa1eb64c31c54b5f5ca40af2b61b7919d94be4174968581d5f752c0538"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/metadata.json": {
+            "bytes": 7453,
+            "sha256": "7520da1cc97412631a600957ecb18aa5a7a4461ea21e76c237330180193ed898"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/fp8.md": {
+            "bytes": 6287,
+            "sha256": "31d9c211c11c2224007abbdcd4059913c4a54833fe2d2f926411c5238fcd3501"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/cutlass-bsd.md": {
+            "bytes": 1547,
+            "sha256": "42fec630f410aa308f70a51a89fadcd19586fa620f9831a32bee528a9a10000e"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/deepgemm-mit.md": {
+            "bytes": 1065,
+            "sha256": "6c73a30c14a2938abaa6afe98ebb19e6d81dac5bd4ace0cb05c61c20afa37b85"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/kernelwiki-kda-mit.md": {
+            "bytes": 1063,
+            "sha256": "84aab57158a914f448b38d8be29dd842b634a08c192d844f416a256beed4ae94"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/notices.md": {
+            "bytes": 3240,
+            "sha256": "876ec016644bf5331b3f257985e126c0e61980e3bfd82569b7b0c3f6b2c6c64c"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/resources-and-scheduling.md": {
+            "bytes": 8919,
+            "sha256": "1ce25b426736072435a1ead9a2a8e1f56df5d0abb8198e8c4fdcb34321db4964"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/source-inventory.json": {
+            "bytes": 120506,
+            "sha256": "1dc09c895c970a0c9708158bcc6dd65cdb736274fd2c35d44542c70407d3849d"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/sources.md": {
+            "bytes": 9004,
+            "sha256": "6b74b9e774f2a80db0860510d9e6940d36284430f29c1e124f115a9eaf25833d"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/synchronization.md": {
+            "bytes": 6843,
+            "sha256": "4322f6b3c0e4368bf8c192b4ec752529767f50d19c0c38dad8fdf9cd781e41d9"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/SKILL.md": {
+            "bytes": 7348,
+            "sha256": "fd1f5a7e6a55714730044155f5b38c1d700383d4e16c3e91963bbfbe1c773af5"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/metadata.json": {
+            "bytes": 572,
+            "sha256": "19e40763fedc20990651bb989c32a7693306c5c54af46c3d4be77fd4f87b6c03"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/SKILL.md": {
+            "bytes": 23213,
+            "sha256": "2c807f3df2da2f533d25183de5771f3d412058d16410994292ef8f48c49f9b3a"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/metadata.json": {
+            "bytes": 1684,
+            "sha256": "9ac6af135f0b6d76a76bc201b6bbd8ac9afe5201dd1e2ea82c0ac4826de401da"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-LICENSE.md": {
+            "bytes": 11357,
+            "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-native-offload-readset.json": {
+            "bytes": 6267,
+            "sha256": "cc6c2011be746002ba3ff96ccefab1809863748ddd17eb102b75e7b0e134fe55"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-native-offload.md": {
+            "bytes": 8966,
+            "sha256": "6c2a72b5ba3708a01d229e0576570837a617e6a42dbb0b5312a26a65a6c31542"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/SKILL.md": {
+            "bytes": 16072,
+            "sha256": "36ef88826a32ac3147b869001004ded17a8d5a491e25bfe105005ba39ea9d1eb"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/metadata.json": {
+            "bytes": 539,
+            "sha256": "ae71c389dfdb1a19d267feecb4c6f1fbb4c9afad501ff6d8fab7e38b29bcc37a"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/references/deepinfer-zero-overhead-scheduling.md": {
+            "bytes": 10894,
+            "sha256": "f452bf40132c9195d739f1c86b3aa18857138c7ed0f1861af5853d93b9955924"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/os-process-placement/SKILL.md": {
+            "bytes": 10327,
+            "sha256": "2003492e4d89fde0443b180d74933e764bd1979435e109c8d26e20eb8f6e5053"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/os-process-placement/metadata.json": {
+            "bytes": 1399,
+            "sha256": "ec02995c075f0a540921f57b4846f2cafe334eb90f44389f17c3c50457c3e9c5"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/os-process-placement/references/llama-numa-thread-readset.md": {
+            "bytes": 6617,
+            "sha256": "748b254b3b043a5ddb0f480eb4a566854d27eedd7b111499863e9bf0a5a6951e"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md": {
+            "bytes": 8813,
+            "sha256": "63b883c10a8d3e7d918abfc7175b72e2b401ff1e02697f94b8382ec1c8f369b7"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/metadata.json": {
+            "bytes": 558,
+            "sha256": "0a41c7b00978332b54ee1e4a7c1ade88ecde3e961128f24d1b4f7708caff5ff5"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md": {
+            "bytes": 6402,
+            "sha256": "0a5f48d9352c5bf460775ef47f9136829978811f4a16e84983e299782cae305d"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-serving-preemption/metadata.json": {
+            "bytes": 537,
+            "sha256": "77ce5ba16245cbd80d46d0abdf80aaf26fc0924fa2d2a5624ff430de68a1daf0"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-serving-preemption/references/source-notes.md": {
+            "bytes": 30405,
+            "sha256": "f29fcecfba0490008a354c11b42403983fe165e56ff907ad61cc6040d2b6407f"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/SKILL.md": {
+            "bytes": 3181,
+            "sha256": "1e64f28ea71672fb53c56f6108b87864738c6d03d6dcd32eedaef9e0625e96e3"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/metadata.json": {
+            "bytes": 478,
+            "sha256": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/references/source-notes.md": {
+            "bytes": 10548,
+            "sha256": "61b1609f71c37b61d136b8c4d39f793de87cffb1cf31c8ec737a1c3fa6fb1b4e"
+          },
+          "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/SKILL.md": {
+            "bytes": 3884,
+            "sha256": "ac773c6f1c110a664dfca28a3ae9129b10c07614e60ca3640dea776220e7b352"
+          },
+          "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/metadata.json": {
+            "bytes": 648,
+            "sha256": "261ce9f644ea6eebb90c830f41fb8b320d2b401456e5bd8d494095a44b8620d0"
+          },
+          "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/references/source-notes.md": {
+            "bytes": 8415,
+            "sha256": "9696d3c0ef96b6611787446bd6f96d4d6255ca8d107105de8c1d9af83edb9fa5"
+          },
+          "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/SKILL.md": {
+            "bytes": 6416,
+            "sha256": "6b7db1eb2780dec27858cb7aa0a3c093c5b5dfa4c30261aebfeaa255213ff688"
+          },
+          "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/metadata.json": {
+            "bytes": 582,
+            "sha256": "4dc8659806539b4784ea9348a01d4194ac70d0bb297a916f90ff6419efa5c328"
+          },
+          "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/references/dlrover-persistence-replicas.md": {
+            "bytes": 8398,
+            "sha256": "055f4357f79026fcbcc766bbe77db610376ec353623f8898f5d14a30ee69e7bc"
+          },
+          "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/references/sources.md": {
+            "bytes": 7984,
+            "sha256": "9425b1aab963a2c8d35f3db897bd15ee69699bf5b0d0c8425c3c8bba3e9a0233"
+          },
+          "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/SKILL.md": {
+            "bytes": 3068,
+            "sha256": "25d04aed9782b2a8ff1186828becdc7fce93c4418039ceaef9b8dcc5ec9bcd24"
+          },
+          "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/metadata.json": {
+            "bytes": 478,
+            "sha256": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7"
+          },
+          "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/references/source-notes.md": {
+            "bytes": 8521,
+            "sha256": "9f9184d46b19312138a6e805e439e43dc0594a0cc6b46675255a582dc946ca5b"
+          },
+          "src/inferrsi/data/skill_library/methods/dense-performance-evidence/SKILL.md": {
+            "bytes": 8305,
+            "sha256": "53cc1867a13f54d1006d50517dea8848869ca50df3f0cd610b2424cc8d18c933"
+          },
+          "src/inferrsi/data/skill_library/methods/dense-performance-evidence/metadata.json": {
+            "bytes": 707,
+            "sha256": "a0d1e3b17140295510a764045d1b222e19be3a3406e6d1e7dddaad49ff474490"
+          },
+          "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/SKILL.md": {
+            "bytes": 6667,
+            "sha256": "82d0099d3c33706f46c657d4fa7f76d3d486e019b36760d1de001921328a60cc"
+          },
+          "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/metadata.json": {
+            "bytes": 547,
+            "sha256": "37379a4d4a058ca46da2b20a920dc01be9ec11e557c32b2332702ec76027f4b5"
+          },
+          "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/references/source-notes.md": {
+            "bytes": 44796,
+            "sha256": "ea416831fd7d0874e06d0a76fb0c7196df53072b717673a0cdf635db6bfd917b"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-benchmark-design/SKILL.md": {
+            "bytes": 8974,
+            "sha256": "19ebc1f567aadf49f7a534a18fdb629c72c0e4ce043d360983136491cb72bfab"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-benchmark-design/metadata.json": {
+            "bytes": 5923,
+            "sha256": "18310402a8ade03fd48ac687f79ba9d447871e90337de84c0e1bc15c0b827306"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-benchmark-design/references/benchmark-commands.md": {
+            "bytes": 8466,
+            "sha256": "0898424e0bff3a6246e1eeeda6abd6c5bd974c047fe32f3bbd5ab9d31ce17cd3"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-benchmark-design/references/task-benchmark-selection.md": {
+            "bytes": 25086,
+            "sha256": "0d4bb080a754e26846e81680023db0dff5e6a4c96be4327c411c3b72ac4ef8e2"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/SKILL.md": {
+            "bytes": 6644,
+            "sha256": "89a6a82c692f7f5192f4d7f4c3a5ff08a4df3b1c4bdaa0aafb7e04292d8aa5b4"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/admission-drain-occupancy.json": {
+            "bytes": 1354,
+            "sha256": "3319e23bbe2bd40e2d45fb89dc87bdc92d4df2b5ddd8086fcf84813b43885b7c"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/admission-drain-occupancy.md": {
+            "bytes": 3134,
+            "sha256": "e88bbac5d253a8e7e3a3d5e21047f23f2970c11ac23887d4303ab94fca52dd27"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/drama3-rejected-directions.json": {
+            "bytes": 1459,
+            "sha256": "2cc0a526fe0509551607a03e7626dd3d7844e643c4afb194837f75cddead8b12"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/drama3-rejected-directions.md": {
+            "bytes": 3557,
+            "sha256": "6f143bc363a3290fc008c08e57cb075f1776d1c8cca8a6a59297d5b1fbbf61bf"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-ascend-rejected-directions.json": {
+            "bytes": 1293,
+            "sha256": "b3d538f79c56724ccb2895cbf50cf2e6b0fec3979be62a684a8230cd932dff55"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-ascend-rejected-directions.md": {
+            "bytes": 2564,
+            "sha256": "be396983812262515899a3e2ba4f01b23ff92e1b836881882c7580372b8ae552"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-b300-rejected-directions.json": {
+            "bytes": 1196,
+            "sha256": "146ea1564a46ed03092c7dc5796b318fc05b04b1821105ee00eb7fa7fa1333e2"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-b300-rejected-directions.md": {
+            "bytes": 2656,
+            "sha256": "cce2930f544a0f78f21797fba4d8ded298e4a81f76c34d7977b0e31d361bc738"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/metadata.json": {
+            "bytes": 535,
+            "sha256": "fb55067fb45970319787068a3206110641cceaf0ccbabd706be6d0b3d2afc2ed"
+          },
+          "src/inferrsi/data/skill_library/methods/kernel-writing-guide/SKILL.md": {
+            "bytes": 10029,
+            "sha256": "01fb1712d7eb988c5e412f02340d2e48b4f8564fcf4ab232dc5897869d0d50a0"
+          },
+          "src/inferrsi/data/skill_library/methods/kernel-writing-guide/metadata.json": {
+            "bytes": 653,
+            "sha256": "0189d184337662056fc6952c7beffc1bf6154c4fa173ae3e3e463bd22d70f36c"
+          },
+          "src/inferrsi/data/skill_library/methods/kernel-writing-guide/references/akg-affine-tail-guards.md": {
+            "bytes": 11941,
+            "sha256": "113deb2db10b171db3425bb3f1792bee763706b794c8cc2e222b4cc0d0114118"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/SKILL.md": {
+            "bytes": 12914,
+            "sha256": "882e7850b9b49ad692a2d11618701afe0f4e83727ea78df114be2029923c725a"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/metadata.json": {
+            "bytes": 1403,
+            "sha256": "9f1fd8a2af52bc964021e21f16ecbdb891e08f2057370e700d67f86605d333db"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 3597,
+            "sha256": "157e74ba180426b279b548a729769f152bf18f9a78ffca4ebfcaf26b7d0c5292"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/MECHANISMS.md": {
+            "bytes": 5875,
+            "sha256": "18bb2f5d7e52773c516420a142edfb531d0b9c822a2934d60f3234ac1cb9d40e"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/NEXT_GATE_LEADS.md": {
+            "bytes": 941,
+            "sha256": "16a264c7159f9c2c700ce8ba46e63f3e636b277c987da4970f6d2fa085fcb053"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/knowledge/adaptation-interfaces.md": {
+            "bytes": 3893,
+            "sha256": "5e761957cb5ecd93926e6ffc13deee7c830c6e8f0ee3c3846e9dd3b50dc74446"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/references/source-notes.md": {
+            "bytes": 13370,
+            "sha256": "54d83c3713045d59a48821bcf93598374ba4a630e3891ad39498f919317c4fb8"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/updates/optimization-search-policy.md": {
+            "bytes": 4238,
+            "sha256": "d3db44fa9a6831b86a02c638b323ed943e69fe256866213f10d68c72d762879e"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/SKILL.md": {
+            "bytes": 15324,
+            "sha256": "c5f78961c23e3c8d659f5675b522518ab06e1f2a8ea8bc74c5d428ffb41693d5"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/metadata.json": {
+            "bytes": 3669,
+            "sha256": "60b5b24ead590b0fb222398fba47822bf6723690da683e256a450bb5c0eeb806"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 7349,
+            "sha256": "083e462e0a204dce599362d46712fcfb392d053c5cda56216e22e28d4b8fa0fa"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/MECHANISMS.md": {
+            "bytes": 4003,
+            "sha256": "742c249082ceca230326f4407a40130fe624553b796ae5a157a8a4c15a7e1398"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/README.md": {
+            "bytes": 1204,
+            "sha256": "17b2d1c7c14124d59ac8a0d5ad4200ad2803d991e784a1a98cd97188cc364999"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/READSET.md": {
+            "bytes": 3937,
+            "sha256": "66459ce393858b2b2160f1e2973373bfc4b9902a12644631934b4ebe3140296a"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/STATIC_CHECKS.md": {
+            "bytes": 1111,
+            "sha256": "cb0aa7d148352ebb7b1678fc42a557a099575054716a341dfa25b6a998029064"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/cases/decision-cases.md": {
+            "bytes": 5491,
+            "sha256": "781a675a6533e0a18c044aaaee0d010da7013b6504654b8b8292a35176db801e"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/govern-optimization-experiments/procedure.md": {
+            "bytes": 3823,
+            "sha256": "ce036a4d8a0908da204515c47e2ff9623548d09de264fa65bd58fedce52a78a0"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/govern-optimization-experiments/references/source-notes.md": {
+            "bytes": 21659,
+            "sha256": "f9a8254e45473ddb2b73e576957b6d3db83c62b99dff1922f4d1e606e569559e"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/knowledge/bounds-predictions-and-decisions.md": {
+            "bytes": 5297,
+            "sha256": "5780d6eb5257612ec45deb94f85f334a6ecabf7e466d1678706041f6549bda79"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/redirect-cross-layer-optimization/procedure.md": {
+            "bytes": 3648,
+            "sha256": "6593a63fce9c474b515663a560c94b7df048dc5f6e77a95b8e54e4a64cd10a09"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/redirect-cross-layer-optimization/references/source-notes.md": {
+            "bytes": 26469,
+            "sha256": "6ff3a7bc31f1172db92aedaf7df1f1aac688bf769227b583a4cd82bfad52da67"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/agentic-kernel.md": {
+            "bytes": 41353,
+            "sha256": "7190573b85f94a4391bb0d887ccbd37a31b3f08bb0531114ad7afd72fe1567a7"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/flashvector.md": {
+            "bytes": 9830,
+            "sha256": "3926369acd88ff3c370bee5718e74dd15f3219c3e4f6311095e30e917bffa4cf"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/inferencebench.md": {
+            "bytes": 9294,
+            "sha256": "21f47111197a0b7f8d6384ea918d139dc8318a62ca3313ce060048271f50b639"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/prior-coverage.md": {
+            "bytes": 3254,
+            "sha256": "35d47ac078591221fcb666219bcd824f9400fc17aafecdc0499d8336f7178238"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/vidur.md": {
+            "bytes": 7736,
+            "sha256": "d38602af5219f5feeb49e6cd6bf53d66098c77fad141dae676f1d8a6743cf25d"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/roofline-contract.md": {
+            "bytes": 7091,
+            "sha256": "419ec4bcaec45ac579a133bce083d4f48779f26e664d7a86f720fa65f343b1d1"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/roofline-example.json": {
+            "bytes": 2015,
+            "sha256": "597cd3205df043cff3ae9d15924c77934e8d8eb300c00c4840f6bb2921f90faf"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/scripts/roofline.py": {
+            "bytes": 9004,
+            "sha256": "524de48a6764401695fe1444f1fa3d38f363069b0cd2fba4715a86ca4b28a5d4"
+          },
+          "src/inferrsi/data/skill_library/methods/plan-activation-residency/SKILL.md": {
+            "bytes": 3507,
+            "sha256": "542d8a1e6963f5019986e1fbd79d3a247a045a51487c09313026087ac457c15e"
+          },
+          "src/inferrsi/data/skill_library/methods/plan-activation-residency/metadata.json": {
+            "bytes": 478,
+            "sha256": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7"
+          },
+          "src/inferrsi/data/skill_library/methods/plan-activation-residency/references/source-notes.md": {
+            "bytes": 13907,
+            "sha256": "5ff52655cfd06fcf970ed2a7d4d4fe1c31115b102af2d41497de0c74848ed2b3"
+          },
+          "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/SKILL.md": {
+            "bytes": 13800,
+            "sha256": "7e20d3390f2501263e5e298e026c47cac16c234f2397833131e15ae5d34c0b97"
+          },
+          "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/metadata.json": {
+            "bytes": 1903,
+            "sha256": "d0e4d6950250e4cec43d82e23324c9301a925d4c22be30b4d2713d790cb6504b"
+          },
+          "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/references/capability-presence-audit.md": {
+            "bytes": 11461,
+            "sha256": "c21eb4a6aabe811a49a6540a34c54dbcfdcf7efae3052a79b54034c7838026a5"
+          },
+          "src/inferrsi/data/skill_library/methods/skill-evolution-playbook/SKILL.md": {
+            "bytes": 9448,
+            "sha256": "9cd37f21fe0d862069e4798edca8a04d59f8b840e55d521901594a63dddd9ba3"
+          },
+          "src/inferrsi/data/skill_library/methods/skill-evolution-playbook/metadata.json": {
+            "bytes": 527,
+            "sha256": "92b36338b9257e2e41b95e4370694d7de52ed84bc7971056b3bfced07b0993d8"
+          },
+          "src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md": {
+            "bytes": 6865,
+            "sha256": "64bd2c6b795d966045d1d3cca55976d3d49b08b06df7c7e06a70c2824cf55b7e"
+          },
+          "src/inferrsi/data/skill_library/methods/training-input-pipeline/metadata.json": {
+            "bytes": 587,
+            "sha256": "80659099fce6a6ec05e9084b75452c877a5812e30f60a78f31ef8c7efcdc1bac"
+          },
+          "src/inferrsi/data/skill_library/methods/training-input-pipeline/references/sources.md": {
+            "bytes": 2637,
+            "sha256": "60e0f3e88655d8e1d5eb4797f39ca4370deda01614f265b8d2ea4559d8ac020f"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/SKILL.md": {
+            "bytes": 7200,
+            "sha256": "bdfa5a4ef37d9f303c3a23cd6f98c21137b97ac5ed58e0ed5fac6d71d33e52ef"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/metadata.json": {
+            "bytes": 976,
+            "sha256": "d42b614af9cee6ee0d4a25dc5d57c7b98438feed1e807a7c9f1755109a86e9cf"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/mindspeed-historical.md": {
+            "bytes": 3384,
+            "sha256": "eb97f935ed30517a8fc9edec5e9cddc04f5d679d156c16d75bd29c534593360a"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 7368,
+            "sha256": "c55c6ad2131ef788866411dcc7056f221420ebff30c95c6bc919b09a4818cb73"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/cases/paper-grounded-cases.md": {
+            "bytes": 5649,
+            "sha256": "19bd25b1a944adb2351ebc6b1d795150ea033041a44f6cc261e881677812a101"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/knowledge/offload-delta.md": {
+            "bytes": 2026,
+            "sha256": "f8740fc5c8bbbe5ac13ff1cfec7ca499569d0b46e8cd56cf1f5a44c3a5e72242"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/notes/training-memory-scheduling.md": {
+            "bytes": 18660,
+            "sha256": "9aa6d3e01452dff95c8a8940ff62ead8c146ce215c9175b78c9f2ef508a7b2de"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/references/source-readset.md": {
+            "bytes": 22946,
+            "sha256": "2a5c166767f56fa47e7eba7386f1f36da8cb73702e5b5bbe14d6a456ca9ebfae"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/sources.md": {
+            "bytes": 2752,
+            "sha256": "6d248f58af8b350fc66c22898988d6f451ac60d03bcd805cd6e84df5e55174ac"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/stas-decision-notes.md": {
+            "bytes": 4086,
+            "sha256": "f1a993e6c4ca9a035a2effcaf5500081f227b0cc56cc885a6b22fda0041fc744"
+          },
+          "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md": {
+            "bytes": 25969,
+            "sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213"
+          },
+          "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json": {
+            "bytes": 2418,
+            "sha256": "798cbff6f10b7c24b188080ebf233b188b0e98ef2c6fdb5e00e1fe3eff727c0f"
+          },
+          "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md": {
+            "bytes": 8656,
+            "sha256": "5a39f14b102b0cee304541e011aa31e44e6b13527d1bee9987c3f3b66c064d3d"
+          },
+          "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md": {
+            "bytes": 6392,
+            "sha256": "6c640f28ee3f0253a17b81461c865cb9f947367002cb75bf72550c4572d02e97"
+          },
+          "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/SKILL.md": {
+            "bytes": 7181,
+            "sha256": "a46110891eca860915e314d6921a22856d631453b22b17837fe30262d6780535"
+          },
+          "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/metadata.json": {
+            "bytes": 2175,
+            "sha256": "4e6100dfb9fc7bc00b19915f4fd38ad509704f2a609c658adf4d632d4e05b6c1"
+          },
+          "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/references/afd-plugin-LICENSE.md": {
+            "bytes": 12747,
+            "sha256": "0e8566ea362276878e1f38f97da49964dbea301c15d4aa9d727f97a4c39fd943"
+          },
+          "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/references/source-notes.md": {
+            "bytes": 17240,
+            "sha256": "5073d87450cb60d1b245308dff8807e73749c07de976d50f2fa26694fad81a72"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/LICENSE": {
+            "bytes": 35149,
+            "sha256": "3972dc9744f6499f0f9b2dbf76696f2ae7ad8af9b23dde66d6af86c9dfb36986"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/SKILL.md": {
+            "bytes": 4507,
+            "sha256": "c993bd6ff8e9b5f82f39f175ec8242341136e4ee0808af414e1d94ad3e196c5e"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/metadata.json": {
+            "bytes": 2032,
+            "sha256": "74742851a002a51c208362cdcd2a59fdab5d1734d8492b1f93fbec3cf21e0841"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/references/graph-cache-invalidation.md": {
+            "bytes": 7679,
+            "sha256": "614c9a4e6943d24d7d7be7d632682ce8b1dcc239124a37d9dc1d487113772eac"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/references/model-residency.md": {
+            "bytes": 7191,
+            "sha256": "014c68386e7cdfdb7b4f566836e8b2ff61bd5a5a506e7b4723f83fe80c5d4de7"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/references/repo-provenance.md": {
+            "bytes": 5648,
+            "sha256": "ca996ef515af42a3ce304747fa3f12a31c00533deaace5ba2ce21aefa0a62e2e"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/references/validation-boundaries.md": {
+            "bytes": 3523,
+            "sha256": "1913dedbeb558d872d6a0e015cc11d822910c3cb259a236db29fb60ad8a64263"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md": {
+            "bytes": 12294,
+            "sha256": "4703beb6e5194ef2ff2c0b5ec69f64e35417a4bc0dfb50a10c80a920b77ef3b5"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/metadata.json": {
+            "bytes": 2333,
+            "sha256": "014be24d0e669f4acbc0717dc323eec784026cf185186619d5a19d6d84496131"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/source-notes.md": {
+            "bytes": 12036,
+            "sha256": "fd917bd5ac80bee9c9f4525f8e0300f6d67dc144e2a16632af42a67f3d53d098"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/validation.md": {
+            "bytes": 5402,
+            "sha256": "60e71e70c5519bcdf6c171eec2be2f342e8801bedd746d6005f11adc646dfaff"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/SKILL.md": {
+            "bytes": 11202,
+            "sha256": "f157562d1edc905cb4fc07246567e3c5b844efb9fbf5fe39f0d40553a1b1a307"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/metadata.json": {
+            "bytes": 2156,
+            "sha256": "4d68bf4c1767eaee12ade963bbd551d0e61d1bcd81a984bb02fd3aae46965919"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/references/source-notes.md": {
+            "bytes": 11633,
+            "sha256": "6fa955022f16cd598f44fd6521901d1cb868b06cec545cabe2b82769a641b5d6"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/references/validation.md": {
+            "bytes": 7196,
+            "sha256": "673d93f6700052974b340fab25b3438e0e6344960569ac9abec7d7720f662288"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/LICENSE": {
+            "bytes": 1077,
+            "sha256": "663e51edcd9a3a8f858018ce6924ad2895a24c269c26c39ad75325bcd0f7e5bc"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/SKILL.md": {
+            "bytes": 5432,
+            "sha256": "fdb4d9483c6efa836f9e59617da605fa9c2186e11c2041b05d5c81850ba0475d"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/metadata.json": {
+            "bytes": 4923,
+            "sha256": "fe626a77af3f084f36fd73ffdb6f67976e7dceeafb94ba22078ef03c3a96cfae"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/THIRD_PARTY_NOTICES.md": {
+            "bytes": 1774,
+            "sha256": "a0d77db7d2aca0eda811ab3c386cc32b3143f2c2609a042cf4642954f0f6fb09"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/minimax-h3-apple.md": {
+            "bytes": 12059,
+            "sha256": "57ce01754ba96c378f13a91305ef3d418b926f99b2753f0af99315572bce063a"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/minimax-h3-weight-inventory.json": {
+            "bytes": 10664,
+            "sha256": "fbaacaeec340d3ce893556ee864405a0746af4d13530a38cb3d5aa9d60929271"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/phase-residency.md": {
+            "bytes": 9408,
+            "sha256": "2836a550eb2be7f5a862f51ec7b135ebd9715bf3c4faa84a7ecbc9d9c11f5d8d"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/precision-and-streaming.md": {
+            "bytes": 9440,
+            "sha256": "7e166c89193f4c7aed852ec95346a2de2b5008dac002e2e1862be1b3e57f6093"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/source-identity.json": {
+            "bytes": 7747,
+            "sha256": "730f8dd23648a3f8f47b9726cc613d893a79d1817d515eec609a9b06d8a9ff90"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/source-notes.md": {
+            "bytes": 9448,
+            "sha256": "7304bf9369efc157dc7aa40686dd828cc14db45c49442e908b3d3cc4b329eb86"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/validation.md": {
+            "bytes": 7880,
+            "sha256": "0754873a758e048480938314894e25b504ae0c269549312e7f8e6020d83fe294"
+          },
+          "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/SKILL.md": {
+            "bytes": 9338,
+            "sha256": "95e81b6bc096a954fa73139ea5bc001d1efa6c99747161da62f012b494e5c013"
+          },
+          "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/metadata.json": {
+            "bytes": 1718,
+            "sha256": "1d30b31dc8352fc6a8d7b8abbef8118c840e47e5b90795dae0c758bf9717278c"
+          },
+          "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/references/source-notes.md": {
+            "bytes": 15234,
+            "sha256": "3d12a401e9decb55a2f4f3b2bfa0d360ef5110f63a0469dfe285654e78054190"
+          },
+          "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md": {
+            "bytes": 10812,
+            "sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27"
+          },
+          "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/metadata.json": {
+            "bytes": 572,
+            "sha256": "c221189b8d85fac3bd41ef1c7cd560f1aebc8cfa0381c182d1d571a7169be829"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/SKILL.md": {
+            "bytes": 3788,
+            "sha256": "bf1177cb9985c12a9bf11edb64771018e1e314c4713920de8fd4a989ff70115f"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/metadata.json": {
+            "bytes": 1299,
+            "sha256": "d86cecb615e93ba0f6be09d1ead102cd5a17068c07e1be2549ac5cd8f7519279"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/cache-stop.md": {
+            "bytes": 2026,
+            "sha256": "bba6b5c73261065309cbcf11b1e591eef48916f0c79ceaaf74d98a9fa96b0992"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/generation-batch.md": {
+            "bytes": 1733,
+            "sha256": "f16d3c57d45e63bcaffd3871aa41c76972ee11f7dbf1e12b45d9b170303cad9b"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/source-map.md": {
+            "bytes": 6358,
+            "sha256": "91d26181f246307f2113f94b12ac9587d689b654dcd216e9f14b11cf5ebb650c"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/validation.md": {
+            "bytes": 1665,
+            "sha256": "73b5dd347017baba519ecc85415aead9993c1751d116b9ba4f07689f58c5f783"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/LICENSE": {
+            "bytes": 11352,
+            "sha256": "a54c8553df5d4313d293d1aef3bf2f13952bda8b7c7a1fdd729b2bbe27ee6015"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/SKILL.md": {
+            "bytes": 2304,
+            "sha256": "d75f5b2881b240ec2d9d4f5cc8e3f7d53e4378cb01732a2fb9699da2901ba38c"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/references/repo-provenance.md": {
+            "bytes": 1778,
+            "sha256": "5e8dbe168a8885e209d4b4532944ffa7fead658b23fd6588054d57b22476f63b"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/references/repo-routing-metadata.json": {
+            "bytes": 511,
+            "sha256": "6820707cb9bd877d1635928cf5a0efe38e4fe37dc04036dc70d53e479ed9bf7c"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/references/snapshot.md": {
+            "bytes": 3488,
+            "sha256": "0fb5d776349de845dc314f16ce73a06c9caf5db84dc1c293327d3b3ca1337c80"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-architecture/SKILL.md": {
+            "bytes": 2355,
+            "sha256": "766b2d66d2056dd779abd9e492cb3af6e25899298f44964f565509f130a7cb24"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-architecture/references/runtime.md": {
+            "bytes": 7734,
+            "sha256": "36bba9874fe054acc5217f9c265f5ae309c83943e0d46228370ae1bcdc69421e"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-extension/SKILL.md": {
+            "bytes": 2193,
+            "sha256": "9af38bca840d0004fc2ed049141476fc40fd40dcfa843f38a404a7a314c19425"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-extension/references/integration.md": {
+            "bytes": 5775,
+            "sha256": "6ce9383ca17bacff60a180f8fb8c43ea47eaa37117fb8da6051ad86135d228ef"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-serving/SKILL.md": {
+            "bytes": 2335,
+            "sha256": "990b6bdebfe22612de49dbebcb16810e8abef94d70691aa883b7937c85210eb6"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-serving/references/recipes-and-tuning.md": {
+            "bytes": 8918,
+            "sha256": "94afb762bbe185f58606154b3f0a5af56e7089cb84ea94e26cfcb9f93c95439f"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/LICENSE": {
+            "bytes": 11346,
+            "sha256": "1495e1e757ef4d0925a2350563cf5754bb23c51701a8ec4fb3c5cdcbedae6747"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/SKILL.md": {
+            "bytes": 5466,
+            "sha256": "ae72d1f0cc750699bf79d7a138a97219209bca7b7e5f19cd427d306245eaf841"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/LICENSE": {
+            "bytes": 11346,
+            "sha256": "1495e1e757ef4d0925a2350563cf5754bb23c51701a8ec4fb3c5cdcbedae6747"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/SKILL.md": {
+            "bytes": 5738,
+            "sha256": "b0a4ba4751087126143ef86983798deb0ed00e24b6ff06bf6ef76a9df6e274f2"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/metadata.json": {
+            "bytes": 2844,
+            "sha256": "6ed57dfb4f25ecb34b81f2188b055e05d172ba2581e6a3b605b814b2f733a4d6"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/references/source-notes.md": {
+            "bytes": 19512,
+            "sha256": "d0fcfd1177ae2f351041c5b2abba7f5121cec133006d9e4dca8f74ddb567332d"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/architecture-and-memory.md": {
+            "bytes": 7735,
+            "sha256": "798e747022b7ec135cc4861c2c9ac44fb14b503bb77a0068ee7fbea494b764a7"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/native-candidates.md": {
+            "bytes": 3961,
+            "sha256": "abafcc9a34df196a6365145554887b8d37afb36c6e236c92e50a4a06c8a4ec93"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/parallelism-and-extension.md": {
+            "bytes": 6324,
+            "sha256": "e3be7f4b51264ac68682a64d264e4d96857d5bb9f9d99cae5f327c56984a9e6b"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/repo-provenance.md": {
+            "bytes": 2274,
+            "sha256": "534d95ae8d0b29bcbc8f6707b3d38fb964e31f5b66ba13517a9a6b8f309bbb38"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/repo-routing-metadata.json": {
+            "bytes": 625,
+            "sha256": "b2b879c1cc42ddc2a5b95ae24115eb891562fa9a6e96d151f4ddf6f7719fe73c"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/source-map.md": {
+            "bytes": 8945,
+            "sha256": "c2adad64ceb0c305cdc4859c03012b3e0a4ab26cd0cb289bbd76e192b0cc3794"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/tuning-and-validation.md": {
+            "bytes": 10832,
+            "sha256": "49b3343b3292cc098a787de9d24d968b444d0ef4f73fedb5d8183390e8f39c9c"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-parallel-extension/SKILL.md": {
+            "bytes": 2316,
+            "sha256": "18bd54ebdd629e1ba6dfa589f391c7cc3c5355785c2ae82f741b64ee8022464f"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-runtime-memory/SKILL.md": {
+            "bytes": 2154,
+            "sha256": "33b1a65aba00fca819d2b6cc42202e7708e9155e4d8f4373a2387f4ee2f2fa3e"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-tuning-validation/SKILL.md": {
+            "bytes": 2018,
+            "sha256": "744b2ef9122101d3f50cbbb40fe020e2b3d3888adbe2d86041041e7c7ce424fc"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE": {
+            "bytes": 1062,
+            "sha256": "b53fa08f515cb5a6fff7b9fd8fcd0961a4b80df29d74372d25e1f9171aa042ee"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md": {
+            "bytes": 5504,
+            "sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json": {
+            "bytes": 2202,
+            "sha256": "f7b7581a3f44ba9774480c559408fc66d6a4d6d1dbbeabe4529f1eb7d4dd5ad6"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md": {
+            "bytes": 8136,
+            "sha256": "7b537cb972347dd68e303960f142892346d8b19b884e4d08016ea36a23bdaa7a"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json": {
+            "bytes": 3617,
+            "sha256": "af4830370b6eddc2d176a9ae14da4ab56a90b0fe048a01a89ee051df4365d8e9"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md": {
+            "bytes": 6589,
+            "sha256": "b67affbd7d237d82b7f51be49f1537e9820fe56ec70a71ee76809f9e599ca693"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md": {
+            "bytes": 3394,
+            "sha256": "5dcf5fa7bf734fbe13214f57d2dfb40ff8215d9cd34e072a9e615bf103b82fc2"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/LICENSE": {
+            "bytes": 1093,
+            "sha256": "949b9b0aada7476006469e58ee4ffb0e0f68c2d9137e90a356757b61a09d37f0"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/SKILL.md": {
+            "bytes": 2813,
+            "sha256": "0e5daeb6409c916be45ecd7ce2531ea5f1c60a4a53e2b0ed25dc1e2cdc5c17ed"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/design-space-analysis.md": {
+            "bytes": 27736,
+            "sha256": "11ecafe71530ddd4839268ea1b24b3361a3895012a705859ca06dae29fa72630"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/repo-provenance.md": {
+            "bytes": 2913,
+            "sha256": "05f17d0ae996ad6c12314bdd29f3427574f6aa611ea60effd18598a0f386638b"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/repo-routing-metadata.json": {
+            "bytes": 332,
+            "sha256": "a10ba49b7e6565109b566c105e17551d042075a1ff649c6f371134e3b7402c74"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/trial-format.md": {
+            "bytes": 2291,
+            "sha256": "92d8cf38549268ae4eef382890662d03e18a7dff858ad45ee5eb6f47eb39c8c0"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/troubleshooting.md": {
+            "bytes": 2079,
+            "sha256": "7301cfa5a5f6cfbf9576c3845a96c6b4de8f8a6509f2330970a8e2e4669b3d5a"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/scripts/summarize_trials.py": {
+            "bytes": 5170,
+            "sha256": "6a243a08da15659ced6519f49f52a49a675bd9e952740c3ab03dd21964eefbe2"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/design-space/SKILL.md": {
+            "bytes": 1826,
+            "sha256": "857bea3c64db45dc035728a41523142ac65778b41f63092148b8ab936f971d76"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/design-space/references/design-space.md": {
+            "bytes": 24231,
+            "sha256": "0a428dcd12d5290b554e9da6795a110f94d98f128771602e1ea5f608048b135e"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/tune-execution/SKILL.md": {
+            "bytes": 1145,
+            "sha256": "cdf51c537e7237296d52ef6cda9f2e4c9d9ce3fc87265a369cd9eb7497c53512"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/tune-execution/references/tuning.md": {
+            "bytes": 11310,
+            "sha256": "7b69aa7872d60dd3fb91e41b77a632572070dcf80dd5b5f6dc1722d8540177b1"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/validate-engine/SKILL.md": {
+            "bytes": 4613,
+            "sha256": "90bfd975e25f18895ce0e85f6f35d24149ce026305a4d479aea4924510053460"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/validate-engine/references/validation.md": {
+            "bytes": 16895,
+            "sha256": "dae846f4c47c68731b96ec1d50cb008c03871c72078f1c5b0f4a9423be9ba397"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/LICENSE": {
+            "bytes": 11357,
+            "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/SKILL.md": {
+            "bytes": 4945,
+            "sha256": "df3e63dc28de2dac2aa92824fd88719159138650d96487332e0effca3f0e2a80"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/references/evidence.md": {
+            "bytes": 4689,
+            "sha256": "ae0d51ba56b34ed81fba34de8fd3685bebf2f160b188fa53719936002881d220"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/references/repo-provenance.md": {
+            "bytes": 2221,
+            "sha256": "43bbfe431434ca4ad6157ba35cbb9ce2d89e4065f427abfdd5f372f51ba0c9c9"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/references/repo-routing-metadata.json": {
+            "bytes": 342,
+            "sha256": "7784b4d98d096718c4f4cc42ce86d5ec113c469844e12de4690b0fccca59d329"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-architecture/SKILL.md": {
+            "bytes": 1336,
+            "sha256": "fb3ff50a7b9122a2545a97ef5179011237c4c4ca214dde7236ad7bf79a7c8623"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-architecture/references/architecture.md": {
+            "bytes": 5600,
+            "sha256": "db845eb4207ddeedb8ae381f1b91704a42cb53ba0ee6c7ee581510ba1a412190"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-serving/SKILL.md": {
+            "bytes": 1343,
+            "sha256": "8ab3209b838635395b22deca6aa879172016531d45c2374eb8e498e7f71bb40f"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-serving/references/serving.md": {
+            "bytes": 5049,
+            "sha256": "9bd5eff300d9755137a7600c8f7a7753bb22d09c8853e08062d7e535662ba8a3"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/SKILL.md": {
+            "bytes": 2227,
+            "sha256": "93c83bc4b799b32f50ff54b870cbbde050a790af16d9b7b062039c4987c80148"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/references/abort-acknowledgment.md": {
+            "bytes": 8547,
+            "sha256": "e95ae5b2cc410917ddf5e9066890d0bef9264104529bdca5ac12bdf56fe80ae3"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/references/validation.md": {
+            "bytes": 4632,
+            "sha256": "1542061f9c76fff2acb6529a58238301484b0e23ec0c697d2fdf129e200ea973"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/LICENSE": {
+            "bytes": 11324,
+            "sha256": "50e6751797c50dedd75ef1b8a0d9e42f5f8472e9fbce91f34718e9f97b0c780a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/SKILL.md": {
+            "bytes": 2863,
+            "sha256": "38e522b8abeb0dc140a2194340432ed41dc8b569d33b63d5d4c59aa1092cd2d1"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/SKILL.md": {
+            "bytes": 6664,
+            "sha256": "313733a65cc48442b18dda1c4665139a689963a90e1032a2235bc6f40e3a8e06"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/metadata.json": {
+            "bytes": 781,
+            "sha256": "b43fc3a7ccae59c664fcdf9ff77962edcc07c2a3df8b59324db551642bd23e23"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/references/source-notes.md": {
+            "bytes": 13853,
+            "sha256": "3665b034594fc8cc767c408002f4e16f8803369242405027c7f37ce6c9ecebe9"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/SKILL.md": {
+            "bytes": 6565,
+            "sha256": "5fb9ea1d1373e110d9b506f69963c4857338b7dbf5e38ec9a49b59a0ff683ed0"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/metadata.json": {
+            "bytes": 1372,
+            "sha256": "78459416734f61a8faa819d9a1a3ad7b9a985e8bf8e149813c604d15be6bb92f"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 9006,
+            "sha256": "82b24be82451ee121892f61575d0cf4ecb9920813a25ccb26cce2b45dda6d8b3"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/migration-is-not-free-memory.md": {
+            "bytes": 2145,
+            "sha256": "2ac168f2725cba7fc92dfcc67c01bf2258dff6a2dbe1ddb063bc60c175d82d0a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/parallelism-change-with-state-cost.md": {
+            "bytes": 2069,
+            "sha256": "5accc81086e4500d66f894ef6af9951e1cbb76941f6c631aab5c5561a786d8d8"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/preemption-under-kv-pressure.md": {
+            "bytes": 2630,
+            "sha256": "462200a6d0b193ef01f32374a463e0f2838ca8edf33d9ab898def65ffd0b490a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/knowledge/scheduling-state-boundaries.md": {
+            "bytes": 7488,
+            "sha256": "d4ef29869959b835a5dcb06ddb2fd8e6cff49bb176994dc1d46eb9a852f6228d"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/fastserve.md": {
+            "bytes": 9289,
+            "sha256": "9b4e275843000ef453d508d9c3c9c3f8ea1de5545fbd01d7abe74baa8a4931a7"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/llumnix.md": {
+            "bytes": 10181,
+            "sha256": "f252ce33ac2db017701ca1d49813c5ec1996c4c92900cf35bb6bda7fc59c0f18"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/loongserve.md": {
+            "bytes": 12422,
+            "sha256": "7bc6d1e256d461b04acc8ed856cf99108faceda9cde729ab95bcca596fe019ca"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/orca.md": {
+            "bytes": 8826,
+            "sha256": "f5a3b5870d64226408a8fcda5d69483769d823f8a5efd7f11698c8da483ab195"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/fastserve-readset.md": {
+            "bytes": 7173,
+            "sha256": "9866bd0c78b2b638998b3b632f6bae26aef032462c81abee336d309af5b07831"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/llumnix-readset.md": {
+            "bytes": 9386,
+            "sha256": "3a2819d65ad24087c08530492ef39d8ea046ed1f780d53512f1d23e2b3578487"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/loongserve-readset.md": {
+            "bytes": 12490,
+            "sha256": "f1a383e7bdbbd5a5a472d7f6e1292c8a17d4910ff736ca5a6aa422e1bee74f8c"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/orca-readset.md": {
+            "bytes": 4796,
+            "sha256": "a497ef23b7e60c0fbc15d95b7e07cd20547cfe56f37c24dfe6ed70d79ef7a3cf"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/source-notes.md": {
+            "bytes": 9575,
+            "sha256": "84033fd514773f10c0bddd9914fabb980497ba2b308f226a922c3e1888413827"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/references/evidence-map.json": {
+            "bytes": 14384,
+            "sha256": "dd8bbab5d285ed3a1a19d2f06ad84e8c444251690d8f673def3e2681765ef6c1"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/references/repo-provenance.json": {
+            "bytes": 1313,
+            "sha256": "4e705a964697712e950e02bedc41320806a41311b7f4dcf32f923d60b1d2e919"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/references/repo-provenance.md": {
+            "bytes": 3261,
+            "sha256": "6e841c38c35ff2cd6ff8139cb1a6c8f75701647897f1089a94a5c62928609941"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/references/repo-routing-metadata.json": {
+            "bytes": 432,
+            "sha256": "a20b54f682eb52311e445ac62b95d1935c001296f5cf23b7764cb04477d535c9"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/SKILL.md": {
+            "bytes": 1835,
+            "sha256": "2a945b5fddacfb90d1cd7cb2b2f767067d641061423a65e1c19afecec7f3c06a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/references/capability-boundaries.md": {
+            "bytes": 6560,
+            "sha256": "c8081d03f04f7481080c46c7dcf728962729bc122f49f9502d50cd1e3d04bcf3"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/references/runtime-and-invariants.md": {
+            "bytes": 8265,
+            "sha256": "afa1f967545021f36bc1b59dbf6aa2c367f4d5ece4f08dce755e2017cf50dbe5"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/extend-validate/SKILL.md": {
+            "bytes": 1731,
+            "sha256": "495679e5b99341bf80daeaacd2d4c2699aba16bc16c447cbd8d2fe17f7559798"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/extend-validate/references/change-boundaries-and-validation.md": {
+            "bytes": 7624,
+            "sha256": "f7fdd2d699f09d96fbf5ca9382f9b56113c870db2688d8e1118eb8163519fc4a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/SKILL.md": {
+            "bytes": 1812,
+            "sha256": "0ef4d2e3a96cbb277557e169a3fc8edf0f0504aad03259771b432a33ee008f78"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/references/commands-and-measurements.md": {
+            "bytes": 5667,
+            "sha256": "76ab51083b14595c49fa1c255cb7e5ebb996f14930a51ab3cbe864cc334b5f80"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/references/experiment-decisions.md": {
+            "bytes": 7644,
+            "sha256": "242593c0d386e57104b57a09bc7db97c2bbbcdb905bb68a397cfe612e07fb083"
+          },
+          "src/inferrsi/data/skill_library/skill-index.json": {
+            "bytes": 7514,
+            "sha256": "8ce35e18f7629260d67a95617b9918eecff73ed8d9370dd3d4595c725ada9aa1"
+          },
+          "src/inferrsi/data/skill_library/workflows/cite-check-downstream-pr/SKILL.md": {
+            "bytes": 1093,
+            "sha256": "485932cdf3b1fba5dc69be73e75d38f86a7b79442d26e82a296c963a7598d226"
+          },
+          "src/inferrsi/data/skill_library/workflows/claim-contract/SKILL.md": {
+            "bytes": 1075,
+            "sha256": "3af1dbbe81d91cb05fc2ef15a2ae25d7592a08558e1a5535f49c52f33a488c55"
+          },
+          "src/inferrsi/data/skill_library/workflows/competitor-scan/SKILL.md": {
+            "bytes": 1088,
+            "sha256": "0ede58ace65a9324a538b4816add1684f230ba536d4364fe5a5dd2c0bbed3776"
+          },
+          "src/inferrsi/data/skill_library/workflows/draft-acceptance-issue/SKILL.md": {
+            "bytes": 1230,
+            "sha256": "ce983acfb7d7d3272775e7cf00c92e4ddadd97511b602575a76e5ab7ade01b56"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/SKILL.md": {
+            "bytes": 5681,
+            "sha256": "364beed2ca617e71f4f955bbd024740d411b94eeb9557d7e055d4463cc22caba"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/cross-repository-mechanisms.md": {
+            "bytes": 17050,
+            "sha256": "28efa1b25f548763180b70d8d0ddebb7d4e7227df189f871fd233fe1d4881303"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/import.md": {
+            "bytes": 3463,
+            "sha256": "c8a57db94768cc6cd4193932def75ba67ec453e56ee0e54f176fb82f1af53091"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/model-aware-preparation.md": {
+            "bytes": 9056,
+            "sha256": "787c017aade68ee1133b979d636fa24062302cef6174c001171c963947ee8da4"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/overview.md": {
+            "bytes": 8099,
+            "sha256": "7ed0a966239b9481b82dd349245cfbfe966c030fc916771c91fd0692464ffc56"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/repository-source-scope.md": {
+            "bytes": 11967,
+            "sha256": "85202ca209171c50573c2fc209acb1e937d8211e1652e4ca9c6cc4bce04d03eb"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/source-manifest.json": {
+            "bytes": 1613,
+            "sha256": "0ce1678803d1ede6b485fb6b4b1c88aba909056ee6e3b5af691899ceb2056240"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/SKILL.md": {
+            "bytes": 6502,
+            "sha256": "74cd863df2e80dc2bda1713d24207809173a3116f30325e3487e1e3517cfec1e"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/metadata.json": {
+            "bytes": 3236,
+            "sha256": "12fc9272bd5fbf02dd9b897d8ca40984b72ae54f20492ace5304711bbc4d71c3"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/COMPARISON.md": {
+            "bytes": 4267,
+            "sha256": "38587678a96581f5c2dc697d982127343565504bffa191352811dffcc165aa23"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 5557,
+            "sha256": "fbc4a96b6f6c477706eed83eaa4fcc29a14f881637629d6c004601ed049f2523"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/QA.md": {
+            "bytes": 1853,
+            "sha256": "3f3785ec39a6938a5d1f2e874e99153fe92c0e666dd3e78a499c8b47d891346a"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/README.md": {
+            "bytes": 2151,
+            "sha256": "221785e91c45588f82b3caab6d52f1ea1d102b16139c20a0802921e2789e6d28"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/UNREAD_LEADS.md": {
+            "bytes": 1038,
+            "sha256": "6c1d87ebe491d27666703149eb05ad582c7e5aff77f0ac5c11a7b89c946f96ee"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/cases/methodology-counterexamples.md": {
+            "bytes": 5213,
+            "sha256": "f9fc21753ea41eef77ca020d78d031e945e5dd835b5d09c5ff0e53749ceb2ca4"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/knowledge/evidence-routing.md": {
+            "bytes": 5282,
+            "sha256": "406d5d74c3fef4c1a23d495492e21a084c3c46495a1a84b385703ed104358874"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/learning-to-commit.md": {
+            "bytes": 2847,
+            "sha256": "9c2d1d738e896d05c00f53e6adf14c9df2501ecc3f0020e71b5a1c2192547086"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/protocol-details.md": {
+            "bytes": 6520,
+            "sha256": "41298c6bde3a28f0712fa06a18bbc4ac59aea6d17cc9eb1318744f4121e3bab6"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/readset.md": {
+            "bytes": 7086,
+            "sha256": "6f953d20d8958c68339c9ceeae348604d0fea3e9c6ccb9b64a63d3b210426a22"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/repo-to-skill.md": {
+            "bytes": 3151,
+            "sha256": "5ebdac2a2bbbe6614f9d4096bc9895bd685acdf76414e2c3c1fb7b50d57c738d"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/skillopt.md": {
+            "bytes": 3040,
+            "sha256": "148075daa67e19f4b4351710498438d17c279a9bce56b2d01cab26ff3be33e99"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/source-manifest.json": {
+            "bytes": 2431,
+            "sha256": "716cd49ed01824ee9f6b6fc55ebf2b24fdd5d5a3e677f6e28e34a472616bbe25"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/trace2skill.md": {
+            "bytes": 2917,
+            "sha256": "230d13b77ba819b5e3edee147523e8cad6bfce95afccc140fb4946d6c8747a2e"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/updates/bounded-updates.md": {
+            "bytes": 7630,
+            "sha256": "d02c5bd0109846d1e1987de074810321cb4af277cd3f66ab2ac268aadd5ee142"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/skill-factory-design-20261014.md": {
+            "bytes": 9465,
+            "sha256": "fb8ce1ad83b7a6ddc657e89ee2eebb710873b213cfd85bdd6d823c8691b3cb9f"
+          },
+          "src/inferrsi/data/skill_library/workflows/rsi-evidence/SKILL.md": {
+            "bytes": 3680,
+            "sha256": "a55610874b658b86a39a1a049efb4e4eab0140f6579ed74dd7718a4ca8cedaca"
+          }
+        },
+        "sha256": "6e19365a0ad8e2951af0f3106f93aa545b4ad00b6e0f0317a20f32065b6a05fb"
+      },
+      "environment": {
+        "filesystem_cache": "not controlled; process cold is not OS-cache cold",
+        "implementation": "CPython",
+        "logical_cpu_count": 8,
+        "machine": "AMD64",
+        "perf_counter_resolution_seconds": 1e-07,
+        "python_executable_sha256": "3470f7919170d235d7e6079691462c4b217745ec67ee612e745730e46d98f238",
+        "python_optimization": 0,
+        "python_version": "3.12.6",
+        "release": "10",
+        "site_initializers_disabled": true,
+        "system": "Windows"
+      },
+      "git_head": "4817f92121df73fd9cd5fb4a319e3c56a68b4a5b",
+      "knowledge_record_count": 108,
+      "library": {
+        "bytes": 2294958,
+        "file_count": 342,
+        "sha256": "be8726296e6186601692dc2d95d99bfbba76f0b4c758de3a88a9b2a523eaa37e"
+      },
+      "modules": {
+        "inferrsi": {
+          "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+          "source": "src/inferrsi/__init__.py"
+        },
+        "inferrsi.catalog": {
+          "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+          "source": "src/inferrsi/catalog.py"
+        },
+        "inferrsi.cli": {
+          "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+          "source": "src/inferrsi/cli.py"
+        },
+        "inferrsi.knowledge_operations": {
+          "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+          "source": "src/inferrsi/knowledge_operations.py"
+        }
+      },
+      "packaged_skill_count": 60,
+      "python_sources": {
+        "bytes": 274444,
+        "file_count": 53,
+        "files": {
+          "src/inferrsi/__init__.py": {
+            "bytes": 102,
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47"
+          },
+          "src/inferrsi/__main__.py": {
+            "bytes": 48,
+            "sha256": "935a1c1166b0c1ea35a82256345000bf2c73ded718d77773bc27a71ecce28f7d"
+          },
+          "src/inferrsi/_skill_selection.py": {
+            "bytes": 5790,
+            "sha256": "4640399f4013425589f6702e2db844820619ba35967d1c3519978eb307168440"
+          },
+          "src/inferrsi/arguments.py": {
+            "bytes": 12200,
+            "sha256": "4456600cf4a362cf069594fe36db46932cf5cb14ddd37683aaf3538f5cf76b4d"
+          },
+          "src/inferrsi/catalog.py": {
+            "bytes": 6855,
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5"
+          },
+          "src/inferrsi/claim.py": {
+            "bytes": 805,
+            "sha256": "5f24c4fa91b48f0bb4d4554531607763fc8fc108e90f332816f06cfb520b5cc1"
+          },
+          "src/inferrsi/cli.py": {
+            "bytes": 14205,
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88"
+          },
+          "src/inferrsi/contracts.py": {
+            "bytes": 119,
+            "sha256": "03000cb9246ce8c0967147533f7ad25df918ba2674839801453ffb870f496490"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/scripts/roofline.py": {
+            "bytes": 9004,
+            "sha256": "524de48a6764401695fe1444f1fa3d38f363069b0cd2fba4715a86ca4b28a5d4"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/scripts/summarize_trials.py": {
+            "bytes": 5170,
+            "sha256": "6a243a08da15659ced6519f49f52a49a675bd9e952740c3ab03dd21964eefbe2"
+          },
+          "src/inferrsi/evaluation.py": {
+            "bytes": 183,
+            "sha256": "482c40710d428206c59fe3696703248cd86c4b3c7d19ad41017c1369bd6d780e"
+          },
+          "src/inferrsi/evidence_paths.py": {
+            "bytes": 859,
+            "sha256": "cc46231e0b8334ef5742defdda20e0093e22b30c86af59df1fb3ae4b5caa70fa"
+          },
+          "src/inferrsi/execution.py": {
+            "bytes": 755,
+            "sha256": "f43d8cbb4d43a72a7ff8a10b29a5fb9cf4cb22af54dd35943283c843fba0da92"
+          },
+          "src/inferrsi/harness/__init__.py": {
+            "bytes": 177,
+            "sha256": "1006e4713003f44bfb32b2e385a918e1652b9e9833801e4555674a71e57fb0d5"
+          },
+          "src/inferrsi/harness/evaluation.py": {
+            "bytes": 4160,
+            "sha256": "9b00b2e93959f16b7b76bb4b257df7dcf48520a06fd6adae96e9a165a05e6e80"
+          },
+          "src/inferrsi/harness/improvement.py": {
+            "bytes": 5777,
+            "sha256": "3faa0f5928a707a98f35f45d1ca18e226c898941e6a07b3829eebeb53e70aeb8"
+          },
+          "src/inferrsi/harness/reporting.py": {
+            "bytes": 3566,
+            "sha256": "38c58f02c44e8c8aeacf46527e142be0e3142a83e4b2693246d0ca76c0ddd7c3"
+          },
+          "src/inferrsi/harness/runner.py": {
+            "bytes": 6567,
+            "sha256": "a566307b9b586655941b627f428b5121ed38d7ac80bccd1e1799e443fb0de911"
+          },
+          "src/inferrsi/harness/scenarios/__init__.py": {
+            "bytes": 79,
+            "sha256": "4e97913c806077f0cadfeb29d986d2fa75b55ab8766cb86e6e3e9a195a9a1f5a"
+          },
+          "src/inferrsi/harness/scenarios/external.py": {
+            "bytes": 1016,
+            "sha256": "84d5f6920501a9d882cc4fffbc721d2051fdbf7c90694bde9cbfba3c2b866304"
+          },
+          "src/inferrsi/harness/scenarios/media.py": {
+            "bytes": 2914,
+            "sha256": "67d979a3ed6f13a8e3e1e40a304f78e5a3e6c2556b3aa8a11c41156745d7050a"
+          },
+          "src/inferrsi/harness/skill_evaluation.py": {
+            "bytes": 11093,
+            "sha256": "b710b8e53758757efd68c703449eb4bab7eb7ec50ae3dfc52a010cdebd8c3bf5"
+          },
+          "src/inferrsi/infra/__init__.py": {
+            "bytes": 85,
+            "sha256": "8779897733779266841b0f79b84b1dc43f0fa57c013855c906c809829ecfc77c"
+          },
+          "src/inferrsi/infra/artifacts.py": {
+            "bytes": 1040,
+            "sha256": "e72e1bebdbb88e42de430238489252386851f9e6645ed7780f2aad2e90338120"
+          },
+          "src/inferrsi/infra/campaign.py": {
+            "bytes": 1028,
+            "sha256": "be505fba1592611bd84ceaca55afbac027205413a2639baefa0c3bb343230209"
+          },
+          "src/inferrsi/infra/commands.py": {
+            "bytes": 811,
+            "sha256": "1890f115b3a2c5c6832173bf60432ad595bd6bf1c69783ecd460c6f8961194f1"
+          },
+          "src/inferrsi/infra/environment.py": {
+            "bytes": 2067,
+            "sha256": "60e46c2aefc216db202b41f44542e7b893e08e0721c667b9ccc50411d1ac84ee"
+          },
+          "src/inferrsi/infra/executors/__init__.py": {
+            "bytes": 64,
+            "sha256": "327c3cd5c821c832ff260a41dba0ee1e439cdabefc9edfb9cad1cd38d14bc87a"
+          },
+          "src/inferrsi/infra/executors/local.py": {
+            "bytes": 3654,
+            "sha256": "37bcbca9f3261c8c58a8bf806c9da40720b18d0676c82e8ed24c3901ac9bbe77"
+          },
+          "src/inferrsi/infra/http.py": {
+            "bytes": 634,
+            "sha256": "f65d6c2b55b6a89694caaf6dcdf7d74f417e9ced103eca1b3f218e305c6b6cb2"
+          },
+          "src/inferrsi/infra/inspection.py": {
+            "bytes": 2415,
+            "sha256": "d1c66deb0cced529bca996d2f88b60f51fa3417ff1cae365c650fec808946ecd"
+          },
+          "src/inferrsi/infra/platforms/__init__.py": {
+            "bytes": 1494,
+            "sha256": "9e88e8b9a396ed90c221e1585135c0a1b58da41c1bb73ed9237c53b6684abfed"
+          },
+          "src/inferrsi/infra/platforms/cuda.py": {
+            "bytes": 3309,
+            "sha256": "9e0e4276e1b2984605493b14b82fba3e86f6635204719c318302e624f66f8b6f"
+          },
+          "src/inferrsi/infra/platforms/mlx.py": {
+            "bytes": 1971,
+            "sha256": "f3b81470db28bd6183ccca2e763c5e079686bba6cdfe6c08ebe2cdf64216f2e3"
+          },
+          "src/inferrsi/infra/processes.py": {
+            "bytes": 8151,
+            "sha256": "3b96ebc7a50913e78e86ff3dcba74c8ee8d03d547d9591ca6488f390ac1e6189"
+          },
+          "src/inferrsi/infra/targets/__init__.py": {
+            "bytes": 186,
+            "sha256": "1f3ea75dded7c6cc967754c7b6d4f30ef00da8191a749938aabc3c6bb8f86eb3"
+          },
+          "src/inferrsi/infra/targets/git.py": {
+            "bytes": 2910,
+            "sha256": "6327c4f126086f310b6de309fc7659c9759da28ed76efaeeb94a3d2fba29a0c2"
+          },
+          "src/inferrsi/inspection.py": {
+            "bytes": 692,
+            "sha256": "afa44bf4cdfbb58a318c21f9b292ca142d1bbb22750e35a4cc8244da5350ebb3"
+          },
+          "src/inferrsi/knowledge_operations.py": {
+            "bytes": 27444,
+            "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b"
+          },
+          "src/inferrsi/knowledge_store.py": {
+            "bytes": 13700,
+            "sha256": "5e794e28906ce4687aff7f14a507e267184640fc8a4fac19e0ced279ce6a6cb6"
+          },
+          "src/inferrsi/model/__init__.py": {
+            "bytes": 77,
+            "sha256": "3c1dba4f5159fd9fbc8d4d96e2a83c8441491c1b1dc7b42031a3fc1782ed9d9f"
+          },
+          "src/inferrsi/model/anthropic_agent.py": {
+            "bytes": 24696,
+            "sha256": "b0ef9cd542544292eeb90ab05ef410be9ed6e5d1a40d729890f49562c28c12f0"
+          },
+          "src/inferrsi/model/protocols/__init__.py": {
+            "bytes": 85,
+            "sha256": "5a2f54483b164de04e309dbebb87a700b213682b0f44a50d89a752dd71a4ce47"
+          },
+          "src/inferrsi/model/protocols/image.py": {
+            "bytes": 2047,
+            "sha256": "251a1f8f885da2cd8e9d3baba766c525afe293091606b4e8668c43ba2cb7cb5f"
+          },
+          "src/inferrsi/model/protocols/metrics.py": {
+            "bytes": 610,
+            "sha256": "17db16d6ef15b41fb60d8f09507893accc43f9cb3ad686b8dc4e9839713ebc56"
+          },
+          "src/inferrsi/model/validators/__init__.py": {
+            "bytes": 65,
+            "sha256": "1d4b6ffa968e283219ddcc8a09de9280238a8d65ae2dc8cd2b60889877a8eb0b"
+          },
+          "src/inferrsi/model/validators/flare.py": {
+            "bytes": 19975,
+            "sha256": "3696b0609c772f2222b448833608565a033123d7e92b11fbc545dd356fec590f"
+          },
+          "src/inferrsi/model/validators/image.py": {
+            "bytes": 3204,
+            "sha256": "79470b6b06a22d53ec6c40636a0b3d821a2e6ee2f7bb7aed3dac7241504add1a"
+          },
+          "src/inferrsi/reporting.py": {
+            "bytes": 127,
+            "sha256": "1f771890ab3e69b1aa9246b548a7825596bdf904f6c395dd218208084266853d"
+          },
+          "src/inferrsi/research.py": {
+            "bytes": 5213,
+            "sha256": "9bfe59dd17b5476c18ce50d671711053c62800521c50d6b710e8140a2b4dca04"
+          },
+          "src/inferrsi/runspec.py": {
+            "bytes": 4398,
+            "sha256": "dca0952425caf42095ef0c7e2feb8ef4740cc422ecb61e777f203962d9b87709"
+          },
+          "src/inferrsi/skill_catalog.py": {
+            "bytes": 9938,
+            "sha256": "65f5a1f3f6894009abb98ec2fec30a63cc1bb2f6ce6aeb01f9f7a3dce0439f2d"
+          },
+          "src/inferrsi/summary_chart.py": {
+            "bytes": 40910,
+            "sha256": "d29023d60f6be804c1b01c9eb73765d4824d6c3bb870fe22c99a24544bc1d8bc"
+          }
+        },
+        "sha256": "4faa529dc47f23c24f8c5b8e0110d8e413aab20df5ad61db591fb751564a6448"
+      },
+      "tracked_source_dirty": false
+    },
+    "B": {
+      "corpus": {
+        "bytes": 2883003,
+        "file_count": 458,
+        "files": {
+          "src/inferrsi/data/knowledge_packs/README.md": {
+            "bytes": 14172,
+            "sha256": "b877e67d9cccf0dcbe57419eeb3ff9374e5bd03ff8eec63adef1025c446140e9"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/diffusers/cuda/diffusion/group-offload-transfer-overlap.json": {
+            "bytes": 3277,
+            "sha256": "9bec3d912192efa2f2c853d3fb70bdae21a75875630a02294fcd084b18c30414"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/flashattention/cuda/attention/generation-specific-dispatch.json": {
+            "bytes": 3085,
+            "sha256": "90290d7bc80b1626863e0520c91bb196d93ebfe08190a450e9de59c6505ae8ad"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/flashinfer/cuda/llm/paged-decode-dispatch.json": {
+            "bytes": 3745,
+            "sha256": "0f2986494fcc9b751df71e598fd61caa68e28bc4b0ac401e4cc763bc3f75283b"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/cross-stream-dependency.json": {
+            "bytes": 3809,
+            "sha256": "b829315e55adf8e11f0bd581d5f45dd130d7eb3f81ce7de532f63c7b9c27cbb5"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/evaluation-completion.json": {
+            "bytes": 3841,
+            "sha256": "19d2e506ac4f50895776a83a062059aa7a052175c8e4b21dd0d46e1a5a6716a1"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/post-detachment-buffer-lifetime.json": {
+            "bytes": 4663,
+            "sha256": "18d78c0e542dacb42ca33e5264d92aef80fa06d163fd16d4feafd5ed0827ab99"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/pytorch/cuda/compiler/cudagraph-shapes-and-lifetimes.json": {
+            "bytes": 3623,
+            "sha256": "7fef4122ff794967e61f55d02a4a6b558d7e498d9513607f5ca86db6386928a8"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/transformers/common/llm/kv-cache-strategy-contracts.json": {
+            "bytes": 3072,
+            "sha256": "a2df26052112a9f3fc6242b2bd636ab13991442db57e56646168d6c52032d38d"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/triton/common/kernel/bandwidth-fusion-resource-limits.json": {
+            "bytes": 2941,
+            "sha256": "6e9673d1a600a6ad34820e4325f2a6417f44f0f96d8ed0bcf21466d998f19ee4"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/cache_dit/common/diffusion/residual-cache-calibration.json": {
+            "bytes": 3841,
+            "sha256": "bb9509ad0225a91b52bc6787477581d05f0ecd5e44fa59f08f718eea36a2ac45"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/bounded-windowed-replay.json": {
+            "bytes": 10959,
+            "sha256": "43613225e9ccdbd34c0f7e3e0767887b9bbfd0f38b2c21684677d99e31a432f0"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/speculative-verification-budget.json": {
+            "bytes": 10177,
+            "sha256": "a729e1fc00be44e1480a5dc65bbcdd58df8599681df74dbc651a5a49f27e9028"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/zero-overhead-deferred-scheduling.json": {
+            "bytes": 8191,
+            "sha256": "ed5426af145f066f6183bd316c4073c46f324f5ee83aa9bd616387c0d0e7149c"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/lmdeploy/cuda/llm/kv-quantization-capacity-cost.json": {
+            "bytes": 3366,
+            "sha256": "873997b38981ccc03ff73e6fb84e96022f293a019843efd8c4aa50f27878ac9a"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/codec-constants-and-state.json": {
+            "bytes": 4250,
+            "sha256": "1cecd7905fef5af3ef8952a4d803900a7df91d662ebae34db47cc0dbf352d4e7"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/fixed-customvoice-prompt.json": {
+            "bytes": 3660,
+            "sha256": "ea5e8fdf54cb731b255b73aa0cf47ef53fac103baa2bd5246be45f95d54cb593"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/graph-batching-state-ownership.json": {
+            "bytes": 3950,
+            "sha256": "130741a9f7ac9d366784d6c4dd42860262235caff0c277de1e8a8f8ec933b37f"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/sglang/cuda/llm/hierarchical-prefix-cache.json": {
+            "bytes": 3641,
+            "sha256": "b50ca58df28d4343545d07adf0da23f8f181297b30aba7870e5690630e7d7315"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/tensorrt_llm/cuda/llm/prefill-decode-disaggregation.json": {
+            "bytes": 2967,
+            "sha256": "349856dd98628f94987e014889cd1b15d0e3ae6ab6c576cdde917f28fbafb102"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/xdit/cuda/diffusion/parallelism-freshness-and-topology.json": {
+            "bytes": 3236,
+            "sha256": "3674b5916cb7b384951bf3e449b6796b161878daabcc1d7e7f6d6f8ff861cc17"
+          },
+          "src/inferrsi/data/repo_packs/afd_plugin/pack.json": {
+            "bytes": 502,
+            "sha256": "00042abeabdc9c0e24f1a4edaba47e233df398aef0ad4e181fc70191d383da5f"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-aclgraph-capture-replay.json": {
+            "bytes": 4311,
+            "sha256": "327414443ea9537ae320aa405912c338351bc06cefdc0b99352e1eea36f61dc3"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-device-health.json": {
+            "bytes": 4115,
+            "sha256": "633fa5fdcea1e1e9edda8a218d6a3fde0d12637848348e1e2fe7d1b0bd744a16"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-distributed-parallel.json": {
+            "bytes": 5841,
+            "sha256": "709ca0d061e4d6917fe8fa049fe93165e71dfecf96c7bf264ced25346381f603"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-graph-mode.json": {
+            "bytes": 4135,
+            "sha256": "bb82ab3d628b91be957e557545810ef5c20790be7d43296096f20a21f45ec391"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-model-adaptation.json": {
+            "bytes": 4137,
+            "sha256": "7ea1ab0f0044afff935a91071b36ac09bc46b5ef044fe2a68ab0608256ddd853"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-profiling.json": {
+            "bytes": 4186,
+            "sha256": "0d27058fd8d0ef8b524b7a60d1cb76950622746c94b5a03915e64e6edb4ab3d4"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-tuning.json": {
+            "bytes": 4439,
+            "sha256": "fcb49d71f59d8b39c1eaa03d24e95e2cbfe5f9d7f6e0f22df03ee0d60f6a568a"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-optimization-casebook.json": {
+            "bytes": 4371,
+            "sha256": "05783630bfaf3c7ef99b848865f4bbab50d0b52e80ad04fcd028c624d7638e9b"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-performance-analysis.json": {
+            "bytes": 7200,
+            "sha256": "b6eea897c12d0c6ac9d172089535c82c25eb71c09d7a4972e64812cc4682b5c9"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascendc-tiling-design.json": {
+            "bytes": 4505,
+            "sha256": "e69b6e89dd342ba55ae54b5fac14aa1b91ee2a361da40cb669c287ce519f4517"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/cann-attribution-overlap.json": {
+            "bytes": 4915,
+            "sha256": "8ac01b60f5cd37fd594d6cb5119e4e6525199d017296fc8779dc68a576f5525c"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/torchnpu-device-memory-lifecycle.json": {
+            "bytes": 4349,
+            "sha256": "7a79d0eb2c50381d40086e6db72d4bdd2cec5d2261bfb00cf73a60701a94da2a"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/torchnpu-pinned-host-memory-lifecycle.json": {
+            "bytes": 4431,
+            "sha256": "8bb34dcfbebb48667a94434b036a1a408b970af9890b047a3de325f364ed5f7c"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/assess-serving-preemption.json": {
+            "bytes": 4198,
+            "sha256": "e3d2a1a72d0d980f31fc85deee27db26bc0daa0c709e8ae99e42deb814a4a7f4"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/assess-split-backward-schedule.json": {
+            "bytes": 3915,
+            "sha256": "a8561794786eb0a897d831f00c92b4fdd3b0492316c1b1d8ad13c1f8f75ac815"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-h3-runtime-residency.json": {
+            "bytes": 12169,
+            "sha256": "602dd625874bd4a89861d5bb25881cc63e4b5a7fdedc2867089f0952943f6eff"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-llama-runtime-state.json": {
+            "bytes": 5816,
+            "sha256": "64f6ab8a95172378350323cead08bc6c201c624b711e6242938a6a21a8e83ad7"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-sd-cpp-residency.json": {
+            "bytes": 8349,
+            "sha256": "95ed97b717255be3b817e5c69debd6d41197559dbf21771507d5da79134a72e1"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-sglang-diffusion-lifetimes.json": {
+            "bytes": 7276,
+            "sha256": "5e5889223878c6cf42a49c6f09c60e9d9add88cb7e7d0df050c4e75d88a05aa6"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-vllm-kv-lifetimes.json": {
+            "bytes": 4275,
+            "sha256": "00424005a5eeb8b15f239edd2cb2c71047787e3ecf55924860cb7a78d137b586"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/bound-windowed-attention-replay.json": {
+            "bytes": 7877,
+            "sha256": "98b2291147a796ef59cf20a075ff662998ee6a529f85b49e9c6fb62fe41fad9f"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/budget-speculative-verification.json": {
+            "bytes": 7490,
+            "sha256": "66448f59460c9dffac46e74160388fab650cf8bbbd5bdc95c1465b70daf0c2dc"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/calibrate-serving-simulation.json": {
+            "bytes": 3992,
+            "sha256": "055b70e0b86f65156775b519e4b8cbf06a956a8993b06a66175167cedb78f2cb"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/cgroup-resource-controls.json": {
+            "bytes": 4801,
+            "sha256": "ad406b6dd5d9415bd28d448e64269a388b50e11f9eb7f5a9adb6b0a6246fd303"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/checkpoint-recovery-consistency.json": {
+            "bytes": 4740,
+            "sha256": "ccb85314a831bde58f8cc515580d001b174c11dca0bbcbc6221303acf629046d"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/choose-parallel-training-plan.json": {
+            "bytes": 3901,
+            "sha256": "40afcb41c59988fc1af8fdd00d2340800c54e96847dd0de16ecddf7ac221d1b4"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/cite-check-downstream-pr.json": {
+            "bytes": 2533,
+            "sha256": "dabff14684d5ae9a2b05a9743ff5b76d00a90fdb29319588aba67f5087aad706"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/claim-contract.json": {
+            "bytes": 2469,
+            "sha256": "78f15f6533434fc29adf1df651f7a15eb8ba81733f8b02c710d90102541d707b"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/comfy-ui.json": {
+            "bytes": 6919,
+            "sha256": "1496807d958df055f720ca93e6dcf7f7e8f506f3f12c94c9a2166f789b00058d"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/competitor-scan.json": {
+            "bytes": 2449,
+            "sha256": "c2e97e49e5185710cdfaa08ac1c2b1c19b958e68ebfca885d5aee0ed8c9f8c8b"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dcp-async-save-performance.json": {
+            "bytes": 4538,
+            "sha256": "359a5acf8304a78fc074688676f9a0be62f56f463d0960a544e409f49843bdaf"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dense-performance-evidence.json": {
+            "bytes": 4227,
+            "sha256": "d87f93a9dc9ddc583e3929baaac85f7a425666dd6dd71665933f18344ccec834"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/diagnose-vllm-admission-pressure.json": {
+            "bytes": 9789,
+            "sha256": "a1de40def775b2e840c27330b67e6f37c3c244db1731fab2bd3e8a787143210e"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/diffusers-invocation-state.json": {
+            "bytes": 6109,
+            "sha256": "914d2ffe9776ba931bed780a2a270c1525e3c6cbaf0735b8531064d3180f3233"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/draft-acceptance-issue.json": {
+            "bytes": 2549,
+            "sha256": "6aaa63dbfe6cda982ccd3a4e368527b80fd2873672c000c66d06d28e40a2bf14"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/evaluate-elastic-serving-placement.json": {
+            "bytes": 4334,
+            "sha256": "fab1fa814e789ec2d86587fcf7de7c38174464b9e9ba0cf53ef3fb470053e7c4"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/generation-cache-lifecycle.json": {
+            "bytes": 5972,
+            "sha256": "e5679cdecb722bbd17280913913da8473052ad09e9e7a13b3a9b95463a139ef6"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/host-memory-management.json": {
+            "bytes": 6530,
+            "sha256": "a36aea21169281866b2468860d3ffeb912e61edd571308b02f82939f17f694a8"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/host-overhead-optimization.json": {
+            "bytes": 4243,
+            "sha256": "05fd7f054966b96892980d585352bc1c3fcd47f439cb34268fdf07feee58a998"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/inference-benchmark-design.json": {
+            "bytes": 10735,
+            "sha256": "091bb086c6f0123dbb5db78968c13f846cad135e8969982efe557644b7c7613b"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/inference-optimization-casebook.json": {
+            "bytes": 6531,
+            "sha256": "c512025d7a59934c91665e24752d9393ca9b6a9dfb624c72789294defae93c0e"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/kernel-writing-guide.json": {
+            "bytes": 4545,
+            "sha256": "f1f0ca1d11d3550c9b8c5fdbed8f6cd09b877ac14341f973d58dca02b5d8995d"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/kv-tier-management.json": {
+            "bytes": 4609,
+            "sha256": "6141f61fb6b0d147b704ab7a2627e9babdb53a90e15b659152a3d565967afddf"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/lazy-evaluation-stream-lifetimes.json": {
+            "bytes": 5934,
+            "sha256": "34f47009a8a33f8907bd78f110b95fcd62409bef1b38e1055acb46b2eab285ff"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/npu-skill-factory.json": {
+            "bytes": 11868,
+            "sha256": "24c1d3bff288e86b99ea674e3252ec6f54d9ba944274aef6a7666197afd6788a"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/optimization-search-policy.json": {
+            "bytes": 7101,
+            "sha256": "d7bf71d5491765f62eec02fe5013dc1a5ef56e1516892da48593c6165a86366d"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/os-process-placement.json": {
+            "bytes": 6304,
+            "sha256": "e6e8fe5d7d66f3e98f05762d94441726dfa757ce7fbffd2b35402b3f327f8fb3"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/performance-ceiling-analysis.json": {
+            "bytes": 13695,
+            "sha256": "47c034d96026e1c2a02848d9c9c3caf4024a5033664bfd089ad30d9f23a80deb"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/plan-activation-residency.json": {
+            "bytes": 3871,
+            "sha256": "436e364020cc854fd52a10ee76e4593c7519bb0546c8aa5238fcfd89158e8ab0"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/repository-evolution-analysis.json": {
+            "bytes": 5865,
+            "sha256": "e0af402c3320ebd9bd36c774cb52c0708721a86ca446c05f6e94306888838e30"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/rsi-evidence.json": {
+            "bytes": 2475,
+            "sha256": "c8b4bb5a8c88009ac5b1526e82639715381c940d3c3eb295f0333cae6c2d3ca9"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/skill-evolution-playbook.json": {
+            "bytes": 4091,
+            "sha256": "0e063c2e7686de0c27c061e1aac0eb2d8c4fa83246e32ddeeb0e28126c176425"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/storage-io-measurement.json": {
+            "bytes": 4783,
+            "sha256": "80e589e3a0b67d3a045e03feeadc45e720ceb277e6124c917eec04863eb67419"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/training-input-pipeline.json": {
+            "bytes": 4296,
+            "sha256": "ae696bea8d2b7bfdaa29b6e5bc0bd861abcc0e42e7d9bec6ae20c6e7d6f27669"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/training-resource-budget.json": {
+            "bytes": 6944,
+            "sha256": "ea5e5e864e3aa34bd00be67c8e741877ae01480f039652073b65b71dfcd54c97"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/weight-loading-optimization.json": {
+            "bytes": 8139,
+            "sha256": "7f2d1fbc6b1cfd80c3d6ed8df107d4ca9b97655626ee078c345445a303b1b367"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/audit-afd-p2p-exchanges.json": {
+            "bytes": 6640,
+            "sha256": "9ac45aa784b430ad1ca9ffddecce8dbae9740cfe9c98c10b330e01d350015a38"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/diagnose-attention-bottlenecks.json": {
+            "bytes": 7602,
+            "sha256": "e525d75578de1cc164e6553d01efbd0a014e74646200580f59b1161c030413ad"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/host-overhead-profiling.json": {
+            "bytes": 3965,
+            "sha256": "e9e6ae8c5b9986e83cade3bd029baae112859e01aa6c96f741a264f3bee44adc"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/reason-hopper-kernels.json": {
+            "bytes": 15086,
+            "sha256": "758a649e651d8f9635ac2df45d18043e156c8d54b8f8dbc95f8e68ebcb866a28"
+          },
+          "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/parallelism-and-extension.json": {
+            "bytes": 3399,
+            "sha256": "1253f5f01bcf1498703c7f34960f0c50580df5ba8f140af557a903e296ad81d3"
+          },
+          "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/runtime-radix-hicache.json": {
+            "bytes": 3317,
+            "sha256": "5a0b2aed51649af214cdc1034f57650b7ce195c934a17452ee67a35131f80cdd"
+          },
+          "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/tuning-and-validation.json": {
+            "bytes": 3385,
+            "sha256": "ef4d4f7182da4433d59d72dda0d21b4aeafadd73260be11ab8d4cb36c20fafc6"
+          },
+          "src/inferrsi/data/repo_packs/sglang/pack.json": {
+            "bytes": 620,
+            "sha256": "bce2e0c42796a9ef499d459c37b56d1d56f4b19344841359c534fca29c0359ce"
+          },
+          "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/audio/serving-and-streaming-quality.json": {
+            "bytes": 3440,
+            "sha256": "97fc1d5a11a39a0634dae9e9602a4ec47a4d2348e121076c2d5c836cf0692114"
+          },
+          "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/omni/extension-lifecycle-validation.json": {
+            "bytes": 3482,
+            "sha256": "7da8dc5d2655099de2bcd5f0fa922d9e1bcb0bfd2f571f026a004a7f402a75b6"
+          },
+          "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/omni/pipeline-state-and-transport.json": {
+            "bytes": 3404,
+            "sha256": "0c57692772019504264e39bf0444a69e20cff86f6e112b072c53cf7a4299cc82"
+          },
+          "src/inferrsi/data/repo_packs/sglang_omni/pack.json": {
+            "bytes": 630,
+            "sha256": "8e70336bc2aac0c74732a026337813652b5d5c6e1fc60f5571e9442345077b96"
+          },
+          "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/architecture-search-space.json": {
+            "bytes": 3294,
+            "sha256": "2c5ad72d5bab4f7082fc3be1532c347444cde38f5a84639fe16b22134b68685e"
+          },
+          "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/controlled-execution-tuning.json": {
+            "bytes": 3255,
+            "sha256": "6a2bf611a432213919e1c237687507ceb0ab753cccad0e934dc77f86604805e9"
+          },
+          "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/engine-validation-contract.json": {
+            "bytes": 3295,
+            "sha256": "ef89775bc82965fe87a832c7ec8a5cf55cb9eba4ecfc6279caea9ce3b0410c41"
+          },
+          "src/inferrsi/data/repo_packs/strata/pack.json": {
+            "bytes": 617,
+            "sha256": "643b32defb8db9aa230f624856f4fb4b595b5c23db58d9453091f5420cff71c3"
+          },
+          "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/bounded-serving-measurement.json": {
+            "bytes": 3239,
+            "sha256": "fb00ddac14ef6b06f48da9ba335529bd55d5d83c2a2fe331f0aa1ca16afceb37"
+          },
+          "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/extension-correctness-gates.json": {
+            "bytes": 3343,
+            "sha256": "4ea7083d277df205ea4c361bee87a8207951596e9ef6cfa136cc7d2aeab02f46"
+          },
+          "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/runtime-and-kv-ownership.json": {
+            "bytes": 3280,
+            "sha256": "e6142c1065754716a5f26f42119722908dc8fff63e667c986d57f1227b8610bd"
+          },
+          "src/inferrsi/data/repo_packs/vllm/pack.json": {
+            "bytes": 508,
+            "sha256": "1705d96710249c4514c15d8d011b43c1a1ae7c15fbc9ddf11ef2d0e6b2ab7d07"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/minimax-h3-reference-kv-reuse.json": {
+            "bytes": 4047,
+            "sha256": "80b853fc594d387ab48bed04dca245e0a22b8cdaf7c9b853551ab3bafa302e84"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/minimax-h3-sp-head-buckets.json": {
+            "bytes": 3742,
+            "sha256": "c4d29d1bceb5e2cf86151bf53c9dafb3d83428fb1e9d2a848d852bb8824bf857"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/qwen-image-21-eager-t2i.json": {
+            "bytes": 5268,
+            "sha256": "ef107d43cc448219eda4d64ab961b61cda6350e20d30d94de1ff7c4ffff7e775"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/audio/streaming-serving-measurement.json": {
+            "bytes": 3389,
+            "sha256": "2680f32f28907f6abc7a7e1e3ecab4f46ee412f40f4898ed2f371ffadf47ae4d"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/measurement-and-profiling.json": {
+            "bytes": 4472,
+            "sha256": "c8b25583cd160024c7472a20debea34cb6d4bc79714b4bad2953555adc9f794e"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/production-integration-gates.json": {
+            "bytes": 6112,
+            "sha256": "649b49ff64d3e9e88bd620027dabf9d126ca69bbe91b448f28e845ecbe1f8e2f"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/step-state-and-output-validation.json": {
+            "bytes": 5090,
+            "sha256": "e1ffe28c4fa54af739cdef2fedf4755bba18e078f8380d4fde8138ea712e5778"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/upstream-review-gates.json": {
+            "bytes": 7614,
+            "sha256": "5127709bb85ce31008d19468e1b4fcb5454a71776f69f991841cd9ef7558f9c4"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/omni/stage-ownership-and-search.json": {
+            "bytes": 3427,
+            "sha256": "e75d6c2829ae22881893f5da2ad7edba099fd64ccfb0fc93d30473e06e9d4d80"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/audio/concurrency-scaling-state-and-overlap.json": {
+            "bytes": 25324,
+            "sha256": "cb47fabb232db81ed525cbb54371becc7e4a28b2bae9d8492f60dfb6f31563e8"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-adaln-schedule-cache.json": {
+            "bytes": 4253,
+            "sha256": "ddbba3bb72b5789938595a82d3280e861f75f6ed54f05a1d2b7dcb1089d30f63"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-cuda-mxfp8.json": {
+            "bytes": 4468,
+            "sha256": "e4c538f03f4ab90a4acc8e3d8721bbd226aec1ee9bde45757a7c27d3d970821d"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-fp32-modulation.json": {
+            "bytes": 3625,
+            "sha256": "abb8d94a1b025f0e2c8e085275e96ee85d68aeb27cb12d0918e4647f8255b693"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-qk-norm-rope.json": {
+            "bytes": 3706,
+            "sha256": "4c1c4acfb9cc0baa90bc28fed1903020ea1ea38755b84315711795db45e48bb6"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-ref2va-teacache.json": {
+            "bytes": 3835,
+            "sha256": "8aeb3f99ac4227cf8b3a5cece13d938b19082e265a557eddae27b7e495bcc22e"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-vae-stacked-tiling.json": {
+            "bytes": 3508,
+            "sha256": "8cc18eacbbfd8fc4d3a6b4152c4b584d559fc9b15377da56ef072e542d807c93"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/pack.json": {
+            "bytes": 566,
+            "sha256": "58c971e9dcf6ac15752629ed4d9426aeb5b5d857fc37b5bcf921aa2e81c0cdd5"
+          },
+          "src/inferrsi/data/repo_packs/vllm_rlt/pack.json": {
+            "bytes": 649,
+            "sha256": "6097041b8d026740060e527f49b1b4e288315b1c3227f7583805b86d8fa415f7"
+          },
+          "src/inferrsi/data/skill_library/README.md": {
+            "bytes": 18819,
+            "sha256": "ae583f3578425df1dc91ba32b8b6543c0469c9a817ebea873a77051d6b13e1ce"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/SKILL.md": {
+            "bytes": 5494,
+            "sha256": "7581b56336d07213a87d01477793ae1a8acf0b7cf876fce36c9a761ee32c1ff4"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/metadata.json": {
+            "bytes": 1429,
+            "sha256": "68ac7f64b7ab947fc2db664bc3990f102c1cc5539b42350aa5818d0bbf17d86c"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/ownership.md": {
+            "bytes": 2081,
+            "sha256": "40688cc44183027efbf2fca8d3548c299cdf2e33fef54f76ab3549353eb9bd1b"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/source-map.md": {
+            "bytes": 4635,
+            "sha256": "10562a75683bd9bd07ec2e37a1cbd0f93f2534d6f85b55a566adb2d97485215f"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/validation.md": {
+            "bytes": 1477,
+            "sha256": "5a0b854b87d091d7d505b0bdd5c03330052507a851ab6edf35b93a6675f2d4f7"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/SKILL.md": {
+            "bytes": 4010,
+            "sha256": "a95569bcc0da6e715fd74783a6a1e0489e92b680c582f15dafbf98df589d505a"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/metadata.json": {
+            "bytes": 1519,
+            "sha256": "9c2293c8956508de987c886c7224d673d22912e7e915b026b05abdd287e97401"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/execution-lifetime.md": {
+            "bytes": 2147,
+            "sha256": "c4bc41e08aca68f661325f61e6f0db1760a079dc362f295c847bdc374fee3bf6"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/source-map.md": {
+            "bytes": 5918,
+            "sha256": "798b9cb89ed6b885eb47464ca9f137d2802a54571b9327e631d7742d3e5abd5d"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/validation.md": {
+            "bytes": 1608,
+            "sha256": "ebfc00bf7d6ab0a1dc415953ae84d56ccdc713d5614aff85e433c4a843793bff"
+          },
+          "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md": {
+            "bytes": 7322,
+            "sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845"
+          },
+          "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json": {
+            "bytes": 587,
+            "sha256": "db7ad368394ccbfdeff6a16e3423c300d51305b1f95ff5c3936df91738a8cc35"
+          },
+          "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md": {
+            "bytes": 5314,
+            "sha256": "22afe4f4d77f441096cc657d55d763473bfcbea8fb531c8e493fe49ab84955f7"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/SKILL.md": {
+            "bytes": 6745,
+            "sha256": "d494fc7e539cbb7388189bd7b3f04bb5c4983fdd7a82484348aad1cee7f741f2"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/metadata.json": {
+            "bytes": 558,
+            "sha256": "71c68431663ed2336cccc985a354adab6b13b4cfd6123a18a5d8d53c28268509"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/references/source-notes.md": {
+            "bytes": 13004,
+            "sha256": "6ffbfb68cc7e04923b4fbca2a73a7555b7c3b45a6f40819bb5f3213ef05cf88d"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/SKILL.md": {
+            "bytes": 5977,
+            "sha256": "03829039860ece35519f2e4c339a48b5014ec25e07a2c150a3efb6a3a4e35b4a"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/metadata.json": {
+            "bytes": 559,
+            "sha256": "80d80c15514919826fa72d6f6cf26b76db8accd85d21187fc239232ca454e7bf"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/SKILL.md": {
+            "bytes": 14703,
+            "sha256": "4ce9dcadd1ae3c7a8cc633336760d0a3c382f354f1b7028e3a10d28fa0d0c9fa"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/metadata.json": {
+            "bytes": 1403,
+            "sha256": "b6cb25879b1bca5f5af08b48776f2e664cf5dcd11a0efb1c4f56879e2afa9006"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/references/awesome-ascend-communication-tests.md": {
+            "bytes": 9388,
+            "sha256": "0cb27a3d24409171bb414d7ebf0c49ad62be5f46482f29624695496ba9eb3e63"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/SKILL.md": {
+            "bytes": 5839,
+            "sha256": "9203cc5a14a1d65df5fbf275734b93be023e06527459b2f79864f28a338cb44a"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/metadata.json": {
+            "bytes": 536,
+            "sha256": "72b41e7845915d34b09219624e82976386f3125816293540c2c38143fe6b477f"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/SKILL.md": {
+            "bytes": 6440,
+            "sha256": "8fdf46fb2743b67a66bc069d22a45305a222debfb606777bc451b4b8f31d9cc6"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/metadata.json": {
+            "bytes": 536,
+            "sha256": "be681a3206e5e183b4c9670c9d7784bcffce18c891449e5b15f94ca5f1324b9a"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/SKILL.md": {
+            "bytes": 12988,
+            "sha256": "401981cf7828edcab0c9ea882f7ecd83badfa3c2a95006e912cf99d67b3d6a12"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/metadata.json": {
+            "bytes": 760,
+            "sha256": "efd37bb23acb500dad4ec8ebe7ddfe5c3f196374052825f923ac85da8445f06b"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/SKILL.md": {
+            "bytes": 12236,
+            "sha256": "38b151bc6b4e9734c6fa41efb7a30343559400c0c75bbdd26bd8a968b8ea4592"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/metadata.json": {
+            "bytes": 557,
+            "sha256": "04d2002fc8a0a7becf86b0b73daaf2da8716ab466d7f988da43d41a10f26680e"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/SKILL.md": {
+            "bytes": 9172,
+            "sha256": "647f5e9a288f2c4e1617ba53df59d9d827df749eb4fe86182d21a9c41175caf8"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/metadata.json": {
+            "bytes": 536,
+            "sha256": "f425f7c1abbd616387273e8932c6bd02f4b1da012f099286ad4171e0f6037b22"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/SKILL.md": {
+            "bytes": 9880,
+            "sha256": "a057b7f86a24f34a8f84abbbb4910d58d4e50ff47060e1c404600017e3febe89"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/metadata.json": {
+            "bytes": 2297,
+            "sha256": "9e705625c8308013c287576b0247da6f6116d817a4a535cbecff80abd6ea35ee"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-CANN-2.0-LICENSE.md": {
+            "bytes": 5262,
+            "sha256": "89b1c114df62cfd183004f38d18d07f186436c3b37203dd69c3e84becf9d2282"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-stage-evidence.md": {
+            "bytes": 11789,
+            "sha256": "f56e4125325c9c1b9235a1d64fd732861c7686444492422daff44ba8dd648be4"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md": {
+            "bytes": 11860,
+            "sha256": "93aa1b775ced304cebfe6c4dec0503ddb2260ecb3cb01edc17ce329446c5ce2b"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/metadata.json": {
+            "bytes": 552,
+            "sha256": "f6d94f104cb70969024b2f4542078a240e200712c66279eca8b94f60bee1dd71"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md": {
+            "bytes": 8972,
+            "sha256": "0e6f094dc554abea12e07def50f20b233b47127779a8aefad595b4ab1efd24c2"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/metadata.json": {
+            "bytes": 569,
+            "sha256": "acccc8eb86de0df6caf78a9b3c6db76a2a686adc4a5ba0029a4d62d50ff9a8f8"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md": {
+            "bytes": 5989,
+            "sha256": "8428df7157af024fa17cfe0b3818ed54bd9f468306dda46eb8d4f0edf4f3f230"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/metadata.json": {
+            "bytes": 579,
+            "sha256": "c650dce88c07b636d68e063dbda4d319f6f7f3ecbcdd5e47f393cbfc3af7c71d"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/references/source-notes.md": {
+            "bytes": 11470,
+            "sha256": "80cba4d3a884e7ee2618326cf74b9f9c3a34e4d9cf443707f6bbe673f2d5ba65"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/SKILL.md": {
+            "bytes": 6159,
+            "sha256": "23e93ea7c98f99743509fdc04139d78ed2e7123c63e80b366c6beaba7eb70d0d"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/metadata.json": {
+            "bytes": 581,
+            "sha256": "97836a38465106d816a9ace894a7b277d6d330f9accade8f8493ce6f0fb57ef9"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/references/source-notes.md": {
+            "bytes": 13493,
+            "sha256": "435193590b4a2ba32e2e4c6fa9d7e8f889528d412cbbb48e35ec5433f7dab384"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/SKILL.md": {
+            "bytes": 7137,
+            "sha256": "a25738dbb3b9daa15c5827cb536a20fe13f7c8b9dd14aa3fb7cfa7e132879431"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/metadata.json": {
+            "bytes": 1007,
+            "sha256": "3abe626a6e4c73c943429f609ce2c62aa77aae7e1726475ebbbdfbe1794f6f7d"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/mechanisms.md": {
+            "bytes": 17376,
+            "sha256": "fe08b2292fa587be3874774ddf41063b84c579a0085df422ec6710d8aa90b8b7"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 4024,
+            "sha256": "371b01ee201ef51fecbfd6da892d517aaba295930b9395025f03f34e21c030a1"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/NEXT_GATE_LEADS.md": {
+            "bytes": 1623,
+            "sha256": "35b8b58ae5db02b2df51ccbe5cdf870a7383d17f345f9ee5eb7b0a4146dfc74b"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/README.md": {
+            "bytes": 2888,
+            "sha256": "e0b1ec1b39c10067f5ddce58320851ad089568c4b2296a24777389097cb5b388"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/io-reduction-can-plateau.md": {
+            "bytes": 2671,
+            "sha256": "5cc35f514224ac6e6efd2da6db221bba468b2f4db66807295e67b240bcdfdba5"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/overlap-and-fp8-counter-signals.md": {
+            "bytes": 3692,
+            "sha256": "80c8a60a6df55f3d84fab23e03567d8b38e23bcdbd63cc09881e46db1e2888ab"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/partitioning-after-fusion.md": {
+            "bytes": 3132,
+            "sha256": "4d4cabe0991eab554dcfbd4fa9c8199e61fa4767709b0c4cd646e2588bb398a2"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/knowledge/attention-cost-model-boundaries.md": {
+            "bytes": 5223,
+            "sha256": "5ed8f888443e38ccb72545b6025447aab7b92936f78c151fed680e582126239b"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/sources.md": {
+            "bytes": 9069,
+            "sha256": "53436e0de4e6dda2d29426591bf61f8526378cdc6f9ec7967b88b1bcd4c61e6f"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/host-overhead-profiling/SKILL.md": {
+            "bytes": 11199,
+            "sha256": "326e566fbe30d3ba04e4b50c3e87aad9df037f0ac4ad6fe129165ae97dc5f91d"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/host-overhead-profiling/metadata.json": {
+            "bytes": 527,
+            "sha256": "76e34c1c64aba473a9e126b032a776417caa0f913f6ac3cf4dba3a5d123b1ee0"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/SKILL.md": {
+            "bytes": 4905,
+            "sha256": "cec4e3aa1eb64c31c54b5f5ca40af2b61b7919d94be4174968581d5f752c0538"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/metadata.json": {
+            "bytes": 7453,
+            "sha256": "7520da1cc97412631a600957ecb18aa5a7a4461ea21e76c237330180193ed898"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/fp8.md": {
+            "bytes": 6287,
+            "sha256": "31d9c211c11c2224007abbdcd4059913c4a54833fe2d2f926411c5238fcd3501"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/cutlass-bsd.md": {
+            "bytes": 1547,
+            "sha256": "42fec630f410aa308f70a51a89fadcd19586fa620f9831a32bee528a9a10000e"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/deepgemm-mit.md": {
+            "bytes": 1065,
+            "sha256": "6c73a30c14a2938abaa6afe98ebb19e6d81dac5bd4ace0cb05c61c20afa37b85"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/kernelwiki-kda-mit.md": {
+            "bytes": 1063,
+            "sha256": "84aab57158a914f448b38d8be29dd842b634a08c192d844f416a256beed4ae94"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/notices.md": {
+            "bytes": 3240,
+            "sha256": "876ec016644bf5331b3f257985e126c0e61980e3bfd82569b7b0c3f6b2c6c64c"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/resources-and-scheduling.md": {
+            "bytes": 8919,
+            "sha256": "1ce25b426736072435a1ead9a2a8e1f56df5d0abb8198e8c4fdcb34321db4964"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/source-inventory.json": {
+            "bytes": 120506,
+            "sha256": "1dc09c895c970a0c9708158bcc6dd65cdb736274fd2c35d44542c70407d3849d"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/sources.md": {
+            "bytes": 9004,
+            "sha256": "6b74b9e774f2a80db0860510d9e6940d36284430f29c1e124f115a9eaf25833d"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/synchronization.md": {
+            "bytes": 6843,
+            "sha256": "4322f6b3c0e4368bf8c192b4ec752529767f50d19c0c38dad8fdf9cd781e41d9"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/SKILL.md": {
+            "bytes": 7348,
+            "sha256": "fd1f5a7e6a55714730044155f5b38c1d700383d4e16c3e91963bbfbe1c773af5"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/metadata.json": {
+            "bytes": 572,
+            "sha256": "19e40763fedc20990651bb989c32a7693306c5c54af46c3d4be77fd4f87b6c03"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/SKILL.md": {
+            "bytes": 23213,
+            "sha256": "2c807f3df2da2f533d25183de5771f3d412058d16410994292ef8f48c49f9b3a"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/metadata.json": {
+            "bytes": 1684,
+            "sha256": "9ac6af135f0b6d76a76bc201b6bbd8ac9afe5201dd1e2ea82c0ac4826de401da"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-LICENSE.md": {
+            "bytes": 11357,
+            "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-native-offload-readset.json": {
+            "bytes": 6267,
+            "sha256": "cc6c2011be746002ba3ff96ccefab1809863748ddd17eb102b75e7b0e134fe55"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-native-offload.md": {
+            "bytes": 8966,
+            "sha256": "6c2a72b5ba3708a01d229e0576570837a617e6a42dbb0b5312a26a65a6c31542"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/SKILL.md": {
+            "bytes": 16072,
+            "sha256": "36ef88826a32ac3147b869001004ded17a8d5a491e25bfe105005ba39ea9d1eb"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/metadata.json": {
+            "bytes": 539,
+            "sha256": "ae71c389dfdb1a19d267feecb4c6f1fbb4c9afad501ff6d8fab7e38b29bcc37a"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/references/deepinfer-zero-overhead-scheduling.md": {
+            "bytes": 10894,
+            "sha256": "f452bf40132c9195d739f1c86b3aa18857138c7ed0f1861af5853d93b9955924"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/os-process-placement/SKILL.md": {
+            "bytes": 10327,
+            "sha256": "2003492e4d89fde0443b180d74933e764bd1979435e109c8d26e20eb8f6e5053"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/os-process-placement/metadata.json": {
+            "bytes": 1399,
+            "sha256": "ec02995c075f0a540921f57b4846f2cafe334eb90f44389f17c3c50457c3e9c5"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/os-process-placement/references/llama-numa-thread-readset.md": {
+            "bytes": 6617,
+            "sha256": "748b254b3b043a5ddb0f480eb4a566854d27eedd7b111499863e9bf0a5a6951e"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md": {
+            "bytes": 8813,
+            "sha256": "63b883c10a8d3e7d918abfc7175b72e2b401ff1e02697f94b8382ec1c8f369b7"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/metadata.json": {
+            "bytes": 558,
+            "sha256": "0a41c7b00978332b54ee1e4a7c1ade88ecde3e961128f24d1b4f7708caff5ff5"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md": {
+            "bytes": 6402,
+            "sha256": "0a5f48d9352c5bf460775ef47f9136829978811f4a16e84983e299782cae305d"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-serving-preemption/metadata.json": {
+            "bytes": 537,
+            "sha256": "77ce5ba16245cbd80d46d0abdf80aaf26fc0924fa2d2a5624ff430de68a1daf0"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-serving-preemption/references/source-notes.md": {
+            "bytes": 30405,
+            "sha256": "f29fcecfba0490008a354c11b42403983fe165e56ff907ad61cc6040d2b6407f"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/SKILL.md": {
+            "bytes": 3181,
+            "sha256": "1e64f28ea71672fb53c56f6108b87864738c6d03d6dcd32eedaef9e0625e96e3"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/metadata.json": {
+            "bytes": 478,
+            "sha256": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/references/source-notes.md": {
+            "bytes": 10548,
+            "sha256": "61b1609f71c37b61d136b8c4d39f793de87cffb1cf31c8ec737a1c3fa6fb1b4e"
+          },
+          "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/SKILL.md": {
+            "bytes": 3884,
+            "sha256": "ac773c6f1c110a664dfca28a3ae9129b10c07614e60ca3640dea776220e7b352"
+          },
+          "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/metadata.json": {
+            "bytes": 648,
+            "sha256": "261ce9f644ea6eebb90c830f41fb8b320d2b401456e5bd8d494095a44b8620d0"
+          },
+          "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/references/source-notes.md": {
+            "bytes": 8415,
+            "sha256": "9696d3c0ef96b6611787446bd6f96d4d6255ca8d107105de8c1d9af83edb9fa5"
+          },
+          "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/SKILL.md": {
+            "bytes": 6416,
+            "sha256": "6b7db1eb2780dec27858cb7aa0a3c093c5b5dfa4c30261aebfeaa255213ff688"
+          },
+          "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/metadata.json": {
+            "bytes": 582,
+            "sha256": "4dc8659806539b4784ea9348a01d4194ac70d0bb297a916f90ff6419efa5c328"
+          },
+          "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/references/dlrover-persistence-replicas.md": {
+            "bytes": 8398,
+            "sha256": "055f4357f79026fcbcc766bbe77db610376ec353623f8898f5d14a30ee69e7bc"
+          },
+          "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/references/sources.md": {
+            "bytes": 7984,
+            "sha256": "9425b1aab963a2c8d35f3db897bd15ee69699bf5b0d0c8425c3c8bba3e9a0233"
+          },
+          "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/SKILL.md": {
+            "bytes": 3068,
+            "sha256": "25d04aed9782b2a8ff1186828becdc7fce93c4418039ceaef9b8dcc5ec9bcd24"
+          },
+          "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/metadata.json": {
+            "bytes": 478,
+            "sha256": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7"
+          },
+          "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/references/source-notes.md": {
+            "bytes": 8521,
+            "sha256": "9f9184d46b19312138a6e805e439e43dc0594a0cc6b46675255a582dc946ca5b"
+          },
+          "src/inferrsi/data/skill_library/methods/dense-performance-evidence/SKILL.md": {
+            "bytes": 8305,
+            "sha256": "53cc1867a13f54d1006d50517dea8848869ca50df3f0cd610b2424cc8d18c933"
+          },
+          "src/inferrsi/data/skill_library/methods/dense-performance-evidence/metadata.json": {
+            "bytes": 707,
+            "sha256": "a0d1e3b17140295510a764045d1b222e19be3a3406e6d1e7dddaad49ff474490"
+          },
+          "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/SKILL.md": {
+            "bytes": 6667,
+            "sha256": "82d0099d3c33706f46c657d4fa7f76d3d486e019b36760d1de001921328a60cc"
+          },
+          "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/metadata.json": {
+            "bytes": 547,
+            "sha256": "37379a4d4a058ca46da2b20a920dc01be9ec11e557c32b2332702ec76027f4b5"
+          },
+          "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/references/source-notes.md": {
+            "bytes": 44796,
+            "sha256": "ea416831fd7d0874e06d0a76fb0c7196df53072b717673a0cdf635db6bfd917b"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-benchmark-design/SKILL.md": {
+            "bytes": 8974,
+            "sha256": "19ebc1f567aadf49f7a534a18fdb629c72c0e4ce043d360983136491cb72bfab"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-benchmark-design/metadata.json": {
+            "bytes": 5923,
+            "sha256": "18310402a8ade03fd48ac687f79ba9d447871e90337de84c0e1bc15c0b827306"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-benchmark-design/references/benchmark-commands.md": {
+            "bytes": 8466,
+            "sha256": "0898424e0bff3a6246e1eeeda6abd6c5bd974c047fe32f3bbd5ab9d31ce17cd3"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-benchmark-design/references/task-benchmark-selection.md": {
+            "bytes": 25086,
+            "sha256": "0d4bb080a754e26846e81680023db0dff5e6a4c96be4327c411c3b72ac4ef8e2"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/SKILL.md": {
+            "bytes": 6644,
+            "sha256": "89a6a82c692f7f5192f4d7f4c3a5ff08a4df3b1c4bdaa0aafb7e04292d8aa5b4"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/admission-drain-occupancy.json": {
+            "bytes": 1354,
+            "sha256": "3319e23bbe2bd40e2d45fb89dc87bdc92d4df2b5ddd8086fcf84813b43885b7c"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/admission-drain-occupancy.md": {
+            "bytes": 3134,
+            "sha256": "e88bbac5d253a8e7e3a3d5e21047f23f2970c11ac23887d4303ab94fca52dd27"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/drama3-rejected-directions.json": {
+            "bytes": 1459,
+            "sha256": "2cc0a526fe0509551607a03e7626dd3d7844e643c4afb194837f75cddead8b12"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/drama3-rejected-directions.md": {
+            "bytes": 3557,
+            "sha256": "6f143bc363a3290fc008c08e57cb075f1776d1c8cca8a6a59297d5b1fbbf61bf"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-ascend-rejected-directions.json": {
+            "bytes": 1293,
+            "sha256": "b3d538f79c56724ccb2895cbf50cf2e6b0fec3979be62a684a8230cd932dff55"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-ascend-rejected-directions.md": {
+            "bytes": 2564,
+            "sha256": "be396983812262515899a3e2ba4f01b23ff92e1b836881882c7580372b8ae552"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-b300-rejected-directions.json": {
+            "bytes": 1196,
+            "sha256": "146ea1564a46ed03092c7dc5796b318fc05b04b1821105ee00eb7fa7fa1333e2"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-b300-rejected-directions.md": {
+            "bytes": 2656,
+            "sha256": "cce2930f544a0f78f21797fba4d8ded298e4a81f76c34d7977b0e31d361bc738"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/metadata.json": {
+            "bytes": 535,
+            "sha256": "fb55067fb45970319787068a3206110641cceaf0ccbabd706be6d0b3d2afc2ed"
+          },
+          "src/inferrsi/data/skill_library/methods/kernel-writing-guide/SKILL.md": {
+            "bytes": 10029,
+            "sha256": "01fb1712d7eb988c5e412f02340d2e48b4f8564fcf4ab232dc5897869d0d50a0"
+          },
+          "src/inferrsi/data/skill_library/methods/kernel-writing-guide/metadata.json": {
+            "bytes": 653,
+            "sha256": "0189d184337662056fc6952c7beffc1bf6154c4fa173ae3e3e463bd22d70f36c"
+          },
+          "src/inferrsi/data/skill_library/methods/kernel-writing-guide/references/akg-affine-tail-guards.md": {
+            "bytes": 11941,
+            "sha256": "113deb2db10b171db3425bb3f1792bee763706b794c8cc2e222b4cc0d0114118"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/SKILL.md": {
+            "bytes": 12914,
+            "sha256": "882e7850b9b49ad692a2d11618701afe0f4e83727ea78df114be2029923c725a"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/metadata.json": {
+            "bytes": 1403,
+            "sha256": "9f1fd8a2af52bc964021e21f16ecbdb891e08f2057370e700d67f86605d333db"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 3597,
+            "sha256": "157e74ba180426b279b548a729769f152bf18f9a78ffca4ebfcaf26b7d0c5292"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/MECHANISMS.md": {
+            "bytes": 5875,
+            "sha256": "18bb2f5d7e52773c516420a142edfb531d0b9c822a2934d60f3234ac1cb9d40e"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/NEXT_GATE_LEADS.md": {
+            "bytes": 941,
+            "sha256": "16a264c7159f9c2c700ce8ba46e63f3e636b277c987da4970f6d2fa085fcb053"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/knowledge/adaptation-interfaces.md": {
+            "bytes": 3893,
+            "sha256": "5e761957cb5ecd93926e6ffc13deee7c830c6e8f0ee3c3846e9dd3b50dc74446"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/references/source-notes.md": {
+            "bytes": 13370,
+            "sha256": "54d83c3713045d59a48821bcf93598374ba4a630e3891ad39498f919317c4fb8"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/updates/optimization-search-policy.md": {
+            "bytes": 4238,
+            "sha256": "d3db44fa9a6831b86a02c638b323ed943e69fe256866213f10d68c72d762879e"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/SKILL.md": {
+            "bytes": 15324,
+            "sha256": "c5f78961c23e3c8d659f5675b522518ab06e1f2a8ea8bc74c5d428ffb41693d5"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/metadata.json": {
+            "bytes": 3669,
+            "sha256": "60b5b24ead590b0fb222398fba47822bf6723690da683e256a450bb5c0eeb806"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 7349,
+            "sha256": "083e462e0a204dce599362d46712fcfb392d053c5cda56216e22e28d4b8fa0fa"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/MECHANISMS.md": {
+            "bytes": 4003,
+            "sha256": "742c249082ceca230326f4407a40130fe624553b796ae5a157a8a4c15a7e1398"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/README.md": {
+            "bytes": 1204,
+            "sha256": "17b2d1c7c14124d59ac8a0d5ad4200ad2803d991e784a1a98cd97188cc364999"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/READSET.md": {
+            "bytes": 3937,
+            "sha256": "66459ce393858b2b2160f1e2973373bfc4b9902a12644631934b4ebe3140296a"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/STATIC_CHECKS.md": {
+            "bytes": 1111,
+            "sha256": "cb0aa7d148352ebb7b1678fc42a557a099575054716a341dfa25b6a998029064"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/cases/decision-cases.md": {
+            "bytes": 5491,
+            "sha256": "781a675a6533e0a18c044aaaee0d010da7013b6504654b8b8292a35176db801e"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/govern-optimization-experiments/procedure.md": {
+            "bytes": 3823,
+            "sha256": "ce036a4d8a0908da204515c47e2ff9623548d09de264fa65bd58fedce52a78a0"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/govern-optimization-experiments/references/source-notes.md": {
+            "bytes": 21659,
+            "sha256": "f9a8254e45473ddb2b73e576957b6d3db83c62b99dff1922f4d1e606e569559e"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/knowledge/bounds-predictions-and-decisions.md": {
+            "bytes": 5297,
+            "sha256": "5780d6eb5257612ec45deb94f85f334a6ecabf7e466d1678706041f6549bda79"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/redirect-cross-layer-optimization/procedure.md": {
+            "bytes": 3648,
+            "sha256": "6593a63fce9c474b515663a560c94b7df048dc5f6e77a95b8e54e4a64cd10a09"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/redirect-cross-layer-optimization/references/source-notes.md": {
+            "bytes": 26469,
+            "sha256": "6ff3a7bc31f1172db92aedaf7df1f1aac688bf769227b583a4cd82bfad52da67"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/agentic-kernel.md": {
+            "bytes": 41353,
+            "sha256": "7190573b85f94a4391bb0d887ccbd37a31b3f08bb0531114ad7afd72fe1567a7"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/flashvector.md": {
+            "bytes": 9830,
+            "sha256": "3926369acd88ff3c370bee5718e74dd15f3219c3e4f6311095e30e917bffa4cf"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/inferencebench.md": {
+            "bytes": 9294,
+            "sha256": "21f47111197a0b7f8d6384ea918d139dc8318a62ca3313ce060048271f50b639"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/prior-coverage.md": {
+            "bytes": 3254,
+            "sha256": "35d47ac078591221fcb666219bcd824f9400fc17aafecdc0499d8336f7178238"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/vidur.md": {
+            "bytes": 7736,
+            "sha256": "d38602af5219f5feeb49e6cd6bf53d66098c77fad141dae676f1d8a6743cf25d"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/roofline-contract.md": {
+            "bytes": 7091,
+            "sha256": "419ec4bcaec45ac579a133bce083d4f48779f26e664d7a86f720fa65f343b1d1"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/roofline-example.json": {
+            "bytes": 2015,
+            "sha256": "597cd3205df043cff3ae9d15924c77934e8d8eb300c00c4840f6bb2921f90faf"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/scripts/roofline.py": {
+            "bytes": 9004,
+            "sha256": "524de48a6764401695fe1444f1fa3d38f363069b0cd2fba4715a86ca4b28a5d4"
+          },
+          "src/inferrsi/data/skill_library/methods/plan-activation-residency/SKILL.md": {
+            "bytes": 3507,
+            "sha256": "542d8a1e6963f5019986e1fbd79d3a247a045a51487c09313026087ac457c15e"
+          },
+          "src/inferrsi/data/skill_library/methods/plan-activation-residency/metadata.json": {
+            "bytes": 478,
+            "sha256": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7"
+          },
+          "src/inferrsi/data/skill_library/methods/plan-activation-residency/references/source-notes.md": {
+            "bytes": 13907,
+            "sha256": "5ff52655cfd06fcf970ed2a7d4d4fe1c31115b102af2d41497de0c74848ed2b3"
+          },
+          "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/SKILL.md": {
+            "bytes": 13800,
+            "sha256": "7e20d3390f2501263e5e298e026c47cac16c234f2397833131e15ae5d34c0b97"
+          },
+          "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/metadata.json": {
+            "bytes": 1903,
+            "sha256": "d0e4d6950250e4cec43d82e23324c9301a925d4c22be30b4d2713d790cb6504b"
+          },
+          "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/references/capability-presence-audit.md": {
+            "bytes": 11461,
+            "sha256": "c21eb4a6aabe811a49a6540a34c54dbcfdcf7efae3052a79b54034c7838026a5"
+          },
+          "src/inferrsi/data/skill_library/methods/skill-evolution-playbook/SKILL.md": {
+            "bytes": 9448,
+            "sha256": "9cd37f21fe0d862069e4798edca8a04d59f8b840e55d521901594a63dddd9ba3"
+          },
+          "src/inferrsi/data/skill_library/methods/skill-evolution-playbook/metadata.json": {
+            "bytes": 527,
+            "sha256": "92b36338b9257e2e41b95e4370694d7de52ed84bc7971056b3bfced07b0993d8"
+          },
+          "src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md": {
+            "bytes": 6865,
+            "sha256": "64bd2c6b795d966045d1d3cca55976d3d49b08b06df7c7e06a70c2824cf55b7e"
+          },
+          "src/inferrsi/data/skill_library/methods/training-input-pipeline/metadata.json": {
+            "bytes": 587,
+            "sha256": "80659099fce6a6ec05e9084b75452c877a5812e30f60a78f31ef8c7efcdc1bac"
+          },
+          "src/inferrsi/data/skill_library/methods/training-input-pipeline/references/sources.md": {
+            "bytes": 2637,
+            "sha256": "60e0f3e88655d8e1d5eb4797f39ca4370deda01614f265b8d2ea4559d8ac020f"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/SKILL.md": {
+            "bytes": 7200,
+            "sha256": "bdfa5a4ef37d9f303c3a23cd6f98c21137b97ac5ed58e0ed5fac6d71d33e52ef"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/metadata.json": {
+            "bytes": 976,
+            "sha256": "d42b614af9cee6ee0d4a25dc5d57c7b98438feed1e807a7c9f1755109a86e9cf"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/mindspeed-historical.md": {
+            "bytes": 3384,
+            "sha256": "eb97f935ed30517a8fc9edec5e9cddc04f5d679d156c16d75bd29c534593360a"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 7368,
+            "sha256": "c55c6ad2131ef788866411dcc7056f221420ebff30c95c6bc919b09a4818cb73"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/cases/paper-grounded-cases.md": {
+            "bytes": 5649,
+            "sha256": "19bd25b1a944adb2351ebc6b1d795150ea033041a44f6cc261e881677812a101"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/knowledge/offload-delta.md": {
+            "bytes": 2026,
+            "sha256": "f8740fc5c8bbbe5ac13ff1cfec7ca499569d0b46e8cd56cf1f5a44c3a5e72242"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/notes/training-memory-scheduling.md": {
+            "bytes": 18660,
+            "sha256": "9aa6d3e01452dff95c8a8940ff62ead8c146ce215c9175b78c9f2ef508a7b2de"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/references/source-readset.md": {
+            "bytes": 22946,
+            "sha256": "2a5c166767f56fa47e7eba7386f1f36da8cb73702e5b5bbe14d6a456ca9ebfae"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/sources.md": {
+            "bytes": 2752,
+            "sha256": "6d248f58af8b350fc66c22898988d6f451ac60d03bcd805cd6e84df5e55174ac"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/stas-decision-notes.md": {
+            "bytes": 4086,
+            "sha256": "f1a993e6c4ca9a035a2effcaf5500081f227b0cc56cc885a6b22fda0041fc744"
+          },
+          "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md": {
+            "bytes": 25969,
+            "sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213"
+          },
+          "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json": {
+            "bytes": 2418,
+            "sha256": "798cbff6f10b7c24b188080ebf233b188b0e98ef2c6fdb5e00e1fe3eff727c0f"
+          },
+          "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md": {
+            "bytes": 8656,
+            "sha256": "5a39f14b102b0cee304541e011aa31e44e6b13527d1bee9987c3f3b66c064d3d"
+          },
+          "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md": {
+            "bytes": 6392,
+            "sha256": "6c640f28ee3f0253a17b81461c865cb9f947367002cb75bf72550c4572d02e97"
+          },
+          "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/SKILL.md": {
+            "bytes": 7181,
+            "sha256": "a46110891eca860915e314d6921a22856d631453b22b17837fe30262d6780535"
+          },
+          "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/metadata.json": {
+            "bytes": 2175,
+            "sha256": "4e6100dfb9fc7bc00b19915f4fd38ad509704f2a609c658adf4d632d4e05b6c1"
+          },
+          "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/references/afd-plugin-LICENSE.md": {
+            "bytes": 12747,
+            "sha256": "0e8566ea362276878e1f38f97da49964dbea301c15d4aa9d727f97a4c39fd943"
+          },
+          "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/references/source-notes.md": {
+            "bytes": 17240,
+            "sha256": "5073d87450cb60d1b245308dff8807e73749c07de976d50f2fa26694fad81a72"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/LICENSE": {
+            "bytes": 35149,
+            "sha256": "3972dc9744f6499f0f9b2dbf76696f2ae7ad8af9b23dde66d6af86c9dfb36986"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/SKILL.md": {
+            "bytes": 4507,
+            "sha256": "c993bd6ff8e9b5f82f39f175ec8242341136e4ee0808af414e1d94ad3e196c5e"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/metadata.json": {
+            "bytes": 2032,
+            "sha256": "74742851a002a51c208362cdcd2a59fdab5d1734d8492b1f93fbec3cf21e0841"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/references/graph-cache-invalidation.md": {
+            "bytes": 7679,
+            "sha256": "614c9a4e6943d24d7d7be7d632682ce8b1dcc239124a37d9dc1d487113772eac"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/references/model-residency.md": {
+            "bytes": 7191,
+            "sha256": "014c68386e7cdfdb7b4f566836e8b2ff61bd5a5a506e7b4723f83fe80c5d4de7"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/references/repo-provenance.md": {
+            "bytes": 5648,
+            "sha256": "ca996ef515af42a3ce304747fa3f12a31c00533deaace5ba2ce21aefa0a62e2e"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/references/validation-boundaries.md": {
+            "bytes": 3523,
+            "sha256": "1913dedbeb558d872d6a0e015cc11d822910c3cb259a236db29fb60ad8a64263"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md": {
+            "bytes": 12294,
+            "sha256": "4703beb6e5194ef2ff2c0b5ec69f64e35417a4bc0dfb50a10c80a920b77ef3b5"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/metadata.json": {
+            "bytes": 2333,
+            "sha256": "014be24d0e669f4acbc0717dc323eec784026cf185186619d5a19d6d84496131"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/source-notes.md": {
+            "bytes": 12036,
+            "sha256": "fd917bd5ac80bee9c9f4525f8e0300f6d67dc144e2a16632af42a67f3d53d098"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/validation.md": {
+            "bytes": 5402,
+            "sha256": "60e71e70c5519bcdf6c171eec2be2f342e8801bedd746d6005f11adc646dfaff"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/SKILL.md": {
+            "bytes": 11202,
+            "sha256": "f157562d1edc905cb4fc07246567e3c5b844efb9fbf5fe39f0d40553a1b1a307"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/metadata.json": {
+            "bytes": 2156,
+            "sha256": "4d68bf4c1767eaee12ade963bbd551d0e61d1bcd81a984bb02fd3aae46965919"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/references/source-notes.md": {
+            "bytes": 11633,
+            "sha256": "6fa955022f16cd598f44fd6521901d1cb868b06cec545cabe2b82769a641b5d6"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/references/validation.md": {
+            "bytes": 7196,
+            "sha256": "673d93f6700052974b340fab25b3438e0e6344960569ac9abec7d7720f662288"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/LICENSE": {
+            "bytes": 1077,
+            "sha256": "663e51edcd9a3a8f858018ce6924ad2895a24c269c26c39ad75325bcd0f7e5bc"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/SKILL.md": {
+            "bytes": 5432,
+            "sha256": "fdb4d9483c6efa836f9e59617da605fa9c2186e11c2041b05d5c81850ba0475d"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/metadata.json": {
+            "bytes": 4923,
+            "sha256": "fe626a77af3f084f36fd73ffdb6f67976e7dceeafb94ba22078ef03c3a96cfae"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/THIRD_PARTY_NOTICES.md": {
+            "bytes": 1774,
+            "sha256": "a0d77db7d2aca0eda811ab3c386cc32b3143f2c2609a042cf4642954f0f6fb09"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/minimax-h3-apple.md": {
+            "bytes": 12059,
+            "sha256": "57ce01754ba96c378f13a91305ef3d418b926f99b2753f0af99315572bce063a"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/minimax-h3-weight-inventory.json": {
+            "bytes": 10664,
+            "sha256": "fbaacaeec340d3ce893556ee864405a0746af4d13530a38cb3d5aa9d60929271"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/phase-residency.md": {
+            "bytes": 9408,
+            "sha256": "2836a550eb2be7f5a862f51ec7b135ebd9715bf3c4faa84a7ecbc9d9c11f5d8d"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/precision-and-streaming.md": {
+            "bytes": 9440,
+            "sha256": "7e166c89193f4c7aed852ec95346a2de2b5008dac002e2e1862be1b3e57f6093"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/source-identity.json": {
+            "bytes": 7747,
+            "sha256": "730f8dd23648a3f8f47b9726cc613d893a79d1817d515eec609a9b06d8a9ff90"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/source-notes.md": {
+            "bytes": 9448,
+            "sha256": "7304bf9369efc157dc7aa40686dd828cc14db45c49442e908b3d3cc4b329eb86"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/validation.md": {
+            "bytes": 7880,
+            "sha256": "0754873a758e048480938314894e25b504ae0c269549312e7f8e6020d83fe294"
+          },
+          "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/SKILL.md": {
+            "bytes": 9338,
+            "sha256": "95e81b6bc096a954fa73139ea5bc001d1efa6c99747161da62f012b494e5c013"
+          },
+          "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/metadata.json": {
+            "bytes": 1718,
+            "sha256": "1d30b31dc8352fc6a8d7b8abbef8118c840e47e5b90795dae0c758bf9717278c"
+          },
+          "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/references/source-notes.md": {
+            "bytes": 15234,
+            "sha256": "3d12a401e9decb55a2f4f3b2bfa0d360ef5110f63a0469dfe285654e78054190"
+          },
+          "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md": {
+            "bytes": 10812,
+            "sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27"
+          },
+          "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/metadata.json": {
+            "bytes": 572,
+            "sha256": "c221189b8d85fac3bd41ef1c7cd560f1aebc8cfa0381c182d1d571a7169be829"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/SKILL.md": {
+            "bytes": 3788,
+            "sha256": "bf1177cb9985c12a9bf11edb64771018e1e314c4713920de8fd4a989ff70115f"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/metadata.json": {
+            "bytes": 1299,
+            "sha256": "d86cecb615e93ba0f6be09d1ead102cd5a17068c07e1be2549ac5cd8f7519279"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/cache-stop.md": {
+            "bytes": 2026,
+            "sha256": "bba6b5c73261065309cbcf11b1e591eef48916f0c79ceaaf74d98a9fa96b0992"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/generation-batch.md": {
+            "bytes": 1733,
+            "sha256": "f16d3c57d45e63bcaffd3871aa41c76972ee11f7dbf1e12b45d9b170303cad9b"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/source-map.md": {
+            "bytes": 6358,
+            "sha256": "91d26181f246307f2113f94b12ac9587d689b654dcd216e9f14b11cf5ebb650c"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/validation.md": {
+            "bytes": 1665,
+            "sha256": "73b5dd347017baba519ecc85415aead9993c1751d116b9ba4f07689f58c5f783"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/LICENSE": {
+            "bytes": 11352,
+            "sha256": "a54c8553df5d4313d293d1aef3bf2f13952bda8b7c7a1fdd729b2bbe27ee6015"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/SKILL.md": {
+            "bytes": 2304,
+            "sha256": "d75f5b2881b240ec2d9d4f5cc8e3f7d53e4378cb01732a2fb9699da2901ba38c"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/references/repo-provenance.md": {
+            "bytes": 1778,
+            "sha256": "5e8dbe168a8885e209d4b4532944ffa7fead658b23fd6588054d57b22476f63b"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/references/repo-routing-metadata.json": {
+            "bytes": 511,
+            "sha256": "6820707cb9bd877d1635928cf5a0efe38e4fe37dc04036dc70d53e479ed9bf7c"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/references/snapshot.md": {
+            "bytes": 3488,
+            "sha256": "0fb5d776349de845dc314f16ce73a06c9caf5db84dc1c293327d3b3ca1337c80"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-architecture/SKILL.md": {
+            "bytes": 2355,
+            "sha256": "766b2d66d2056dd779abd9e492cb3af6e25899298f44964f565509f130a7cb24"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-architecture/references/runtime.md": {
+            "bytes": 7734,
+            "sha256": "36bba9874fe054acc5217f9c265f5ae309c83943e0d46228370ae1bcdc69421e"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-extension/SKILL.md": {
+            "bytes": 2193,
+            "sha256": "9af38bca840d0004fc2ed049141476fc40fd40dcfa843f38a404a7a314c19425"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-extension/references/integration.md": {
+            "bytes": 5775,
+            "sha256": "6ce9383ca17bacff60a180f8fb8c43ea47eaa37117fb8da6051ad86135d228ef"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-serving/SKILL.md": {
+            "bytes": 2335,
+            "sha256": "990b6bdebfe22612de49dbebcb16810e8abef94d70691aa883b7937c85210eb6"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-serving/references/recipes-and-tuning.md": {
+            "bytes": 8918,
+            "sha256": "94afb762bbe185f58606154b3f0a5af56e7089cb84ea94e26cfcb9f93c95439f"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/LICENSE": {
+            "bytes": 11346,
+            "sha256": "1495e1e757ef4d0925a2350563cf5754bb23c51701a8ec4fb3c5cdcbedae6747"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/SKILL.md": {
+            "bytes": 5466,
+            "sha256": "ae72d1f0cc750699bf79d7a138a97219209bca7b7e5f19cd427d306245eaf841"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/LICENSE": {
+            "bytes": 11346,
+            "sha256": "1495e1e757ef4d0925a2350563cf5754bb23c51701a8ec4fb3c5cdcbedae6747"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/SKILL.md": {
+            "bytes": 5738,
+            "sha256": "b0a4ba4751087126143ef86983798deb0ed00e24b6ff06bf6ef76a9df6e274f2"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/metadata.json": {
+            "bytes": 2844,
+            "sha256": "6ed57dfb4f25ecb34b81f2188b055e05d172ba2581e6a3b605b814b2f733a4d6"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/references/source-notes.md": {
+            "bytes": 19512,
+            "sha256": "d0fcfd1177ae2f351041c5b2abba7f5121cec133006d9e4dca8f74ddb567332d"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/architecture-and-memory.md": {
+            "bytes": 7735,
+            "sha256": "798e747022b7ec135cc4861c2c9ac44fb14b503bb77a0068ee7fbea494b764a7"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/native-candidates.md": {
+            "bytes": 3961,
+            "sha256": "abafcc9a34df196a6365145554887b8d37afb36c6e236c92e50a4a06c8a4ec93"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/parallelism-and-extension.md": {
+            "bytes": 6324,
+            "sha256": "e3be7f4b51264ac68682a64d264e4d96857d5bb9f9d99cae5f327c56984a9e6b"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/repo-provenance.md": {
+            "bytes": 2274,
+            "sha256": "534d95ae8d0b29bcbc8f6707b3d38fb964e31f5b66ba13517a9a6b8f309bbb38"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/repo-routing-metadata.json": {
+            "bytes": 625,
+            "sha256": "b2b879c1cc42ddc2a5b95ae24115eb891562fa9a6e96d151f4ddf6f7719fe73c"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/source-map.md": {
+            "bytes": 8945,
+            "sha256": "c2adad64ceb0c305cdc4859c03012b3e0a4ab26cd0cb289bbd76e192b0cc3794"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/tuning-and-validation.md": {
+            "bytes": 10832,
+            "sha256": "49b3343b3292cc098a787de9d24d968b444d0ef4f73fedb5d8183390e8f39c9c"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-parallel-extension/SKILL.md": {
+            "bytes": 2316,
+            "sha256": "18bd54ebdd629e1ba6dfa589f391c7cc3c5355785c2ae82f741b64ee8022464f"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-runtime-memory/SKILL.md": {
+            "bytes": 2154,
+            "sha256": "33b1a65aba00fca819d2b6cc42202e7708e9155e4d8f4373a2387f4ee2f2fa3e"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-tuning-validation/SKILL.md": {
+            "bytes": 2018,
+            "sha256": "744b2ef9122101d3f50cbbb40fe020e2b3d3888adbe2d86041041e7c7ce424fc"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE": {
+            "bytes": 1062,
+            "sha256": "b53fa08f515cb5a6fff7b9fd8fcd0961a4b80df29d74372d25e1f9171aa042ee"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md": {
+            "bytes": 5504,
+            "sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json": {
+            "bytes": 2202,
+            "sha256": "f7b7581a3f44ba9774480c559408fc66d6a4d6d1dbbeabe4529f1eb7d4dd5ad6"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md": {
+            "bytes": 8136,
+            "sha256": "7b537cb972347dd68e303960f142892346d8b19b884e4d08016ea36a23bdaa7a"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json": {
+            "bytes": 3617,
+            "sha256": "af4830370b6eddc2d176a9ae14da4ab56a90b0fe048a01a89ee051df4365d8e9"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md": {
+            "bytes": 6589,
+            "sha256": "b67affbd7d237d82b7f51be49f1537e9820fe56ec70a71ee76809f9e599ca693"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md": {
+            "bytes": 3394,
+            "sha256": "5dcf5fa7bf734fbe13214f57d2dfb40ff8215d9cd34e072a9e615bf103b82fc2"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/LICENSE": {
+            "bytes": 1093,
+            "sha256": "949b9b0aada7476006469e58ee4ffb0e0f68c2d9137e90a356757b61a09d37f0"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/SKILL.md": {
+            "bytes": 2813,
+            "sha256": "0e5daeb6409c916be45ecd7ce2531ea5f1c60a4a53e2b0ed25dc1e2cdc5c17ed"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/design-space-analysis.md": {
+            "bytes": 27736,
+            "sha256": "11ecafe71530ddd4839268ea1b24b3361a3895012a705859ca06dae29fa72630"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/repo-provenance.md": {
+            "bytes": 2913,
+            "sha256": "05f17d0ae996ad6c12314bdd29f3427574f6aa611ea60effd18598a0f386638b"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/repo-routing-metadata.json": {
+            "bytes": 332,
+            "sha256": "a10ba49b7e6565109b566c105e17551d042075a1ff649c6f371134e3b7402c74"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/trial-format.md": {
+            "bytes": 2291,
+            "sha256": "92d8cf38549268ae4eef382890662d03e18a7dff858ad45ee5eb6f47eb39c8c0"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/troubleshooting.md": {
+            "bytes": 2079,
+            "sha256": "7301cfa5a5f6cfbf9576c3845a96c6b4de8f8a6509f2330970a8e2e4669b3d5a"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/scripts/summarize_trials.py": {
+            "bytes": 5170,
+            "sha256": "6a243a08da15659ced6519f49f52a49a675bd9e952740c3ab03dd21964eefbe2"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/design-space/SKILL.md": {
+            "bytes": 1826,
+            "sha256": "857bea3c64db45dc035728a41523142ac65778b41f63092148b8ab936f971d76"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/design-space/references/design-space.md": {
+            "bytes": 24231,
+            "sha256": "0a428dcd12d5290b554e9da6795a110f94d98f128771602e1ea5f608048b135e"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/tune-execution/SKILL.md": {
+            "bytes": 1145,
+            "sha256": "cdf51c537e7237296d52ef6cda9f2e4c9d9ce3fc87265a369cd9eb7497c53512"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/tune-execution/references/tuning.md": {
+            "bytes": 11310,
+            "sha256": "7b69aa7872d60dd3fb91e41b77a632572070dcf80dd5b5f6dc1722d8540177b1"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/validate-engine/SKILL.md": {
+            "bytes": 4613,
+            "sha256": "90bfd975e25f18895ce0e85f6f35d24149ce026305a4d479aea4924510053460"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/validate-engine/references/validation.md": {
+            "bytes": 16895,
+            "sha256": "dae846f4c47c68731b96ec1d50cb008c03871c72078f1c5b0f4a9423be9ba397"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/LICENSE": {
+            "bytes": 11357,
+            "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/SKILL.md": {
+            "bytes": 4945,
+            "sha256": "df3e63dc28de2dac2aa92824fd88719159138650d96487332e0effca3f0e2a80"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/references/evidence.md": {
+            "bytes": 4689,
+            "sha256": "ae0d51ba56b34ed81fba34de8fd3685bebf2f160b188fa53719936002881d220"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/references/repo-provenance.md": {
+            "bytes": 2221,
+            "sha256": "43bbfe431434ca4ad6157ba35cbb9ce2d89e4065f427abfdd5f372f51ba0c9c9"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/references/repo-routing-metadata.json": {
+            "bytes": 342,
+            "sha256": "7784b4d98d096718c4f4cc42ce86d5ec113c469844e12de4690b0fccca59d329"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-architecture/SKILL.md": {
+            "bytes": 1336,
+            "sha256": "fb3ff50a7b9122a2545a97ef5179011237c4c4ca214dde7236ad7bf79a7c8623"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-architecture/references/architecture.md": {
+            "bytes": 5600,
+            "sha256": "db845eb4207ddeedb8ae381f1b91704a42cb53ba0ee6c7ee581510ba1a412190"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-serving/SKILL.md": {
+            "bytes": 1343,
+            "sha256": "8ab3209b838635395b22deca6aa879172016531d45c2374eb8e498e7f71bb40f"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-serving/references/serving.md": {
+            "bytes": 5049,
+            "sha256": "9bd5eff300d9755137a7600c8f7a7753bb22d09c8853e08062d7e535662ba8a3"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/SKILL.md": {
+            "bytes": 2227,
+            "sha256": "93c83bc4b799b32f50ff54b870cbbde050a790af16d9b7b062039c4987c80148"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/references/abort-acknowledgment.md": {
+            "bytes": 8547,
+            "sha256": "e95ae5b2cc410917ddf5e9066890d0bef9264104529bdca5ac12bdf56fe80ae3"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/references/validation.md": {
+            "bytes": 4632,
+            "sha256": "1542061f9c76fff2acb6529a58238301484b0e23ec0c697d2fdf129e200ea973"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/LICENSE": {
+            "bytes": 11324,
+            "sha256": "50e6751797c50dedd75ef1b8a0d9e42f5f8472e9fbce91f34718e9f97b0c780a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/SKILL.md": {
+            "bytes": 2863,
+            "sha256": "38e522b8abeb0dc140a2194340432ed41dc8b569d33b63d5d4c59aa1092cd2d1"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/SKILL.md": {
+            "bytes": 6664,
+            "sha256": "313733a65cc48442b18dda1c4665139a689963a90e1032a2235bc6f40e3a8e06"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/metadata.json": {
+            "bytes": 781,
+            "sha256": "b43fc3a7ccae59c664fcdf9ff77962edcc07c2a3df8b59324db551642bd23e23"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/references/source-notes.md": {
+            "bytes": 13853,
+            "sha256": "3665b034594fc8cc767c408002f4e16f8803369242405027c7f37ce6c9ecebe9"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/SKILL.md": {
+            "bytes": 6565,
+            "sha256": "5fb9ea1d1373e110d9b506f69963c4857338b7dbf5e38ec9a49b59a0ff683ed0"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/metadata.json": {
+            "bytes": 1372,
+            "sha256": "78459416734f61a8faa819d9a1a3ad7b9a985e8bf8e149813c604d15be6bb92f"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 9006,
+            "sha256": "82b24be82451ee121892f61575d0cf4ecb9920813a25ccb26cce2b45dda6d8b3"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/migration-is-not-free-memory.md": {
+            "bytes": 2145,
+            "sha256": "2ac168f2725cba7fc92dfcc67c01bf2258dff6a2dbe1ddb063bc60c175d82d0a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/parallelism-change-with-state-cost.md": {
+            "bytes": 2069,
+            "sha256": "5accc81086e4500d66f894ef6af9951e1cbb76941f6c631aab5c5561a786d8d8"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/preemption-under-kv-pressure.md": {
+            "bytes": 2630,
+            "sha256": "462200a6d0b193ef01f32374a463e0f2838ca8edf33d9ab898def65ffd0b490a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/knowledge/scheduling-state-boundaries.md": {
+            "bytes": 7488,
+            "sha256": "d4ef29869959b835a5dcb06ddb2fd8e6cff49bb176994dc1d46eb9a852f6228d"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/fastserve.md": {
+            "bytes": 9289,
+            "sha256": "9b4e275843000ef453d508d9c3c9c3f8ea1de5545fbd01d7abe74baa8a4931a7"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/llumnix.md": {
+            "bytes": 10181,
+            "sha256": "f252ce33ac2db017701ca1d49813c5ec1996c4c92900cf35bb6bda7fc59c0f18"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/loongserve.md": {
+            "bytes": 12422,
+            "sha256": "7bc6d1e256d461b04acc8ed856cf99108faceda9cde729ab95bcca596fe019ca"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/orca.md": {
+            "bytes": 8826,
+            "sha256": "f5a3b5870d64226408a8fcda5d69483769d823f8a5efd7f11698c8da483ab195"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/fastserve-readset.md": {
+            "bytes": 7173,
+            "sha256": "9866bd0c78b2b638998b3b632f6bae26aef032462c81abee336d309af5b07831"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/llumnix-readset.md": {
+            "bytes": 9386,
+            "sha256": "3a2819d65ad24087c08530492ef39d8ea046ed1f780d53512f1d23e2b3578487"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/loongserve-readset.md": {
+            "bytes": 12490,
+            "sha256": "f1a383e7bdbbd5a5a472d7f6e1292c8a17d4910ff736ca5a6aa422e1bee74f8c"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/orca-readset.md": {
+            "bytes": 4796,
+            "sha256": "a497ef23b7e60c0fbc15d95b7e07cd20547cfe56f37c24dfe6ed70d79ef7a3cf"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/source-notes.md": {
+            "bytes": 9575,
+            "sha256": "84033fd514773f10c0bddd9914fabb980497ba2b308f226a922c3e1888413827"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/references/evidence-map.json": {
+            "bytes": 14384,
+            "sha256": "dd8bbab5d285ed3a1a19d2f06ad84e8c444251690d8f673def3e2681765ef6c1"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/references/repo-provenance.json": {
+            "bytes": 1313,
+            "sha256": "4e705a964697712e950e02bedc41320806a41311b7f4dcf32f923d60b1d2e919"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/references/repo-provenance.md": {
+            "bytes": 3261,
+            "sha256": "6e841c38c35ff2cd6ff8139cb1a6c8f75701647897f1089a94a5c62928609941"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/references/repo-routing-metadata.json": {
+            "bytes": 432,
+            "sha256": "a20b54f682eb52311e445ac62b95d1935c001296f5cf23b7764cb04477d535c9"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/SKILL.md": {
+            "bytes": 1835,
+            "sha256": "2a945b5fddacfb90d1cd7cb2b2f767067d641061423a65e1c19afecec7f3c06a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/references/capability-boundaries.md": {
+            "bytes": 6560,
+            "sha256": "c8081d03f04f7481080c46c7dcf728962729bc122f49f9502d50cd1e3d04bcf3"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/references/runtime-and-invariants.md": {
+            "bytes": 8265,
+            "sha256": "afa1f967545021f36bc1b59dbf6aa2c367f4d5ece4f08dce755e2017cf50dbe5"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/extend-validate/SKILL.md": {
+            "bytes": 1731,
+            "sha256": "495679e5b99341bf80daeaacd2d4c2699aba16bc16c447cbd8d2fe17f7559798"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/extend-validate/references/change-boundaries-and-validation.md": {
+            "bytes": 7624,
+            "sha256": "f7fdd2d699f09d96fbf5ca9382f9b56113c870db2688d8e1118eb8163519fc4a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/SKILL.md": {
+            "bytes": 1812,
+            "sha256": "0ef4d2e3a96cbb277557e169a3fc8edf0f0504aad03259771b432a33ee008f78"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/references/commands-and-measurements.md": {
+            "bytes": 5667,
+            "sha256": "76ab51083b14595c49fa1c255cb7e5ebb996f14930a51ab3cbe864cc334b5f80"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/references/experiment-decisions.md": {
+            "bytes": 7644,
+            "sha256": "242593c0d386e57104b57a09bc7db97c2bbbcdb905bb68a397cfe612e07fb083"
+          },
+          "src/inferrsi/data/skill_library/skill-index.json": {
+            "bytes": 7514,
+            "sha256": "8ce35e18f7629260d67a95617b9918eecff73ed8d9370dd3d4595c725ada9aa1"
+          },
+          "src/inferrsi/data/skill_library/workflows/cite-check-downstream-pr/SKILL.md": {
+            "bytes": 1093,
+            "sha256": "485932cdf3b1fba5dc69be73e75d38f86a7b79442d26e82a296c963a7598d226"
+          },
+          "src/inferrsi/data/skill_library/workflows/claim-contract/SKILL.md": {
+            "bytes": 1075,
+            "sha256": "3af1dbbe81d91cb05fc2ef15a2ae25d7592a08558e1a5535f49c52f33a488c55"
+          },
+          "src/inferrsi/data/skill_library/workflows/competitor-scan/SKILL.md": {
+            "bytes": 1088,
+            "sha256": "0ede58ace65a9324a538b4816add1684f230ba536d4364fe5a5dd2c0bbed3776"
+          },
+          "src/inferrsi/data/skill_library/workflows/draft-acceptance-issue/SKILL.md": {
+            "bytes": 1230,
+            "sha256": "ce983acfb7d7d3272775e7cf00c92e4ddadd97511b602575a76e5ab7ade01b56"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/SKILL.md": {
+            "bytes": 5681,
+            "sha256": "364beed2ca617e71f4f955bbd024740d411b94eeb9557d7e055d4463cc22caba"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/cross-repository-mechanisms.md": {
+            "bytes": 17050,
+            "sha256": "28efa1b25f548763180b70d8d0ddebb7d4e7227df189f871fd233fe1d4881303"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/import.md": {
+            "bytes": 3463,
+            "sha256": "c8a57db94768cc6cd4193932def75ba67ec453e56ee0e54f176fb82f1af53091"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/model-aware-preparation.md": {
+            "bytes": 9056,
+            "sha256": "787c017aade68ee1133b979d636fa24062302cef6174c001171c963947ee8da4"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/overview.md": {
+            "bytes": 8099,
+            "sha256": "7ed0a966239b9481b82dd349245cfbfe966c030fc916771c91fd0692464ffc56"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/repository-source-scope.md": {
+            "bytes": 11967,
+            "sha256": "85202ca209171c50573c2fc209acb1e937d8211e1652e4ca9c6cc4bce04d03eb"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/source-manifest.json": {
+            "bytes": 1613,
+            "sha256": "0ce1678803d1ede6b485fb6b4b1c88aba909056ee6e3b5af691899ceb2056240"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/SKILL.md": {
+            "bytes": 6502,
+            "sha256": "74cd863df2e80dc2bda1713d24207809173a3116f30325e3487e1e3517cfec1e"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/metadata.json": {
+            "bytes": 3236,
+            "sha256": "12fc9272bd5fbf02dd9b897d8ca40984b72ae54f20492ace5304711bbc4d71c3"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/COMPARISON.md": {
+            "bytes": 4267,
+            "sha256": "38587678a96581f5c2dc697d982127343565504bffa191352811dffcc165aa23"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 5557,
+            "sha256": "fbc4a96b6f6c477706eed83eaa4fcc29a14f881637629d6c004601ed049f2523"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/QA.md": {
+            "bytes": 1853,
+            "sha256": "3f3785ec39a6938a5d1f2e874e99153fe92c0e666dd3e78a499c8b47d891346a"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/README.md": {
+            "bytes": 2151,
+            "sha256": "221785e91c45588f82b3caab6d52f1ea1d102b16139c20a0802921e2789e6d28"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/UNREAD_LEADS.md": {
+            "bytes": 1038,
+            "sha256": "6c1d87ebe491d27666703149eb05ad582c7e5aff77f0ac5c11a7b89c946f96ee"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/cases/methodology-counterexamples.md": {
+            "bytes": 5213,
+            "sha256": "f9fc21753ea41eef77ca020d78d031e945e5dd835b5d09c5ff0e53749ceb2ca4"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/knowledge/evidence-routing.md": {
+            "bytes": 5282,
+            "sha256": "406d5d74c3fef4c1a23d495492e21a084c3c46495a1a84b385703ed104358874"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/learning-to-commit.md": {
+            "bytes": 2847,
+            "sha256": "9c2d1d738e896d05c00f53e6adf14c9df2501ecc3f0020e71b5a1c2192547086"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/protocol-details.md": {
+            "bytes": 6520,
+            "sha256": "41298c6bde3a28f0712fa06a18bbc4ac59aea6d17cc9eb1318744f4121e3bab6"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/readset.md": {
+            "bytes": 7086,
+            "sha256": "6f953d20d8958c68339c9ceeae348604d0fea3e9c6ccb9b64a63d3b210426a22"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/repo-to-skill.md": {
+            "bytes": 3151,
+            "sha256": "5ebdac2a2bbbe6614f9d4096bc9895bd685acdf76414e2c3c1fb7b50d57c738d"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/skillopt.md": {
+            "bytes": 3040,
+            "sha256": "148075daa67e19f4b4351710498438d17c279a9bce56b2d01cab26ff3be33e99"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/source-manifest.json": {
+            "bytes": 2431,
+            "sha256": "716cd49ed01824ee9f6b6fc55ebf2b24fdd5d5a3e677f6e28e34a472616bbe25"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/trace2skill.md": {
+            "bytes": 2917,
+            "sha256": "230d13b77ba819b5e3edee147523e8cad6bfce95afccc140fb4946d6c8747a2e"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/updates/bounded-updates.md": {
+            "bytes": 7630,
+            "sha256": "d02c5bd0109846d1e1987de074810321cb4af277cd3f66ab2ac268aadd5ee142"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/skill-factory-design-20261014.md": {
+            "bytes": 9465,
+            "sha256": "fb8ce1ad83b7a6ddc657e89ee2eebb710873b213cfd85bdd6d823c8691b3cb9f"
+          },
+          "src/inferrsi/data/skill_library/workflows/rsi-evidence/SKILL.md": {
+            "bytes": 3680,
+            "sha256": "a55610874b658b86a39a1a049efb4e4eab0140f6579ed74dd7718a4ca8cedaca"
+          }
+        },
+        "sha256": "6e19365a0ad8e2951af0f3106f93aa545b4ad00b6e0f0317a20f32065b6a05fb"
+      },
+      "environment": {
+        "filesystem_cache": "not controlled; process cold is not OS-cache cold",
+        "implementation": "CPython",
+        "logical_cpu_count": 8,
+        "machine": "AMD64",
+        "perf_counter_resolution_seconds": 1e-07,
+        "python_executable_sha256": "3470f7919170d235d7e6079691462c4b217745ec67ee612e745730e46d98f238",
+        "python_optimization": 0,
+        "python_version": "3.12.6",
+        "release": "10",
+        "site_initializers_disabled": true,
+        "system": "Windows"
+      },
+      "git_head": "407bb244651bec1c68f4b3e94b0c697b233c4a09",
+      "knowledge_record_count": 108,
+      "library": {
+        "bytes": 2294958,
+        "file_count": 342,
+        "sha256": "be8726296e6186601692dc2d95d99bfbba76f0b4c758de3a88a9b2a523eaa37e"
+      },
+      "modules": {
+        "inferrsi": {
+          "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+          "source": "src/inferrsi/__init__.py"
+        },
+        "inferrsi.catalog": {
+          "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+          "source": "src/inferrsi/catalog.py"
+        },
+        "inferrsi.cli": {
+          "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+          "source": "src/inferrsi/cli.py"
+        },
+        "inferrsi.knowledge_operations": {
+          "sha256": "c71bb2063ab8d919ca5907ac900a3930ec60730515c614fee7375d9efb77173a",
+          "source": "src/inferrsi/knowledge_operations.py"
+        }
+      },
+      "packaged_skill_count": 60,
+      "python_sources": {
+        "bytes": 274624,
+        "file_count": 53,
+        "files": {
+          "src/inferrsi/__init__.py": {
+            "bytes": 102,
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47"
+          },
+          "src/inferrsi/__main__.py": {
+            "bytes": 48,
+            "sha256": "935a1c1166b0c1ea35a82256345000bf2c73ded718d77773bc27a71ecce28f7d"
+          },
+          "src/inferrsi/_skill_selection.py": {
+            "bytes": 5790,
+            "sha256": "4640399f4013425589f6702e2db844820619ba35967d1c3519978eb307168440"
+          },
+          "src/inferrsi/arguments.py": {
+            "bytes": 12200,
+            "sha256": "4456600cf4a362cf069594fe36db46932cf5cb14ddd37683aaf3538f5cf76b4d"
+          },
+          "src/inferrsi/catalog.py": {
+            "bytes": 6855,
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5"
+          },
+          "src/inferrsi/claim.py": {
+            "bytes": 805,
+            "sha256": "5f24c4fa91b48f0bb4d4554531607763fc8fc108e90f332816f06cfb520b5cc1"
+          },
+          "src/inferrsi/cli.py": {
+            "bytes": 14205,
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88"
+          },
+          "src/inferrsi/contracts.py": {
+            "bytes": 119,
+            "sha256": "03000cb9246ce8c0967147533f7ad25df918ba2674839801453ffb870f496490"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/scripts/roofline.py": {
+            "bytes": 9004,
+            "sha256": "524de48a6764401695fe1444f1fa3d38f363069b0cd2fba4715a86ca4b28a5d4"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/scripts/summarize_trials.py": {
+            "bytes": 5170,
+            "sha256": "6a243a08da15659ced6519f49f52a49a675bd9e952740c3ab03dd21964eefbe2"
+          },
+          "src/inferrsi/evaluation.py": {
+            "bytes": 183,
+            "sha256": "482c40710d428206c59fe3696703248cd86c4b3c7d19ad41017c1369bd6d780e"
+          },
+          "src/inferrsi/evidence_paths.py": {
+            "bytes": 859,
+            "sha256": "cc46231e0b8334ef5742defdda20e0093e22b30c86af59df1fb3ae4b5caa70fa"
+          },
+          "src/inferrsi/execution.py": {
+            "bytes": 755,
+            "sha256": "f43d8cbb4d43a72a7ff8a10b29a5fb9cf4cb22af54dd35943283c843fba0da92"
+          },
+          "src/inferrsi/harness/__init__.py": {
+            "bytes": 177,
+            "sha256": "1006e4713003f44bfb32b2e385a918e1652b9e9833801e4555674a71e57fb0d5"
+          },
+          "src/inferrsi/harness/evaluation.py": {
+            "bytes": 4160,
+            "sha256": "9b00b2e93959f16b7b76bb4b257df7dcf48520a06fd6adae96e9a165a05e6e80"
+          },
+          "src/inferrsi/harness/improvement.py": {
+            "bytes": 5777,
+            "sha256": "3faa0f5928a707a98f35f45d1ca18e226c898941e6a07b3829eebeb53e70aeb8"
+          },
+          "src/inferrsi/harness/reporting.py": {
+            "bytes": 3566,
+            "sha256": "38c58f02c44e8c8aeacf46527e142be0e3142a83e4b2693246d0ca76c0ddd7c3"
+          },
+          "src/inferrsi/harness/runner.py": {
+            "bytes": 6567,
+            "sha256": "a566307b9b586655941b627f428b5121ed38d7ac80bccd1e1799e443fb0de911"
+          },
+          "src/inferrsi/harness/scenarios/__init__.py": {
+            "bytes": 79,
+            "sha256": "4e97913c806077f0cadfeb29d986d2fa75b55ab8766cb86e6e3e9a195a9a1f5a"
+          },
+          "src/inferrsi/harness/scenarios/external.py": {
+            "bytes": 1016,
+            "sha256": "84d5f6920501a9d882cc4fffbc721d2051fdbf7c90694bde9cbfba3c2b866304"
+          },
+          "src/inferrsi/harness/scenarios/media.py": {
+            "bytes": 2914,
+            "sha256": "67d979a3ed6f13a8e3e1e40a304f78e5a3e6c2556b3aa8a11c41156745d7050a"
+          },
+          "src/inferrsi/harness/skill_evaluation.py": {
+            "bytes": 11093,
+            "sha256": "b710b8e53758757efd68c703449eb4bab7eb7ec50ae3dfc52a010cdebd8c3bf5"
+          },
+          "src/inferrsi/infra/__init__.py": {
+            "bytes": 85,
+            "sha256": "8779897733779266841b0f79b84b1dc43f0fa57c013855c906c809829ecfc77c"
+          },
+          "src/inferrsi/infra/artifacts.py": {
+            "bytes": 1040,
+            "sha256": "e72e1bebdbb88e42de430238489252386851f9e6645ed7780f2aad2e90338120"
+          },
+          "src/inferrsi/infra/campaign.py": {
+            "bytes": 1028,
+            "sha256": "be505fba1592611bd84ceaca55afbac027205413a2639baefa0c3bb343230209"
+          },
+          "src/inferrsi/infra/commands.py": {
+            "bytes": 811,
+            "sha256": "1890f115b3a2c5c6832173bf60432ad595bd6bf1c69783ecd460c6f8961194f1"
+          },
+          "src/inferrsi/infra/environment.py": {
+            "bytes": 2067,
+            "sha256": "60e46c2aefc216db202b41f44542e7b893e08e0721c667b9ccc50411d1ac84ee"
+          },
+          "src/inferrsi/infra/executors/__init__.py": {
+            "bytes": 64,
+            "sha256": "327c3cd5c821c832ff260a41dba0ee1e439cdabefc9edfb9cad1cd38d14bc87a"
+          },
+          "src/inferrsi/infra/executors/local.py": {
+            "bytes": 3654,
+            "sha256": "37bcbca9f3261c8c58a8bf806c9da40720b18d0676c82e8ed24c3901ac9bbe77"
+          },
+          "src/inferrsi/infra/http.py": {
+            "bytes": 634,
+            "sha256": "f65d6c2b55b6a89694caaf6dcdf7d74f417e9ced103eca1b3f218e305c6b6cb2"
+          },
+          "src/inferrsi/infra/inspection.py": {
+            "bytes": 2415,
+            "sha256": "d1c66deb0cced529bca996d2f88b60f51fa3417ff1cae365c650fec808946ecd"
+          },
+          "src/inferrsi/infra/platforms/__init__.py": {
+            "bytes": 1494,
+            "sha256": "9e88e8b9a396ed90c221e1585135c0a1b58da41c1bb73ed9237c53b6684abfed"
+          },
+          "src/inferrsi/infra/platforms/cuda.py": {
+            "bytes": 3309,
+            "sha256": "9e0e4276e1b2984605493b14b82fba3e86f6635204719c318302e624f66f8b6f"
+          },
+          "src/inferrsi/infra/platforms/mlx.py": {
+            "bytes": 1971,
+            "sha256": "f3b81470db28bd6183ccca2e763c5e079686bba6cdfe6c08ebe2cdf64216f2e3"
+          },
+          "src/inferrsi/infra/processes.py": {
+            "bytes": 8151,
+            "sha256": "3b96ebc7a50913e78e86ff3dcba74c8ee8d03d547d9591ca6488f390ac1e6189"
+          },
+          "src/inferrsi/infra/targets/__init__.py": {
+            "bytes": 186,
+            "sha256": "1f3ea75dded7c6cc967754c7b6d4f30ef00da8191a749938aabc3c6bb8f86eb3"
+          },
+          "src/inferrsi/infra/targets/git.py": {
+            "bytes": 2910,
+            "sha256": "6327c4f126086f310b6de309fc7659c9759da28ed76efaeeb94a3d2fba29a0c2"
+          },
+          "src/inferrsi/inspection.py": {
+            "bytes": 692,
+            "sha256": "afa44bf4cdfbb58a318c21f9b292ca142d1bbb22750e35a4cc8244da5350ebb3"
+          },
+          "src/inferrsi/knowledge_operations.py": {
+            "bytes": 27624,
+            "sha256": "c71bb2063ab8d919ca5907ac900a3930ec60730515c614fee7375d9efb77173a"
+          },
+          "src/inferrsi/knowledge_store.py": {
+            "bytes": 13700,
+            "sha256": "5e794e28906ce4687aff7f14a507e267184640fc8a4fac19e0ced279ce6a6cb6"
+          },
+          "src/inferrsi/model/__init__.py": {
+            "bytes": 77,
+            "sha256": "3c1dba4f5159fd9fbc8d4d96e2a83c8441491c1b1dc7b42031a3fc1782ed9d9f"
+          },
+          "src/inferrsi/model/anthropic_agent.py": {
+            "bytes": 24696,
+            "sha256": "b0ef9cd542544292eeb90ab05ef410be9ed6e5d1a40d729890f49562c28c12f0"
+          },
+          "src/inferrsi/model/protocols/__init__.py": {
+            "bytes": 85,
+            "sha256": "5a2f54483b164de04e309dbebb87a700b213682b0f44a50d89a752dd71a4ce47"
+          },
+          "src/inferrsi/model/protocols/image.py": {
+            "bytes": 2047,
+            "sha256": "251a1f8f885da2cd8e9d3baba766c525afe293091606b4e8668c43ba2cb7cb5f"
+          },
+          "src/inferrsi/model/protocols/metrics.py": {
+            "bytes": 610,
+            "sha256": "17db16d6ef15b41fb60d8f09507893accc43f9cb3ad686b8dc4e9839713ebc56"
+          },
+          "src/inferrsi/model/validators/__init__.py": {
+            "bytes": 65,
+            "sha256": "1d4b6ffa968e283219ddcc8a09de9280238a8d65ae2dc8cd2b60889877a8eb0b"
+          },
+          "src/inferrsi/model/validators/flare.py": {
+            "bytes": 19975,
+            "sha256": "3696b0609c772f2222b448833608565a033123d7e92b11fbc545dd356fec590f"
+          },
+          "src/inferrsi/model/validators/image.py": {
+            "bytes": 3204,
+            "sha256": "79470b6b06a22d53ec6c40636a0b3d821a2e6ee2f7bb7aed3dac7241504add1a"
+          },
+          "src/inferrsi/reporting.py": {
+            "bytes": 127,
+            "sha256": "1f771890ab3e69b1aa9246b548a7825596bdf904f6c395dd218208084266853d"
+          },
+          "src/inferrsi/research.py": {
+            "bytes": 5213,
+            "sha256": "9bfe59dd17b5476c18ce50d671711053c62800521c50d6b710e8140a2b4dca04"
+          },
+          "src/inferrsi/runspec.py": {
+            "bytes": 4398,
+            "sha256": "dca0952425caf42095ef0c7e2feb8ef4740cc422ecb61e777f203962d9b87709"
+          },
+          "src/inferrsi/skill_catalog.py": {
+            "bytes": 9938,
+            "sha256": "65f5a1f3f6894009abb98ec2fec30a63cc1bb2f6ce6aeb01f9f7a3dce0439f2d"
+          },
+          "src/inferrsi/summary_chart.py": {
+            "bytes": 40910,
+            "sha256": "d29023d60f6be804c1b01c9eb73765d4824d6c3bb870fe22c99a24544bc1d8bc"
+          }
+        },
+        "sha256": "d6edbaf975cf3bc00902d6aa7cf2bb955dc58b3e37c74f59bb226c5f0d2fa261"
+      },
+      "tracked_source_dirty": false
+    }
+  },
+  "pairs": [
+    {
+      "pair": 0,
+      "order": "AB",
+      "A": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 956367600,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 1141043400,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 31766500,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 1050037700,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 1203064400,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 32086500,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 968658700,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 1091271600,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 31534700,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 950677800,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 1059228900,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 30147800,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 880051400,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 995498500,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 974618900,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 1106810700,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 385004100,
+            "decode_ns": 14500,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1633381200,
+            "decode_ns": 19500,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 1406939000,
+            "decode_ns": 1639300,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 1400129400,
+            "decode_ns": 3201100,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      },
+      "B": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 653961200,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 832134900,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 31837400,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 568411100,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 718840400,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 32799300,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 582848900,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 675246300,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 30796800,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 570669500,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 678774600,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 31429200,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 501214900,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 610733200,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 560394700,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 672993600,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "c71bb2063ab8d919ca5907ac900a3930ec60730515c614fee7375d9efb77173a",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 346454100,
+            "decode_ns": 12500,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1092093900,
+            "decode_ns": 18700,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 1000223300,
+            "decode_ns": 1445100,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 1068648500,
+            "decode_ns": 3234500,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      }
+    },
+    {
+      "pair": 1,
+      "order": "BA",
+      "B": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 645171700,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 751482200,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 30995100,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 552982100,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 665635000,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 31407600,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 579206800,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 667375900,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 31526700,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 563089300,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 676676400,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 32549000,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 498320500,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 634547400,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 564431800,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 685957800,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "c71bb2063ab8d919ca5907ac900a3930ec60730515c614fee7375d9efb77173a",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 351950600,
+            "decode_ns": 12800,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1104294000,
+            "decode_ns": 19700,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 999673300,
+            "decode_ns": 1677100,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 1054962300,
+            "decode_ns": 3211200,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      },
+      "A": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 1022547900,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 1148098900,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 31570400,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 986274000,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 1106520400,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 31999000,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 934886800,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 1054990700,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 32004700,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 938163800,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 1050558500,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 35688100,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 1395703500,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 1210672400,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 996288000,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 1314731100,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 345315300,
+            "decode_ns": 13700,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1592663300,
+            "decode_ns": 18500,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 1368909000,
+            "decode_ns": 1499800,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 1442175500,
+            "decode_ns": 3353500,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      }
+    },
+    {
+      "pair": 2,
+      "order": "AB",
+      "A": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 1019658800,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 1131529500,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 31037100,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 928077300,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 1089250000,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 32073000,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 958851200,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 1864123200,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 35253200,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 973674300,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 1055920000,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 30935500,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 880547800,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 1160809700,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 1077014100,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 1059509200,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 341368500,
+            "decode_ns": 17100,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1834924400,
+            "decode_ns": 20800,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 1500780600,
+            "decode_ns": 1543000,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 1410598700,
+            "decode_ns": 3274400,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      },
+      "B": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 651157900,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 752426800,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 31183400,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 551837200,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 665519000,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 30730100,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 587324300,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 684910600,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 31537600,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 559766600,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 669647100,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 30977800,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 494445200,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 605624100,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 574527900,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 686745900,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "c71bb2063ab8d919ca5907ac900a3930ec60730515c614fee7375d9efb77173a",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 347527300,
+            "decode_ns": 12300,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1107406400,
+            "decode_ns": 32200,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 988403700,
+            "decode_ns": 1492400,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 995649700,
+            "decode_ns": 3460000,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      }
+    },
+    {
+      "pair": 3,
+      "order": "BA",
+      "B": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 695885500,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 823543400,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 35460800,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 586574300,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 669399000,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 30562400,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 565358900,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 689287500,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 33187300,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 582528200,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 669822000,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 30949200,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 499014400,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 618754000,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 580405200,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 688694200,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "c71bb2063ab8d919ca5907ac900a3930ec60730515c614fee7375d9efb77173a",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 339190100,
+            "decode_ns": 12700,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1102467900,
+            "decode_ns": 19200,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 997983600,
+            "decode_ns": 1459800,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 1398879200,
+            "decode_ns": 5293800,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      },
+      "A": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 1080308400,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 1248603400,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 32242000,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 945464400,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 1070388000,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 31939000,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 971184800,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 1192881000,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 38178700,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 1045329900,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 1110398400,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 32560700,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 894147900,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 1000738300,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 940772400,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 1064217600,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 334869700,
+            "decode_ns": 12500,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1650161300,
+            "decode_ns": 19700,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 1419372900,
+            "decode_ns": 1435600,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 1384117900,
+            "decode_ns": 3097700,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      }
+    },
+    {
+      "pair": 4,
+      "order": "AB",
+      "A": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 1044537800,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 1193196600,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 32405100,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 950531500,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 1074591900,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 32427500,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 956832600,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 1085534300,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 31654200,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 948691900,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 1230276100,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 48085400,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 1205961400,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 1155451600,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 1136543500,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 1430489400,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 474929700,
+            "decode_ns": 14200,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 2095528500,
+            "decode_ns": 19700,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 2333465200,
+            "decode_ns": 1698600,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 1466317600,
+            "decode_ns": 3304500,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      },
+      "B": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 707881100,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 1372943600,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 54156900,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 927507200,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 779058100,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 32138400,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 575574300,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 687621400,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 31939900,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 560411400,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 669644800,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 30640600,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 513356000,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 612426800,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 560862100,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 675544000,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "c71bb2063ab8d919ca5907ac900a3930ec60730515c614fee7375d9efb77173a",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 350606500,
+            "decode_ns": 12700,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1075784800,
+            "decode_ns": 18300,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 1024296000,
+            "decode_ns": 1499100,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 1072420300,
+            "decode_ns": 3339400,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      }
+    },
+    {
+      "pair": 5,
+      "order": "BA",
+      "B": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 632486500,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 754616800,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 31022700,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 563909500,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 668911900,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 31045400,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 568893900,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 670136900,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 31484300,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 561848500,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 679501700,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 31339500,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 501706300,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 608601200,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 566199900,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 684335600,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "c71bb2063ab8d919ca5907ac900a3930ec60730515c614fee7375d9efb77173a",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 339003100,
+            "decode_ns": 12300,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1079955500,
+            "decode_ns": 19300,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 977813600,
+            "decode_ns": 1478700,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 1020530400,
+            "decode_ns": 3135300,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      },
+      "A": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 1011321600,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 1142023100,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 31804500,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 926099400,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 1053262900,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 31969100,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 939493000,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 1038275800,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 30918000,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 972812300,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 1053335900,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 30853800,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 873918000,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 998638400,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 948819300,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 1061702300,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 347404900,
+            "decode_ns": 11900,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1578629300,
+            "decode_ns": 23000,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 1448488200,
+            "decode_ns": 1740100,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 2740670100,
+            "decode_ns": 5495100,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      }
+    },
+    {
+      "pair": 6,
+      "order": "AB",
+      "A": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 1254292400,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 1370531600,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 35327500,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 1132210700,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 1586398800,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 49780800,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 2937072700,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 1895383000,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 54632100,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 1933674700,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 1849419500,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 51135000,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 1546027300,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 1850872500,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 1471614900,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 1569750900,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 487308300,
+            "decode_ns": 14700,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 2132036300,
+            "decode_ns": 28500,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 1524694400,
+            "decode_ns": 1684000,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 1810157800,
+            "decode_ns": 3555700,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      },
+      "B": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 845480100,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 1057018800,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 53433400,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 773968600,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 696639700,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 30554500,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 572869200,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 676191200,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 30565500,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 601597800,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 775391600,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 35053000,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 550203400,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 598239000,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 539063600,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 650242300,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "c71bb2063ab8d919ca5907ac900a3930ec60730515c614fee7375d9efb77173a",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 321772900,
+            "decode_ns": 11300,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1028291200,
+            "decode_ns": 18000,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 992453800,
+            "decode_ns": 1539800,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 1027748000,
+            "decode_ns": 3605300,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      }
+    },
+    {
+      "pair": 7,
+      "order": "BA",
+      "B": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 896956200,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 948916000,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 43508500,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 693833200,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 807950400,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 34323300,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 596549800,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 659215400,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 30534400,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 550042300,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 652163000,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 30594900,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 506349100,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 599767600,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 579831700,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 727820400,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "c71bb2063ab8d919ca5907ac900a3930ec60730515c614fee7375d9efb77173a",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 411988700,
+            "decode_ns": 14000,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1368555300,
+            "decode_ns": 24200,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 1156373900,
+            "decode_ns": 1555700,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 1011463800,
+            "decode_ns": 3452300,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      },
+      "A": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 1054030400,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 1102366800,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 30939200,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 958598400,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 1019487600,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 29719900,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 912099900,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 1004235800,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 29776600,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 907738500,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 1011411700,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 29535300,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 854806500,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 1067116400,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 904300600,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 1036013000,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 330484700,
+            "decode_ns": 11600,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1540994000,
+            "decode_ns": 19900,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 1323875500,
+            "decode_ns": 1495700,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 1347399500,
+            "decode_ns": 3205400,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      }
+    },
+    {
+      "pair": 8,
+      "order": "AB",
+      "A": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 1096094600,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 1650199200,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 58812500,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 1219701200,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 1106551000,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 34012600,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 1637209100,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 1211056000,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 31644300,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 918645100,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 1066841400,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 30300400,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 854697300,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 964483100,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 921108500,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 1054366300,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 313877000,
+            "decode_ns": 12000,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1643906700,
+            "decode_ns": 23500,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 1381322600,
+            "decode_ns": 1537100,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 1377925600,
+            "decode_ns": 3474900,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      },
+      "B": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 665337100,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 795635400,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 31084900,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 579973900,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 779100100,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 38361300,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 720613300,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 862125300,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 37828300,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 692738300,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 823657100,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 39719400,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 633370100,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 686550300,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 779192200,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 1055913500,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "c71bb2063ab8d919ca5907ac900a3930ec60730515c614fee7375d9efb77173a",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 766755400,
+            "decode_ns": 15900,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1648931900,
+            "decode_ns": 30100,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 1566623200,
+            "decode_ns": 3402900,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 1897488600,
+            "decode_ns": 3419900,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      }
+    },
+    {
+      "pair": 9,
+      "order": "BA",
+      "B": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 664792200,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 760201500,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 31725800,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 686850400,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 750841200,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 32096300,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 615956300,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 731885400,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 31104400,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 569430200,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 660626400,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 30118900,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 491233400,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 738894700,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 765871800,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 767936100,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "c71bb2063ab8d919ca5907ac900a3930ec60730515c614fee7375d9efb77173a",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 372959600,
+            "decode_ns": 12400,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1078452800,
+            "decode_ns": 22500,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 945596200,
+            "decode_ns": 1555700,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 986643900,
+            "decode_ns": 3219300,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      },
+      "A": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 1019328300,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 1128305500,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 30871500,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 913271800,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 1037670000,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 31182400,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 964224100,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 1029187900,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 29783700,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 920387800,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 1039500800,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 35144200,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 916443600,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 972369100,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 919718500,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 1048572200,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 339751800,
+            "decode_ns": 12400,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1552064300,
+            "decode_ns": 18800,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 1309874800,
+            "decode_ns": 1447700,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 1416693400,
+            "decode_ns": 3227100,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      }
+    },
+    {
+      "pair": 10,
+      "order": "AB",
+      "A": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 1258358400,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 1182460900,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 31248300,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 958013100,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 1114957900,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 31930100,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 1080005800,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 1092786600,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 36721300,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 1133174600,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 1309366500,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 64293200,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 1456999000,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 1401271300,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 1951861100,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 1272683200,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 365535900,
+            "decode_ns": 11500,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1922624100,
+            "decode_ns": 24000,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 1505908700,
+            "decode_ns": 1519000,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 1541049000,
+            "decode_ns": 3370500,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      },
+      "B": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 703733500,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 770570600,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 30729900,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 557425300,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 735598300,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 34844300,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 684734400,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 751023000,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 38908500,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 681962300,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 762060200,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 31637400,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 513908200,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 674799400,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 662553400,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 696908400,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "c71bb2063ab8d919ca5907ac900a3930ec60730515c614fee7375d9efb77173a",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 374534500,
+            "decode_ns": 13000,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1280257500,
+            "decode_ns": 24400,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 1359075800,
+            "decode_ns": 3027500,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 1363905700,
+            "decode_ns": 4186900,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      }
+    },
+    {
+      "pair": 11,
+      "order": "BA",
+      "B": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 600908800,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 702861700,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 29475100,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 516406200,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 625103200,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 29353700,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 527400700,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 635942500,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 28929100,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 531544700,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 629753200,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 29428600,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 473640900,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 570747100,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 534326300,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 631882800,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "c71bb2063ab8d919ca5907ac900a3930ec60730515c614fee7375d9efb77173a",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 373078100,
+            "decode_ns": 13300,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1048925900,
+            "decode_ns": 22500,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 925746200,
+            "decode_ns": 1453300,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 1060198200,
+            "decode_ns": 3352600,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      },
+      "A": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 946860400,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 1057066600,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 29043300,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 868245500,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 976656700,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 29742000,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 925024200,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 1064697700,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 30292600,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 889579600,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 981015400,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 29415700,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 826415200,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 925480100,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 880194200,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 1003514100,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 324910800,
+            "decode_ns": 12400,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1533454400,
+            "decode_ns": 23100,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 1273409600,
+            "decode_ns": 1437600,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 1338380500,
+            "decode_ns": 3051700,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      }
+    },
+    {
+      "pair": 12,
+      "order": "AB",
+      "A": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 950555900,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 1052582500,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 28803000,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 883287800,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 984565000,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 28824100,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 917332900,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 1040264900,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 29685200,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 967991300,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 1012778900,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 28868100,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 837147000,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 940812000,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 884997000,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 986844800,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 314418000,
+            "decode_ns": 11200,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1521476200,
+            "decode_ns": 30000,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 1292517600,
+            "decode_ns": 1388900,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 1291205200,
+            "decode_ns": 3146600,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      },
+      "B": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 606719100,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 704387400,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 29839700,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 526082000,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 623800900,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 29587000,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 545274100,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 642852400,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 29660200,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 531722800,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 630736800,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 29599100,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 466064200,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 579155200,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 538781200,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 639774000,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "c71bb2063ab8d919ca5907ac900a3930ec60730515c614fee7375d9efb77173a",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 329735200,
+            "decode_ns": 12500,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1174857300,
+            "decode_ns": 21700,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 1040870100,
+            "decode_ns": 1485300,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 937058300,
+            "decode_ns": 3094600,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      }
+    },
+    {
+      "pair": 13,
+      "order": "BA",
+      "B": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 678576000,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 846912900,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 35480600,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 579368200,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 631824700,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 29076800,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 531717400,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 655427900,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 28523100,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 526013100,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 629632400,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 29579300,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 474861700,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 576397600,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 543487200,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 641388500,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "c71bb2063ab8d919ca5907ac900a3930ec60730515c614fee7375d9efb77173a",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 311835900,
+            "decode_ns": 11600,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1037362500,
+            "decode_ns": 34200,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 933928200,
+            "decode_ns": 1411300,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 934717800,
+            "decode_ns": 3097200,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      },
+      "A": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 979972900,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 1117831800,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 30308700,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 900416500,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 974947200,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 28758200,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 885506600,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 1027049800,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 31518900,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 906534000,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 988224300,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 29801300,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 834446100,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 1627686400,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 1080936600,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 988303300,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 309811600,
+            "decode_ns": 12300,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1485501500,
+            "decode_ns": 22500,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 1344235200,
+            "decode_ns": 1444500,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 1296935800,
+            "decode_ns": 3188400,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      }
+    },
+    {
+      "pair": 14,
+      "order": "AB",
+      "A": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 974315400,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 1064671300,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 29782500,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 896813400,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 1001259100,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 29085400,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 901209600,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 1006739700,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 29771400,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 889454000,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 999125000,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 29888800,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 834747500,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 924292700,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 885637200,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 1065348000,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 344128600,
+            "decode_ns": 13700,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1536655700,
+            "decode_ns": 25800,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 1295772200,
+            "decode_ns": 1713100,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 1314117200,
+            "decode_ns": 3156500,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      },
+      "B": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 604931600,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 752898200,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 31382100,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 539635400,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 624022200,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 29507100,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 530799800,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 660216400,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 29258100,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 540331500,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 709112600,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 30741100,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 490558900,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 615002800,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 809745800,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 876725700,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "c71bb2063ab8d919ca5907ac900a3930ec60730515c614fee7375d9efb77173a",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 344250100,
+            "decode_ns": 12500,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1053073800,
+            "decode_ns": 22300,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 930016600,
+            "decode_ns": 1480600,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 954074000,
+            "decode_ns": 3252000,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      }
+    },
+    {
+      "pair": 15,
+      "order": "BA",
+      "B": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 661353000,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 1099425000,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 37261700,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 594856300,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 659687300,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 31207000,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 544464900,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 729165200,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 46337900,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 804370100,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 836220300,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 39896700,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 554316500,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 638138900,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 693186100,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 700716400,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "c71bb2063ab8d919ca5907ac900a3930ec60730515c614fee7375d9efb77173a",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 324561800,
+            "decode_ns": 12800,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1173330700,
+            "decode_ns": 19400,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 947995700,
+            "decode_ns": 1554400,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 1014041100,
+            "decode_ns": 3799200,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      },
+      "A": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 1316113800,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 1175418500,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 30459300,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 914384200,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 1179434800,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 38643800,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 1046032100,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 1091456500,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 31893300,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 1350648300,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 1334450200,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 35917000,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 925053900,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 1017067200,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 948405600,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 1227121500,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 488510300,
+            "decode_ns": 15800,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1869774200,
+            "decode_ns": 19000,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 1686501100,
+            "decode_ns": 2320800,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 2639715300,
+            "decode_ns": 8382300,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      }
+    },
+    {
+      "pair": 16,
+      "order": "AB",
+      "A": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 1174417400,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 1125634500,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 35592100,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 1095833600,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 1016513900,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 31823900,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 1035014600,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 1190594200,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 34721600,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 1044732300,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 1139627400,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 31834400,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 885320600,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 940078000,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 900502500,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 1059190000,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 320069600,
+            "decode_ns": 10800,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1535422900,
+            "decode_ns": 21700,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 1456109300,
+            "decode_ns": 1661900,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 1508172400,
+            "decode_ns": 3386000,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      },
+      "B": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 734361300,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 807084600,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 36407700,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 620199700,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 739885900,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 37746400,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 648441400,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 740842200,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 40185200,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 643295400,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 743033500,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 38585500,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 561250500,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 696165200,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 607461600,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 709383600,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "c71bb2063ab8d919ca5907ac900a3930ec60730515c614fee7375d9efb77173a",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 363430500,
+            "decode_ns": 13200,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1192218100,
+            "decode_ns": 20100,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 1120484200,
+            "decode_ns": 1632200,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 1038853200,
+            "decode_ns": 3425800,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      }
+    },
+    {
+      "pair": 17,
+      "order": "BA",
+      "B": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 643306600,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 780831300,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 32695700,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 600140800,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 730993700,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 32913000,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 606897000,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 706849000,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 35839500,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 627618600,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 789035000,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 36413000,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 558454900,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 613298400,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 554076200,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 647844600,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "c71bb2063ab8d919ca5907ac900a3930ec60730515c614fee7375d9efb77173a",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 354393100,
+            "decode_ns": 13200,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1140824200,
+            "decode_ns": 24800,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 1057875200,
+            "decode_ns": 1595500,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 950722500,
+            "decode_ns": 3023100,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      },
+      "A": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 1019861000,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 1230912100,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 31597700,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 931576800,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 1012387700,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 28770400,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 874810000,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 1081733300,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 30025900,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 915561600,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 1070079700,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 33804000,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 934368300,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 939652000,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 914147500,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 1036657700,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 325473200,
+            "decode_ns": 11900,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 2153863600,
+            "decode_ns": 24500,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 1489875800,
+            "decode_ns": 1783000,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 2408813400,
+            "decode_ns": 5214400,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      }
+    },
+    {
+      "pair": 18,
+      "order": "AB",
+      "A": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 1339223800,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 1415762200,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 53759200,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 1361751300,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 1229528800,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 34040900,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 1348145400,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 2024106600,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 56353700,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 1969647800,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 1762318900,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 80335200,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 1528333400,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 1647646900,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 1440488300,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 1593444600,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 418868200,
+            "decode_ns": 13500,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1900660200,
+            "decode_ns": 34800,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 1647850400,
+            "decode_ns": 1526300,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 1923232200,
+            "decode_ns": 4214100,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      },
+      "B": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 1164681300,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 1134827500,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 50117800,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 877200300,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 996335400,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 44741600,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 826867800,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 1061174900,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 54741600,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 846379800,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 1006765600,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 43421800,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 785029100,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 782934000,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 745440500,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 836730000,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "c71bb2063ab8d919ca5907ac900a3930ec60730515c614fee7375d9efb77173a",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 434632200,
+            "decode_ns": 13700,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1237515400,
+            "decode_ns": 18900,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 993099700,
+            "decode_ns": 1534100,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 981808000,
+            "decode_ns": 3171200,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      }
+    },
+    {
+      "pair": 19,
+      "order": "BA",
+      "B": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 625697300,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 726653800,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 29470200,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 519062300,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 626314500,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 29766600,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 544251700,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 638242200,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 29575300,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 534167300,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 626757800,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 28472200,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 511888000,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 651793700,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 607049500,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 661046000,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "c71bb2063ab8d919ca5907ac900a3930ec60730515c614fee7375d9efb77173a",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 312376100,
+            "decode_ns": 11500,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1042396300,
+            "decode_ns": 26500,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 1026287600,
+            "decode_ns": 1634100,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 968532600,
+            "decode_ns": 3178000,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      },
+      "A": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 968824800,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 1084113700,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 29764100,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 905343300,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 1003497700,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 30114700,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 888461400,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 995548500,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 29162900,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 876563700,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 1009745300,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 30817600,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 844304700,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 929171100,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 888434300,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 1010348000,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 316923900,
+            "decode_ns": 10700,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 1486903800,
+            "decode_ns": 18300,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 1282747100,
+            "decode_ns": 1455600,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 1297898800,
+            "decode_ns": 3228500,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      }
+    }
+  ],
+  "elapsed_ns": 836162085800,
+  "summary": {
+    "primary": {
+      "paired_median_reduction": 0.40006826681186614,
+      "confidence_95": [
+        0.3837892274877763,
+        0.41059207129958375
+      ],
+      "paired_reductions": [
+        0.4007778491130225,
+        0.40642607415007126,
+        0.4064291001593511,
+        0.42023606373908406,
+        0.32710519192428966,
+        0.40762419055158194,
+        0.5685129874771919,
+        0.310213196858093,
+        0.41355995204758556,
+        0.32041790673188464,
+        0.382473882382114,
+        0.4164316199492143,
+        0.42341487351923734,
+        0.38510457259343855,
+        0.39935868451070977,
+        0.4681914863937914,
+        0.39654425028446916,
+        0.3316604754684327,
+        0.363968338719332,
+        0.39880913457301126
+      ],
+      "bootstrap": {
+        "method": "percentile bootstrap of paired reductions; each draw samples 20 reductions with replacement, then computes their median",
+        "iterations": 10000,
+        "seed": 19020261027,
+        "confidence": 0.95,
+        "quantiles": [
+          0.025,
+          0.975
+        ],
+        "quantile_method": "linear interpolation at (n-1)*p"
+      }
+    },
+    "performance_gate_passed": true,
+    "measurement_structure_errors": [],
+    "output_parity_assessed": true,
+    "output_parity_failures": [],
+    "latency_ns": {
+      "api/cold_context_material/dev09/A": {
+        "n": 20,
+        "p50": 1060605750.0,
+        "p95": 1570935585.0
+      },
+      "api/cold_context_material/dev09/B": {
+        "n": 20,
+        "p50": 686351850.0,
+        "p95": 885685090.0000001
+      },
+      "api/cold_default_diagnostic/dev05/A": {
+        "n": 20,
+        "p50": 999688350.0,
+        "p95": 1657808180.0000002
+      },
+      "api/cold_default_diagnostic/dev05/B": {
+        "n": 20,
+        "p50": 614150600.0,
+        "p95": 741096665.0
+      },
+      "api/cold_relevance/dev01/A": {
+        "n": 20,
+        "p50": 1141533250.0,
+        "p95": 1427484050.0000002
+      },
+      "api/cold_relevance/dev01/B": {
+        "n": 20,
+        "p50": 788233350.0,
+        "p95": 1146733305.0000002
+      },
+      "api/cold_relevance/dev05/A": {
+        "n": 20,
+        "p50": 1061825450.0,
+        "p95": 1247372300.0000002
+      },
+      "api/cold_relevance/dev05/B": {
+        "n": 20,
+        "p50": 683019350.0,
+        "p95": 817369650.0000001
+      },
+      "api/cold_relevance/dev09/A": {
+        "n": 20,
+        "p50": 1083633800.0,
+        "p95": 1901819180.0
+      },
+      "api/cold_relevance/dev09/B": {
+        "n": 20,
+        "p50": 680550900.0,
+        "p95": 872077780.0000001
+      },
+      "api/cold_relevance/dev13/A": {
+        "n": 20,
+        "p50": 1057574450.0,
+        "p95": 1766673930.0
+      },
+      "api/cold_relevance/dev13/B": {
+        "n": 20,
+        "p50": 677725500.0,
+        "p95": 844747565.0000001
+      },
+      "api/hot_context_material/dev09/A": {
+        "n": 20,
+        "p50": 944589000.0,
+        "p95": 1495627210.0000002
+      },
+      "api/hot_context_material/dev09/B": {
+        "n": 20,
+        "p50": 577179800.0,
+        "p95": 780719880.0
+      },
+      "api/hot_default_diagnostic/dev05/A": {
+        "n": 20,
+        "p50": 882934200.0,
+        "p95": 1529218095.0
+      },
+      "api/hot_default_diagnostic/dev05/B": {
+        "n": 20,
+        "p50": 504027700.0,
+        "p95": 640953050.0000001
+      },
+      "api/hot_relevance/dev01/A": {
+        "n": 20,
+        "p50": 1021204450.0,
+        "p95": 1317269300.0
+      },
+      "api/hot_relevance/dev01/B": {
+        "n": 20,
+        "p50": 663072600.0,
+        "p95": 910342455.0000002
+      },
+      "api/hot_relevance/dev05/A": {
+        "n": 20,
+        "p50": 938520600.0,
+        "p95": 1226803705.0
+      },
+      "api/hot_relevance/dev05/B": {
+        "n": 20,
+        "p50": 579671050.0,
+        "p95": 879715645.0
+      },
+      "api/hot_relevance/dev09/A": {
+        "n": 20,
+        "p50": 957841900.0,
+        "p95": 1702202280.000001
+      },
+      "api/hot_relevance/dev09/B": {
+        "n": 20,
+        "p50": 577390550.0,
+        "p95": 725926025.0000001
+      },
+      "api/hot_relevance/dev13/A": {
+        "n": 20,
+        "p50": 949684850.0,
+        "p95": 1935473355.0
+      },
+      "api/hot_relevance/dev13/B": {
+        "n": 20,
+        "p50": 566259750.0,
+        "p95": 806470585.0
+      },
+      "api/search/dev01/A": {
+        "n": 20,
+        "p50": 31409350.0,
+        "p95": 54011865.0
+      },
+      "api/search/dev01/B": {
+        "n": 20,
+        "p50": 31781600.0,
+        "p95": 53469575.0
+      },
+      "api/search/dev05/A": {
+        "n": 20,
+        "p50": 31934550.0,
+        "p95": 39200650.00000001
+      },
+      "api/search/dev05/B": {
+        "n": 20,
+        "p50": 31307300.0,
+        "p95": 38680315.00000001
+      },
+      "api/search/dev09/A": {
+        "n": 20,
+        "p50": 31589500.0,
+        "p95": 54718180.0
+      },
+      "api/search/dev09/B": {
+        "n": 20,
+        "p50": 31505500.0,
+        "p95": 46758085.00000001
+      },
+      "api/search/dev13/A": {
+        "n": 20,
+        "p50": 31384950.0,
+        "p95": 65095300.000000015
+      },
+      "api/search/dev13/B": {
+        "n": 20,
+        "p50": 31158650.0,
+        "p95": 40072955.0
+      },
+      "cli/context_material/dev09/decode_ns/A": {
+        "n": 20,
+        "p50": 3289450.0,
+        "p95": 5639460.000000002
+      },
+      "cli/context_material/dev09/decode_ns/B": {
+        "n": 20,
+        "p50": 3295700.0,
+        "p95": 4242245.000000001
+      },
+      "cli/context_material/dev09/process_ns/A": {
+        "n": 20,
+        "p50": 1413646050.0,
+        "p95": 2644763040.0
+      },
+      "cli/context_material/dev09/process_ns/B": {
+        "n": 20,
+        "p50": 1017285750.0,
+        "p95": 1423809670.0000002
+      },
+      "cli/default_diagnostic/dev05/decode_ns/A": {
+        "n": 20,
+        "p50": 22100.0,
+        "p95": 30240.000000000004
+      },
+      "cli/default_diagnostic/dev05/decode_ns/B": {
+        "n": 20,
+        "p50": 22000.0,
+        "p95": 32300.0
+      },
+      "cli/default_diagnostic/dev05/process_ns/A": {
+        "n": 20,
+        "p50": 1613022250.0,
+        "p95": 2133127665.0
+      },
+      "cli/default_diagnostic/dev05/process_ns/B": {
+        "n": 20,
+        "p50": 1103380950.0,
+        "p95": 1382574130.0000002
+      },
+      "cli/relevance/dev05/decode_ns/A": {
+        "n": 20,
+        "p50": 1531700.0,
+        "p95": 1809890.0000000005
+      },
+      "cli/relevance/dev05/decode_ns/B": {
+        "n": 20,
+        "p50": 1536950.0,
+        "p95": 3046270.0000000005
+      },
+      "cli/relevance/dev05/process_ns/A": {
+        "n": 20,
+        "p50": 1413155950.0,
+        "p95": 1718849305.0000005
+      },
+      "cli/relevance/dev05/process_ns/B": {
+        "n": 20,
+        "p50": 998828450.0,
+        "p95": 1369453170.0000002
+      },
+      "cli/search/dev05/decode_ns/A": {
+        "n": 20,
+        "p50": 12400.0,
+        "p95": 15865.000000000002
+      },
+      "cli/search/dev05/decode_ns/B": {
+        "n": 20,
+        "p50": 12700.0,
+        "p95": 14095.000000000002
+      },
+      "cli/search/dev05/process_ns/A": {
+        "n": 20,
+        "p50": 340560150.0,
+        "p95": 487368400.0
+      },
+      "cli/search/dev05/process_ns/B": {
+        "n": 20,
+        "p50": 349066900.0,
+        "p95": 451238360.00000024
+      }
+    }
+  },
+  "source_and_corpus_unchanged_at_end": {
+    "A": true,
+    "B": true
+  },
+  "completed_at_utc": "2026-10-27T11:44:34.072678+00:00"
+}
diff --git a/evidence/knowledge-retrieval-20261027/results/candidate-identity.json b/evidence/knowledge-retrieval-20261027/results/candidate-identity.json
new file mode 100644
index 00000000..9960caad
--- /dev/null
+++ b/evidence/knowledge-retrieval-20261027/results/candidate-identity.json
@@ -0,0 +1,2100 @@
+{
+  "corpus": {
+    "bytes": 2883003,
+    "file_count": 458,
+    "files": {
+      "src/inferrsi/data/knowledge_packs/README.md": {
+        "bytes": 14172,
+        "sha256": "b877e67d9cccf0dcbe57419eeb3ff9374e5bd03ff8eec63adef1025c446140e9"
+      },
+      "src/inferrsi/data/knowledge_packs/dependency/diffusers/cuda/diffusion/group-offload-transfer-overlap.json": {
+        "bytes": 3277,
+        "sha256": "9bec3d912192efa2f2c853d3fb70bdae21a75875630a02294fcd084b18c30414"
+      },
+      "src/inferrsi/data/knowledge_packs/dependency/flashattention/cuda/attention/generation-specific-dispatch.json": {
+        "bytes": 3085,
+        "sha256": "90290d7bc80b1626863e0520c91bb196d93ebfe08190a450e9de59c6505ae8ad"
+      },
+      "src/inferrsi/data/knowledge_packs/dependency/flashinfer/cuda/llm/paged-decode-dispatch.json": {
+        "bytes": 3745,
+        "sha256": "0f2986494fcc9b751df71e598fd61caa68e28bc4b0ac401e4cc763bc3f75283b"
+      },
+      "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/cross-stream-dependency.json": {
+        "bytes": 3809,
+        "sha256": "b829315e55adf8e11f0bd581d5f45dd130d7eb3f81ce7de532f63c7b9c27cbb5"
+      },
+      "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/evaluation-completion.json": {
+        "bytes": 3841,
+        "sha256": "19d2e506ac4f50895776a83a062059aa7a052175c8e4b21dd0d46e1a5a6716a1"
+      },
+      "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/post-detachment-buffer-lifetime.json": {
+        "bytes": 4663,
+        "sha256": "18d78c0e542dacb42ca33e5264d92aef80fa06d163fd16d4feafd5ed0827ab99"
+      },
+      "src/inferrsi/data/knowledge_packs/dependency/pytorch/cuda/compiler/cudagraph-shapes-and-lifetimes.json": {
+        "bytes": 3623,
+        "sha256": "7fef4122ff794967e61f55d02a4a6b558d7e498d9513607f5ca86db6386928a8"
+      },
+      "src/inferrsi/data/knowledge_packs/dependency/transformers/common/llm/kv-cache-strategy-contracts.json": {
+        "bytes": 3072,
+        "sha256": "a2df26052112a9f3fc6242b2bd636ab13991442db57e56646168d6c52032d38d"
+      },
+      "src/inferrsi/data/knowledge_packs/dependency/triton/common/kernel/bandwidth-fusion-resource-limits.json": {
+        "bytes": 2941,
+        "sha256": "6e9673d1a600a6ad34820e4325f2a6417f44f0f96d8ed0bcf21466d998f19ee4"
+      },
+      "src/inferrsi/data/knowledge_packs/peer/cache_dit/common/diffusion/residual-cache-calibration.json": {
+        "bytes": 3841,
+        "sha256": "bb9509ad0225a91b52bc6787477581d05f0ecd5e44fa59f08f718eea36a2ac45"
+      },
+      "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/bounded-windowed-replay.json": {
+        "bytes": 10959,
+        "sha256": "43613225e9ccdbd34c0f7e3e0767887b9bbfd0f38b2c21684677d99e31a432f0"
+      },
+      "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/speculative-verification-budget.json": {
+        "bytes": 10177,
+        "sha256": "a729e1fc00be44e1480a5dc65bbcdd58df8599681df74dbc651a5a49f27e9028"
+      },
+      "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/zero-overhead-deferred-scheduling.json": {
+        "bytes": 8191,
+        "sha256": "ed5426af145f066f6183bd316c4073c46f324f5ee83aa9bd616387c0d0e7149c"
+      },
+      "src/inferrsi/data/knowledge_packs/peer/lmdeploy/cuda/llm/kv-quantization-capacity-cost.json": {
+        "bytes": 3366,
+        "sha256": "873997b38981ccc03ff73e6fb84e96022f293a019843efd8c4aa50f27878ac9a"
+      },
+      "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/codec-constants-and-state.json": {
+        "bytes": 4250,
+        "sha256": "1cecd7905fef5af3ef8952a4d803900a7df91d662ebae34db47cc0dbf352d4e7"
+      },
+      "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/fixed-customvoice-prompt.json": {
+        "bytes": 3660,
+        "sha256": "ea5e8fdf54cb731b255b73aa0cf47ef53fac103baa2bd5246be45f95d54cb593"
+      },
+      "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/graph-batching-state-ownership.json": {
+        "bytes": 3950,
+        "sha256": "130741a9f7ac9d366784d6c4dd42860262235caff0c277de1e8a8f8ec933b37f"
+      },
+      "src/inferrsi/data/knowledge_packs/peer/sglang/cuda/llm/hierarchical-prefix-cache.json": {
+        "bytes": 3641,
+        "sha256": "b50ca58df28d4343545d07adf0da23f8f181297b30aba7870e5690630e7d7315"
+      },
+      "src/inferrsi/data/knowledge_packs/peer/tensorrt_llm/cuda/llm/prefill-decode-disaggregation.json": {
+        "bytes": 2967,
+        "sha256": "349856dd98628f94987e014889cd1b15d0e3ae6ab6c576cdde917f28fbafb102"
+      },
+      "src/inferrsi/data/knowledge_packs/peer/xdit/cuda/diffusion/parallelism-freshness-and-topology.json": {
+        "bytes": 3236,
+        "sha256": "3674b5916cb7b384951bf3e449b6796b161878daabcc1d7e7f6d6f8ff861cc17"
+      },
+      "src/inferrsi/data/repo_packs/afd_plugin/pack.json": {
+        "bytes": 502,
+        "sha256": "00042abeabdc9c0e24f1a4edaba47e233df398aef0ad4e181fc70191d383da5f"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-aclgraph-capture-replay.json": {
+        "bytes": 4311,
+        "sha256": "327414443ea9537ae320aa405912c338351bc06cefdc0b99352e1eea36f61dc3"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-device-health.json": {
+        "bytes": 4115,
+        "sha256": "633fa5fdcea1e1e9edda8a218d6a3fde0d12637848348e1e2fe7d1b0bd744a16"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-distributed-parallel.json": {
+        "bytes": 5841,
+        "sha256": "709ca0d061e4d6917fe8fa049fe93165e71dfecf96c7bf264ced25346381f603"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-graph-mode.json": {
+        "bytes": 4135,
+        "sha256": "bb82ab3d628b91be957e557545810ef5c20790be7d43296096f20a21f45ec391"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-model-adaptation.json": {
+        "bytes": 4137,
+        "sha256": "7ea1ab0f0044afff935a91071b36ac09bc46b5ef044fe2a68ab0608256ddd853"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-profiling.json": {
+        "bytes": 4186,
+        "sha256": "0d27058fd8d0ef8b524b7a60d1cb76950622746c94b5a03915e64e6edb4ab3d4"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-tuning.json": {
+        "bytes": 4439,
+        "sha256": "fcb49d71f59d8b39c1eaa03d24e95e2cbfe5f9d7f6e0f22df03ee0d60f6a568a"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-optimization-casebook.json": {
+        "bytes": 4371,
+        "sha256": "05783630bfaf3c7ef99b848865f4bbab50d0b52e80ad04fcd028c624d7638e9b"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-performance-analysis.json": {
+        "bytes": 7200,
+        "sha256": "b6eea897c12d0c6ac9d172089535c82c25eb71c09d7a4972e64812cc4682b5c9"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascendc-tiling-design.json": {
+        "bytes": 4505,
+        "sha256": "e69b6e89dd342ba55ae54b5fac14aa1b91ee2a361da40cb669c287ce519f4517"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/cann-attribution-overlap.json": {
+        "bytes": 4915,
+        "sha256": "8ac01b60f5cd37fd594d6cb5119e4e6525199d017296fc8779dc68a576f5525c"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/torchnpu-device-memory-lifecycle.json": {
+        "bytes": 4349,
+        "sha256": "7a79d0eb2c50381d40086e6db72d4bdd2cec5d2261bfb00cf73a60701a94da2a"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/torchnpu-pinned-host-memory-lifecycle.json": {
+        "bytes": 4431,
+        "sha256": "8bb34dcfbebb48667a94434b036a1a408b970af9890b047a3de325f364ed5f7c"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/assess-serving-preemption.json": {
+        "bytes": 4198,
+        "sha256": "e3d2a1a72d0d980f31fc85deee27db26bc0daa0c709e8ae99e42deb814a4a7f4"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/assess-split-backward-schedule.json": {
+        "bytes": 3915,
+        "sha256": "a8561794786eb0a897d831f00c92b4fdd3b0492316c1b1d8ad13c1f8f75ac815"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-h3-runtime-residency.json": {
+        "bytes": 12169,
+        "sha256": "602dd625874bd4a89861d5bb25881cc63e4b5a7fdedc2867089f0952943f6eff"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-llama-runtime-state.json": {
+        "bytes": 5816,
+        "sha256": "64f6ab8a95172378350323cead08bc6c201c624b711e6242938a6a21a8e83ad7"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-sd-cpp-residency.json": {
+        "bytes": 8349,
+        "sha256": "95ed97b717255be3b817e5c69debd6d41197559dbf21771507d5da79134a72e1"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-sglang-diffusion-lifetimes.json": {
+        "bytes": 7276,
+        "sha256": "5e5889223878c6cf42a49c6f09c60e9d9add88cb7e7d0df050c4e75d88a05aa6"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-vllm-kv-lifetimes.json": {
+        "bytes": 4275,
+        "sha256": "00424005a5eeb8b15f239edd2cb2c71047787e3ecf55924860cb7a78d137b586"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/bound-windowed-attention-replay.json": {
+        "bytes": 7877,
+        "sha256": "98b2291147a796ef59cf20a075ff662998ee6a529f85b49e9c6fb62fe41fad9f"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/budget-speculative-verification.json": {
+        "bytes": 7490,
+        "sha256": "66448f59460c9dffac46e74160388fab650cf8bbbd5bdc95c1465b70daf0c2dc"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/calibrate-serving-simulation.json": {
+        "bytes": 3992,
+        "sha256": "055b70e0b86f65156775b519e4b8cbf06a956a8993b06a66175167cedb78f2cb"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/cgroup-resource-controls.json": {
+        "bytes": 4801,
+        "sha256": "ad406b6dd5d9415bd28d448e64269a388b50e11f9eb7f5a9adb6b0a6246fd303"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/checkpoint-recovery-consistency.json": {
+        "bytes": 4740,
+        "sha256": "ccb85314a831bde58f8cc515580d001b174c11dca0bbcbc6221303acf629046d"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/choose-parallel-training-plan.json": {
+        "bytes": 3901,
+        "sha256": "40afcb41c59988fc1af8fdd00d2340800c54e96847dd0de16ecddf7ac221d1b4"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/cite-check-downstream-pr.json": {
+        "bytes": 2533,
+        "sha256": "dabff14684d5ae9a2b05a9743ff5b76d00a90fdb29319588aba67f5087aad706"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/claim-contract.json": {
+        "bytes": 2469,
+        "sha256": "78f15f6533434fc29adf1df651f7a15eb8ba81733f8b02c710d90102541d707b"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/comfy-ui.json": {
+        "bytes": 6919,
+        "sha256": "1496807d958df055f720ca93e6dcf7f7e8f506f3f12c94c9a2166f789b00058d"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/competitor-scan.json": {
+        "bytes": 2449,
+        "sha256": "c2e97e49e5185710cdfaa08ac1c2b1c19b958e68ebfca885d5aee0ed8c9f8c8b"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dcp-async-save-performance.json": {
+        "bytes": 4538,
+        "sha256": "359a5acf8304a78fc074688676f9a0be62f56f463d0960a544e409f49843bdaf"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dense-performance-evidence.json": {
+        "bytes": 4227,
+        "sha256": "d87f93a9dc9ddc583e3929baaac85f7a425666dd6dd71665933f18344ccec834"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/diagnose-vllm-admission-pressure.json": {
+        "bytes": 9789,
+        "sha256": "a1de40def775b2e840c27330b67e6f37c3c244db1731fab2bd3e8a787143210e"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/diffusers-invocation-state.json": {
+        "bytes": 6109,
+        "sha256": "914d2ffe9776ba931bed780a2a270c1525e3c6cbaf0735b8531064d3180f3233"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/draft-acceptance-issue.json": {
+        "bytes": 2549,
+        "sha256": "6aaa63dbfe6cda982ccd3a4e368527b80fd2873672c000c66d06d28e40a2bf14"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/evaluate-elastic-serving-placement.json": {
+        "bytes": 4334,
+        "sha256": "fab1fa814e789ec2d86587fcf7de7c38174464b9e9ba0cf53ef3fb470053e7c4"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/generation-cache-lifecycle.json": {
+        "bytes": 5972,
+        "sha256": "e5679cdecb722bbd17280913913da8473052ad09e9e7a13b3a9b95463a139ef6"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/host-memory-management.json": {
+        "bytes": 6530,
+        "sha256": "a36aea21169281866b2468860d3ffeb912e61edd571308b02f82939f17f694a8"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/host-overhead-optimization.json": {
+        "bytes": 4243,
+        "sha256": "05fd7f054966b96892980d585352bc1c3fcd47f439cb34268fdf07feee58a998"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/inference-benchmark-design.json": {
+        "bytes": 10735,
+        "sha256": "091bb086c6f0123dbb5db78968c13f846cad135e8969982efe557644b7c7613b"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/inference-optimization-casebook.json": {
+        "bytes": 6531,
+        "sha256": "c512025d7a59934c91665e24752d9393ca9b6a9dfb624c72789294defae93c0e"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/kernel-writing-guide.json": {
+        "bytes": 4545,
+        "sha256": "f1f0ca1d11d3550c9b8c5fdbed8f6cd09b877ac14341f973d58dca02b5d8995d"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/kv-tier-management.json": {
+        "bytes": 4609,
+        "sha256": "6141f61fb6b0d147b704ab7a2627e9babdb53a90e15b659152a3d565967afddf"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/lazy-evaluation-stream-lifetimes.json": {
+        "bytes": 5934,
+        "sha256": "34f47009a8a33f8907bd78f110b95fcd62409bef1b38e1055acb46b2eab285ff"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/npu-skill-factory.json": {
+        "bytes": 11868,
+        "sha256": "24c1d3bff288e86b99ea674e3252ec6f54d9ba944274aef6a7666197afd6788a"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/optimization-search-policy.json": {
+        "bytes": 7101,
+        "sha256": "d7bf71d5491765f62eec02fe5013dc1a5ef56e1516892da48593c6165a86366d"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/os-process-placement.json": {
+        "bytes": 6304,
+        "sha256": "e6e8fe5d7d66f3e98f05762d94441726dfa757ce7fbffd2b35402b3f327f8fb3"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/performance-ceiling-analysis.json": {
+        "bytes": 13695,
+        "sha256": "47c034d96026e1c2a02848d9c9c3caf4024a5033664bfd089ad30d9f23a80deb"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/plan-activation-residency.json": {
+        "bytes": 3871,
+        "sha256": "436e364020cc854fd52a10ee76e4593c7519bb0546c8aa5238fcfd89158e8ab0"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/repository-evolution-analysis.json": {
+        "bytes": 5865,
+        "sha256": "e0af402c3320ebd9bd36c774cb52c0708721a86ca446c05f6e94306888838e30"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/rsi-evidence.json": {
+        "bytes": 2475,
+        "sha256": "c8b4bb5a8c88009ac5b1526e82639715381c940d3c3eb295f0333cae6c2d3ca9"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/skill-evolution-playbook.json": {
+        "bytes": 4091,
+        "sha256": "0e063c2e7686de0c27c061e1aac0eb2d8c4fa83246e32ddeeb0e28126c176425"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/storage-io-measurement.json": {
+        "bytes": 4783,
+        "sha256": "80e589e3a0b67d3a045e03feeadc45e720ceb277e6124c917eec04863eb67419"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/training-input-pipeline.json": {
+        "bytes": 4296,
+        "sha256": "ae696bea8d2b7bfdaa29b6e5bc0bd861abcc0e42e7d9bec6ae20c6e7d6f27669"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/training-resource-budget.json": {
+        "bytes": 6944,
+        "sha256": "ea5e5e864e3aa34bd00be67c8e741877ae01480f039652073b65b71dfcd54c97"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/weight-loading-optimization.json": {
+        "bytes": 8139,
+        "sha256": "7f2d1fbc6b1cfd80c3d6ed8df107d4ca9b97655626ee078c345445a303b1b367"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/audit-afd-p2p-exchanges.json": {
+        "bytes": 6640,
+        "sha256": "9ac45aa784b430ad1ca9ffddecce8dbae9740cfe9c98c10b330e01d350015a38"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/diagnose-attention-bottlenecks.json": {
+        "bytes": 7602,
+        "sha256": "e525d75578de1cc164e6553d01efbd0a014e74646200580f59b1161c030413ad"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/host-overhead-profiling.json": {
+        "bytes": 3965,
+        "sha256": "e9e6ae8c5b9986e83cade3bd029baae112859e01aa6c96f741a264f3bee44adc"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/reason-hopper-kernels.json": {
+        "bytes": 15086,
+        "sha256": "758a649e651d8f9635ac2df45d18043e156c8d54b8f8dbc95f8e68ebcb866a28"
+      },
+      "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/parallelism-and-extension.json": {
+        "bytes": 3399,
+        "sha256": "1253f5f01bcf1498703c7f34960f0c50580df5ba8f140af557a903e296ad81d3"
+      },
+      "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/runtime-radix-hicache.json": {
+        "bytes": 3317,
+        "sha256": "5a0b2aed51649af214cdc1034f57650b7ce195c934a17452ee67a35131f80cdd"
+      },
+      "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/tuning-and-validation.json": {
+        "bytes": 3385,
+        "sha256": "ef4d4f7182da4433d59d72dda0d21b4aeafadd73260be11ab8d4cb36c20fafc6"
+      },
+      "src/inferrsi/data/repo_packs/sglang/pack.json": {
+        "bytes": 620,
+        "sha256": "bce2e0c42796a9ef499d459c37b56d1d56f4b19344841359c534fca29c0359ce"
+      },
+      "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/audio/serving-and-streaming-quality.json": {
+        "bytes": 3440,
+        "sha256": "97fc1d5a11a39a0634dae9e9602a4ec47a4d2348e121076c2d5c836cf0692114"
+      },
+      "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/omni/extension-lifecycle-validation.json": {
+        "bytes": 3482,
+        "sha256": "7da8dc5d2655099de2bcd5f0fa922d9e1bcb0bfd2f571f026a004a7f402a75b6"
+      },
+      "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/omni/pipeline-state-and-transport.json": {
+        "bytes": 3404,
+        "sha256": "0c57692772019504264e39bf0444a69e20cff86f6e112b072c53cf7a4299cc82"
+      },
+      "src/inferrsi/data/repo_packs/sglang_omni/pack.json": {
+        "bytes": 630,
+        "sha256": "8e70336bc2aac0c74732a026337813652b5d5c6e1fc60f5571e9442345077b96"
+      },
+      "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/architecture-search-space.json": {
+        "bytes": 3294,
+        "sha256": "2c5ad72d5bab4f7082fc3be1532c347444cde38f5a84639fe16b22134b68685e"
+      },
+      "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/controlled-execution-tuning.json": {
+        "bytes": 3255,
+        "sha256": "6a2bf611a432213919e1c237687507ceb0ab753cccad0e934dc77f86604805e9"
+      },
+      "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/engine-validation-contract.json": {
+        "bytes": 3295,
+        "sha256": "ef89775bc82965fe87a832c7ec8a5cf55cb9eba4ecfc6279caea9ce3b0410c41"
+      },
+      "src/inferrsi/data/repo_packs/strata/pack.json": {
+        "bytes": 617,
+        "sha256": "643b32defb8db9aa230f624856f4fb4b595b5c23db58d9453091f5420cff71c3"
+      },
+      "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/bounded-serving-measurement.json": {
+        "bytes": 3239,
+        "sha256": "fb00ddac14ef6b06f48da9ba335529bd55d5d83c2a2fe331f0aa1ca16afceb37"
+      },
+      "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/extension-correctness-gates.json": {
+        "bytes": 3343,
+        "sha256": "4ea7083d277df205ea4c361bee87a8207951596e9ef6cfa136cc7d2aeab02f46"
+      },
+      "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/runtime-and-kv-ownership.json": {
+        "bytes": 3280,
+        "sha256": "e6142c1065754716a5f26f42119722908dc8fff63e667c986d57f1227b8610bd"
+      },
+      "src/inferrsi/data/repo_packs/vllm/pack.json": {
+        "bytes": 508,
+        "sha256": "1705d96710249c4514c15d8d011b43c1a1ae7c15fbc9ddf11ef2d0e6b2ab7d07"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/minimax-h3-reference-kv-reuse.json": {
+        "bytes": 4047,
+        "sha256": "80b853fc594d387ab48bed04dca245e0a22b8cdaf7c9b853551ab3bafa302e84"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/minimax-h3-sp-head-buckets.json": {
+        "bytes": 3742,
+        "sha256": "c4d29d1bceb5e2cf86151bf53c9dafb3d83428fb1e9d2a848d852bb8824bf857"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/qwen-image-21-eager-t2i.json": {
+        "bytes": 5268,
+        "sha256": "ef107d43cc448219eda4d64ab961b61cda6350e20d30d94de1ff7c4ffff7e775"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/audio/streaming-serving-measurement.json": {
+        "bytes": 3389,
+        "sha256": "2680f32f28907f6abc7a7e1e3ecab4f46ee412f40f4898ed2f371ffadf47ae4d"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/measurement-and-profiling.json": {
+        "bytes": 4472,
+        "sha256": "c8b25583cd160024c7472a20debea34cb6d4bc79714b4bad2953555adc9f794e"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/production-integration-gates.json": {
+        "bytes": 6112,
+        "sha256": "649b49ff64d3e9e88bd620027dabf9d126ca69bbe91b448f28e845ecbe1f8e2f"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/step-state-and-output-validation.json": {
+        "bytes": 5090,
+        "sha256": "e1ffe28c4fa54af739cdef2fedf4755bba18e078f8380d4fde8138ea712e5778"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/upstream-review-gates.json": {
+        "bytes": 7614,
+        "sha256": "5127709bb85ce31008d19468e1b4fcb5454a71776f69f991841cd9ef7558f9c4"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/omni/stage-ownership-and-search.json": {
+        "bytes": 3427,
+        "sha256": "e75d6c2829ae22881893f5da2ad7edba099fd64ccfb0fc93d30473e06e9d4d80"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/audio/concurrency-scaling-state-and-overlap.json": {
+        "bytes": 25324,
+        "sha256": "cb47fabb232db81ed525cbb54371becc7e4a28b2bae9d8492f60dfb6f31563e8"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-adaln-schedule-cache.json": {
+        "bytes": 4253,
+        "sha256": "ddbba3bb72b5789938595a82d3280e861f75f6ed54f05a1d2b7dcb1089d30f63"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-cuda-mxfp8.json": {
+        "bytes": 4468,
+        "sha256": "e4c538f03f4ab90a4acc8e3d8721bbd226aec1ee9bde45757a7c27d3d970821d"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-fp32-modulation.json": {
+        "bytes": 3625,
+        "sha256": "abb8d94a1b025f0e2c8e085275e96ee85d68aeb27cb12d0918e4647f8255b693"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-qk-norm-rope.json": {
+        "bytes": 3706,
+        "sha256": "4c1c4acfb9cc0baa90bc28fed1903020ea1ea38755b84315711795db45e48bb6"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-ref2va-teacache.json": {
+        "bytes": 3835,
+        "sha256": "8aeb3f99ac4227cf8b3a5cece13d938b19082e265a557eddae27b7e495bcc22e"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-vae-stacked-tiling.json": {
+        "bytes": 3508,
+        "sha256": "8cc18eacbbfd8fc4d3a6b4152c4b584d559fc9b15377da56ef072e542d807c93"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/pack.json": {
+        "bytes": 566,
+        "sha256": "58c971e9dcf6ac15752629ed4d9426aeb5b5d857fc37b5bcf921aa2e81c0cdd5"
+      },
+      "src/inferrsi/data/repo_packs/vllm_rlt/pack.json": {
+        "bytes": 649,
+        "sha256": "6097041b8d026740060e527f49b1b4e288315b1c3227f7583805b86d8fa415f7"
+      },
+      "src/inferrsi/data/skill_library/README.md": {
+        "bytes": 18819,
+        "sha256": "ae583f3578425df1dc91ba32b8b6543c0469c9a817ebea873a77051d6b13e1ce"
+      },
+      "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/SKILL.md": {
+        "bytes": 5494,
+        "sha256": "7581b56336d07213a87d01477793ae1a8acf0b7cf876fce36c9a761ee32c1ff4"
+      },
+      "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/metadata.json": {
+        "bytes": 1429,
+        "sha256": "68ac7f64b7ab947fc2db664bc3990f102c1cc5539b42350aa5818d0bbf17d86c"
+      },
+      "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/ownership.md": {
+        "bytes": 2081,
+        "sha256": "40688cc44183027efbf2fca8d3548c299cdf2e33fef54f76ab3549353eb9bd1b"
+      },
+      "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/source-map.md": {
+        "bytes": 4635,
+        "sha256": "10562a75683bd9bd07ec2e37a1cbd0f93f2534d6f85b55a566adb2d97485215f"
+      },
+      "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/validation.md": {
+        "bytes": 1477,
+        "sha256": "5a0b854b87d091d7d505b0bdd5c03330052507a851ab6edf35b93a6675f2d4f7"
+      },
+      "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/SKILL.md": {
+        "bytes": 4010,
+        "sha256": "a95569bcc0da6e715fd74783a6a1e0489e92b680c582f15dafbf98df589d505a"
+      },
+      "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/metadata.json": {
+        "bytes": 1519,
+        "sha256": "9c2293c8956508de987c886c7224d673d22912e7e915b026b05abdd287e97401"
+      },
+      "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/execution-lifetime.md": {
+        "bytes": 2147,
+        "sha256": "c4bc41e08aca68f661325f61e6f0db1760a079dc362f295c847bdc374fee3bf6"
+      },
+      "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/source-map.md": {
+        "bytes": 5918,
+        "sha256": "798b9cb89ed6b885eb47464ca9f137d2802a54571b9327e631d7742d3e5abd5d"
+      },
+      "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/validation.md": {
+        "bytes": 1608,
+        "sha256": "ebfc00bf7d6ab0a1dc415953ae84d56ccdc713d5614aff85e433c4a843793bff"
+      },
+      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md": {
+        "bytes": 7322,
+        "sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845"
+      },
+      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json": {
+        "bytes": 587,
+        "sha256": "db7ad368394ccbfdeff6a16e3423c300d51305b1f95ff5c3936df91738a8cc35"
+      },
+      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md": {
+        "bytes": 5314,
+        "sha256": "22afe4f4d77f441096cc657d55d763473bfcbea8fb531c8e493fe49ab84955f7"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/SKILL.md": {
+        "bytes": 6745,
+        "sha256": "d494fc7e539cbb7388189bd7b3f04bb5c4983fdd7a82484348aad1cee7f741f2"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/metadata.json": {
+        "bytes": 558,
+        "sha256": "71c68431663ed2336cccc985a354adab6b13b4cfd6123a18a5d8d53c28268509"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/references/source-notes.md": {
+        "bytes": 13004,
+        "sha256": "6ffbfb68cc7e04923b4fbca2a73a7555b7c3b45a6f40819bb5f3213ef05cf88d"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/SKILL.md": {
+        "bytes": 5977,
+        "sha256": "03829039860ece35519f2e4c339a48b5014ec25e07a2c150a3efb6a3a4e35b4a"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/metadata.json": {
+        "bytes": 559,
+        "sha256": "80d80c15514919826fa72d6f6cf26b76db8accd85d21187fc239232ca454e7bf"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/SKILL.md": {
+        "bytes": 14703,
+        "sha256": "4ce9dcadd1ae3c7a8cc633336760d0a3c382f354f1b7028e3a10d28fa0d0c9fa"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/metadata.json": {
+        "bytes": 1403,
+        "sha256": "b6cb25879b1bca5f5af08b48776f2e664cf5dcd11a0efb1c4f56879e2afa9006"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/references/awesome-ascend-communication-tests.md": {
+        "bytes": 9388,
+        "sha256": "0cb27a3d24409171bb414d7ebf0c49ad62be5f46482f29624695496ba9eb3e63"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/SKILL.md": {
+        "bytes": 5839,
+        "sha256": "9203cc5a14a1d65df5fbf275734b93be023e06527459b2f79864f28a338cb44a"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/metadata.json": {
+        "bytes": 536,
+        "sha256": "72b41e7845915d34b09219624e82976386f3125816293540c2c38143fe6b477f"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/SKILL.md": {
+        "bytes": 6440,
+        "sha256": "8fdf46fb2743b67a66bc069d22a45305a222debfb606777bc451b4b8f31d9cc6"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/metadata.json": {
+        "bytes": 536,
+        "sha256": "be681a3206e5e183b4c9670c9d7784bcffce18c891449e5b15f94ca5f1324b9a"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/SKILL.md": {
+        "bytes": 12988,
+        "sha256": "401981cf7828edcab0c9ea882f7ecd83badfa3c2a95006e912cf99d67b3d6a12"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/metadata.json": {
+        "bytes": 760,
+        "sha256": "efd37bb23acb500dad4ec8ebe7ddfe5c3f196374052825f923ac85da8445f06b"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/SKILL.md": {
+        "bytes": 12236,
+        "sha256": "38b151bc6b4e9734c6fa41efb7a30343559400c0c75bbdd26bd8a968b8ea4592"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/metadata.json": {
+        "bytes": 557,
+        "sha256": "04d2002fc8a0a7becf86b0b73daaf2da8716ab466d7f988da43d41a10f26680e"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/SKILL.md": {
+        "bytes": 9172,
+        "sha256": "647f5e9a288f2c4e1617ba53df59d9d827df749eb4fe86182d21a9c41175caf8"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/metadata.json": {
+        "bytes": 536,
+        "sha256": "f425f7c1abbd616387273e8932c6bd02f4b1da012f099286ad4171e0f6037b22"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/SKILL.md": {
+        "bytes": 9880,
+        "sha256": "a057b7f86a24f34a8f84abbbb4910d58d4e50ff47060e1c404600017e3febe89"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/metadata.json": {
+        "bytes": 2297,
+        "sha256": "9e705625c8308013c287576b0247da6f6116d817a4a535cbecff80abd6ea35ee"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-CANN-2.0-LICENSE.md": {
+        "bytes": 5262,
+        "sha256": "89b1c114df62cfd183004f38d18d07f186436c3b37203dd69c3e84becf9d2282"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-stage-evidence.md": {
+        "bytes": 11789,
+        "sha256": "f56e4125325c9c1b9235a1d64fd732861c7686444492422daff44ba8dd648be4"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md": {
+        "bytes": 11860,
+        "sha256": "93aa1b775ced304cebfe6c4dec0503ddb2260ecb3cb01edc17ce329446c5ce2b"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/metadata.json": {
+        "bytes": 552,
+        "sha256": "f6d94f104cb70969024b2f4542078a240e200712c66279eca8b94f60bee1dd71"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md": {
+        "bytes": 8972,
+        "sha256": "0e6f094dc554abea12e07def50f20b233b47127779a8aefad595b4ab1efd24c2"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/metadata.json": {
+        "bytes": 569,
+        "sha256": "acccc8eb86de0df6caf78a9b3c6db76a2a686adc4a5ba0029a4d62d50ff9a8f8"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md": {
+        "bytes": 5989,
+        "sha256": "8428df7157af024fa17cfe0b3818ed54bd9f468306dda46eb8d4f0edf4f3f230"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/metadata.json": {
+        "bytes": 579,
+        "sha256": "c650dce88c07b636d68e063dbda4d319f6f7f3ecbcdd5e47f393cbfc3af7c71d"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/references/source-notes.md": {
+        "bytes": 11470,
+        "sha256": "80cba4d3a884e7ee2618326cf74b9f9c3a34e4d9cf443707f6bbe673f2d5ba65"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/SKILL.md": {
+        "bytes": 6159,
+        "sha256": "23e93ea7c98f99743509fdc04139d78ed2e7123c63e80b366c6beaba7eb70d0d"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/metadata.json": {
+        "bytes": 581,
+        "sha256": "97836a38465106d816a9ace894a7b277d6d330f9accade8f8493ce6f0fb57ef9"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/references/source-notes.md": {
+        "bytes": 13493,
+        "sha256": "435193590b4a2ba32e2e4c6fa9d7e8f889528d412cbbb48e35ec5433f7dab384"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/SKILL.md": {
+        "bytes": 7137,
+        "sha256": "a25738dbb3b9daa15c5827cb536a20fe13f7c8b9dd14aa3fb7cfa7e132879431"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/metadata.json": {
+        "bytes": 1007,
+        "sha256": "3abe626a6e4c73c943429f609ce2c62aa77aae7e1726475ebbbdfbe1794f6f7d"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/mechanisms.md": {
+        "bytes": 17376,
+        "sha256": "fe08b2292fa587be3874774ddf41063b84c579a0085df422ec6710d8aa90b8b7"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/INTEGRATION_NOTES.md": {
+        "bytes": 4024,
+        "sha256": "371b01ee201ef51fecbfd6da892d517aaba295930b9395025f03f34e21c030a1"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/NEXT_GATE_LEADS.md": {
+        "bytes": 1623,
+        "sha256": "35b8b58ae5db02b2df51ccbe5cdf870a7383d17f345f9ee5eb7b0a4146dfc74b"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/README.md": {
+        "bytes": 2888,
+        "sha256": "e0b1ec1b39c10067f5ddce58320851ad089568c4b2296a24777389097cb5b388"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/io-reduction-can-plateau.md": {
+        "bytes": 2671,
+        "sha256": "5cc35f514224ac6e6efd2da6db221bba468b2f4db66807295e67b240bcdfdba5"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/overlap-and-fp8-counter-signals.md": {
+        "bytes": 3692,
+        "sha256": "80c8a60a6df55f3d84fab23e03567d8b38e23bcdbd63cc09881e46db1e2888ab"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/partitioning-after-fusion.md": {
+        "bytes": 3132,
+        "sha256": "4d4cabe0991eab554dcfbd4fa9c8199e61fa4767709b0c4cd646e2588bb398a2"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/knowledge/attention-cost-model-boundaries.md": {
+        "bytes": 5223,
+        "sha256": "5ed8f888443e38ccb72545b6025447aab7b92936f78c151fed680e582126239b"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/sources.md": {
+        "bytes": 9069,
+        "sha256": "53436e0de4e6dda2d29426591bf61f8526378cdc6f9ec7967b88b1bcd4c61e6f"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/host-overhead-profiling/SKILL.md": {
+        "bytes": 11199,
+        "sha256": "326e566fbe30d3ba04e4b50c3e87aad9df037f0ac4ad6fe129165ae97dc5f91d"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/host-overhead-profiling/metadata.json": {
+        "bytes": 527,
+        "sha256": "76e34c1c64aba473a9e126b032a776417caa0f913f6ac3cf4dba3a5d123b1ee0"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/SKILL.md": {
+        "bytes": 4905,
+        "sha256": "cec4e3aa1eb64c31c54b5f5ca40af2b61b7919d94be4174968581d5f752c0538"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/metadata.json": {
+        "bytes": 7453,
+        "sha256": "7520da1cc97412631a600957ecb18aa5a7a4461ea21e76c237330180193ed898"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/fp8.md": {
+        "bytes": 6287,
+        "sha256": "31d9c211c11c2224007abbdcd4059913c4a54833fe2d2f926411c5238fcd3501"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/cutlass-bsd.md": {
+        "bytes": 1547,
+        "sha256": "42fec630f410aa308f70a51a89fadcd19586fa620f9831a32bee528a9a10000e"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/deepgemm-mit.md": {
+        "bytes": 1065,
+        "sha256": "6c73a30c14a2938abaa6afe98ebb19e6d81dac5bd4ace0cb05c61c20afa37b85"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/kernelwiki-kda-mit.md": {
+        "bytes": 1063,
+        "sha256": "84aab57158a914f448b38d8be29dd842b634a08c192d844f416a256beed4ae94"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/notices.md": {
+        "bytes": 3240,
+        "sha256": "876ec016644bf5331b3f257985e126c0e61980e3bfd82569b7b0c3f6b2c6c64c"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/resources-and-scheduling.md": {
+        "bytes": 8919,
+        "sha256": "1ce25b426736072435a1ead9a2a8e1f56df5d0abb8198e8c4fdcb34321db4964"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/source-inventory.json": {
+        "bytes": 120506,
+        "sha256": "1dc09c895c970a0c9708158bcc6dd65cdb736274fd2c35d44542c70407d3849d"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/sources.md": {
+        "bytes": 9004,
+        "sha256": "6b74b9e774f2a80db0860510d9e6940d36284430f29c1e124f115a9eaf25833d"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/synchronization.md": {
+        "bytes": 6843,
+        "sha256": "4322f6b3c0e4368bf8c192b4ec752529767f50d19c0c38dad8fdf9cd781e41d9"
+      },
+      "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/SKILL.md": {
+        "bytes": 7348,
+        "sha256": "fd1f5a7e6a55714730044155f5b38c1d700383d4e16c3e91963bbfbe1c773af5"
+      },
+      "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/metadata.json": {
+        "bytes": 572,
+        "sha256": "19e40763fedc20990651bb989c32a7693306c5c54af46c3d4be77fd4f87b6c03"
+      },
+      "src/inferrsi/data/skill_library/hardware/host/host-memory-management/SKILL.md": {
+        "bytes": 23213,
+        "sha256": "2c807f3df2da2f533d25183de5771f3d412058d16410994292ef8f48c49f9b3a"
+      },
+      "src/inferrsi/data/skill_library/hardware/host/host-memory-management/metadata.json": {
+        "bytes": 1684,
+        "sha256": "9ac6af135f0b6d76a76bc201b6bbd8ac9afe5201dd1e2ea82c0ac4826de401da"
+      },
+      "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-LICENSE.md": {
+        "bytes": 11357,
+        "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4"
+      },
+      "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-native-offload-readset.json": {
+        "bytes": 6267,
+        "sha256": "cc6c2011be746002ba3ff96ccefab1809863748ddd17eb102b75e7b0e134fe55"
+      },
+      "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-native-offload.md": {
+        "bytes": 8966,
+        "sha256": "6c2a72b5ba3708a01d229e0576570837a617e6a42dbb0b5312a26a65a6c31542"
+      },
+      "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/SKILL.md": {
+        "bytes": 16072,
+        "sha256": "36ef88826a32ac3147b869001004ded17a8d5a491e25bfe105005ba39ea9d1eb"
+      },
+      "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/metadata.json": {
+        "bytes": 539,
+        "sha256": "ae71c389dfdb1a19d267feecb4c6f1fbb4c9afad501ff6d8fab7e38b29bcc37a"
+      },
+      "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/references/deepinfer-zero-overhead-scheduling.md": {
+        "bytes": 10894,
+        "sha256": "f452bf40132c9195d739f1c86b3aa18857138c7ed0f1861af5853d93b9955924"
+      },
+      "src/inferrsi/data/skill_library/hardware/host/os-process-placement/SKILL.md": {
+        "bytes": 10327,
+        "sha256": "2003492e4d89fde0443b180d74933e764bd1979435e109c8d26e20eb8f6e5053"
+      },
+      "src/inferrsi/data/skill_library/hardware/host/os-process-placement/metadata.json": {
+        "bytes": 1399,
+        "sha256": "ec02995c075f0a540921f57b4846f2cafe334eb90f44389f17c3c50457c3e9c5"
+      },
+      "src/inferrsi/data/skill_library/hardware/host/os-process-placement/references/llama-numa-thread-readset.md": {
+        "bytes": 6617,
+        "sha256": "748b254b3b043a5ddb0f480eb4a566854d27eedd7b111499863e9bf0a5a6951e"
+      },
+      "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md": {
+        "bytes": 8813,
+        "sha256": "63b883c10a8d3e7d918abfc7175b72e2b401ff1e02697f94b8382ec1c8f369b7"
+      },
+      "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/metadata.json": {
+        "bytes": 558,
+        "sha256": "0a41c7b00978332b54ee1e4a7c1ade88ecde3e961128f24d1b4f7708caff5ff5"
+      },
+      "src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md": {
+        "bytes": 6402,
+        "sha256": "0a5f48d9352c5bf460775ef47f9136829978811f4a16e84983e299782cae305d"
+      },
+      "src/inferrsi/data/skill_library/methods/assess-serving-preemption/metadata.json": {
+        "bytes": 537,
+        "sha256": "77ce5ba16245cbd80d46d0abdf80aaf26fc0924fa2d2a5624ff430de68a1daf0"
+      },
+      "src/inferrsi/data/skill_library/methods/assess-serving-preemption/references/source-notes.md": {
+        "bytes": 30405,
+        "sha256": "f29fcecfba0490008a354c11b42403983fe165e56ff907ad61cc6040d2b6407f"
+      },
+      "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/SKILL.md": {
+        "bytes": 3181,
+        "sha256": "1e64f28ea71672fb53c56f6108b87864738c6d03d6dcd32eedaef9e0625e96e3"
+      },
+      "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/metadata.json": {
+        "bytes": 478,
+        "sha256": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7"
+      },
+      "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/references/source-notes.md": {
+        "bytes": 10548,
+        "sha256": "61b1609f71c37b61d136b8c4d39f793de87cffb1cf31c8ec737a1c3fa6fb1b4e"
+      },
+      "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/SKILL.md": {
+        "bytes": 3884,
+        "sha256": "ac773c6f1c110a664dfca28a3ae9129b10c07614e60ca3640dea776220e7b352"
+      },
+      "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/metadata.json": {
+        "bytes": 648,
+        "sha256": "261ce9f644ea6eebb90c830f41fb8b320d2b401456e5bd8d494095a44b8620d0"
+      },
+      "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/references/source-notes.md": {
+        "bytes": 8415,
+        "sha256": "9696d3c0ef96b6611787446bd6f96d4d6255ca8d107105de8c1d9af83edb9fa5"
+      },
+      "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/SKILL.md": {
+        "bytes": 6416,
+        "sha256": "6b7db1eb2780dec27858cb7aa0a3c093c5b5dfa4c30261aebfeaa255213ff688"
+      },
+      "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/metadata.json": {
+        "bytes": 582,
+        "sha256": "4dc8659806539b4784ea9348a01d4194ac70d0bb297a916f90ff6419efa5c328"
+      },
+      "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/references/dlrover-persistence-replicas.md": {
+        "bytes": 8398,
+        "sha256": "055f4357f79026fcbcc766bbe77db610376ec353623f8898f5d14a30ee69e7bc"
+      },
+      "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/references/sources.md": {
+        "bytes": 7984,
+        "sha256": "9425b1aab963a2c8d35f3db897bd15ee69699bf5b0d0c8425c3c8bba3e9a0233"
+      },
+      "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/SKILL.md": {
+        "bytes": 3068,
+        "sha256": "25d04aed9782b2a8ff1186828becdc7fce93c4418039ceaef9b8dcc5ec9bcd24"
+      },
+      "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/metadata.json": {
+        "bytes": 478,
+        "sha256": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7"
+      },
+      "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/references/source-notes.md": {
+        "bytes": 8521,
+        "sha256": "9f9184d46b19312138a6e805e439e43dc0594a0cc6b46675255a582dc946ca5b"
+      },
+      "src/inferrsi/data/skill_library/methods/dense-performance-evidence/SKILL.md": {
+        "bytes": 8305,
+        "sha256": "53cc1867a13f54d1006d50517dea8848869ca50df3f0cd610b2424cc8d18c933"
+      },
+      "src/inferrsi/data/skill_library/methods/dense-performance-evidence/metadata.json": {
+        "bytes": 707,
+        "sha256": "a0d1e3b17140295510a764045d1b222e19be3a3406e6d1e7dddaad49ff474490"
+      },
+      "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/SKILL.md": {
+        "bytes": 6667,
+        "sha256": "82d0099d3c33706f46c657d4fa7f76d3d486e019b36760d1de001921328a60cc"
+      },
+      "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/metadata.json": {
+        "bytes": 547,
+        "sha256": "37379a4d4a058ca46da2b20a920dc01be9ec11e557c32b2332702ec76027f4b5"
+      },
+      "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/references/source-notes.md": {
+        "bytes": 44796,
+        "sha256": "ea416831fd7d0874e06d0a76fb0c7196df53072b717673a0cdf635db6bfd917b"
+      },
+      "src/inferrsi/data/skill_library/methods/inference-benchmark-design/SKILL.md": {
+        "bytes": 8974,
+        "sha256": "19ebc1f567aadf49f7a534a18fdb629c72c0e4ce043d360983136491cb72bfab"
+      },
+      "src/inferrsi/data/skill_library/methods/inference-benchmark-design/metadata.json": {
+        "bytes": 5923,
+        "sha256": "18310402a8ade03fd48ac687f79ba9d447871e90337de84c0e1bc15c0b827306"
+      },
+      "src/inferrsi/data/skill_library/methods/inference-benchmark-design/references/benchmark-commands.md": {
+        "bytes": 8466,
+        "sha256": "0898424e0bff3a6246e1eeeda6abd6c5bd974c047fe32f3bbd5ab9d31ce17cd3"
+      },
+      "src/inferrsi/data/skill_library/methods/inference-benchmark-design/references/task-benchmark-selection.md": {
+        "bytes": 25086,
+        "sha256": "0d4bb080a754e26846e81680023db0dff5e6a4c96be4327c411c3b72ac4ef8e2"
+      },
+      "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/SKILL.md": {
+        "bytes": 6644,
+        "sha256": "89a6a82c692f7f5192f4d7f4c3a5ff08a4df3b1c4bdaa0aafb7e04292d8aa5b4"
+      },
+      "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/admission-drain-occupancy.json": {
+        "bytes": 1354,
+        "sha256": "3319e23bbe2bd40e2d45fb89dc87bdc92d4df2b5ddd8086fcf84813b43885b7c"
+      },
+      "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/admission-drain-occupancy.md": {
+        "bytes": 3134,
+        "sha256": "e88bbac5d253a8e7e3a3d5e21047f23f2970c11ac23887d4303ab94fca52dd27"
+      },
+      "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/drama3-rejected-directions.json": {
+        "bytes": 1459,
+        "sha256": "2cc0a526fe0509551607a03e7626dd3d7844e643c4afb194837f75cddead8b12"
+      },
+      "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/drama3-rejected-directions.md": {
+        "bytes": 3557,
+        "sha256": "6f143bc363a3290fc008c08e57cb075f1776d1c8cca8a6a59297d5b1fbbf61bf"
+      },
+      "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-ascend-rejected-directions.json": {
+        "bytes": 1293,
+        "sha256": "b3d538f79c56724ccb2895cbf50cf2e6b0fec3979be62a684a8230cd932dff55"
+      },
+      "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-ascend-rejected-directions.md": {
+        "bytes": 2564,
+        "sha256": "be396983812262515899a3e2ba4f01b23ff92e1b836881882c7580372b8ae552"
+      },
+      "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-b300-rejected-directions.json": {
+        "bytes": 1196,
+        "sha256": "146ea1564a46ed03092c7dc5796b318fc05b04b1821105ee00eb7fa7fa1333e2"
+      },
+      "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-b300-rejected-directions.md": {
+        "bytes": 2656,
+        "sha256": "cce2930f544a0f78f21797fba4d8ded298e4a81f76c34d7977b0e31d361bc738"
+      },
+      "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/metadata.json": {
+        "bytes": 535,
+        "sha256": "fb55067fb45970319787068a3206110641cceaf0ccbabd706be6d0b3d2afc2ed"
+      },
+      "src/inferrsi/data/skill_library/methods/kernel-writing-guide/SKILL.md": {
+        "bytes": 10029,
+        "sha256": "01fb1712d7eb988c5e412f02340d2e48b4f8564fcf4ab232dc5897869d0d50a0"
+      },
+      "src/inferrsi/data/skill_library/methods/kernel-writing-guide/metadata.json": {
+        "bytes": 653,
+        "sha256": "0189d184337662056fc6952c7beffc1bf6154c4fa173ae3e3e463bd22d70f36c"
+      },
+      "src/inferrsi/data/skill_library/methods/kernel-writing-guide/references/akg-affine-tail-guards.md": {
+        "bytes": 11941,
+        "sha256": "113deb2db10b171db3425bb3f1792bee763706b794c8cc2e222b4cc0d0114118"
+      },
+      "src/inferrsi/data/skill_library/methods/optimization-search-policy/SKILL.md": {
+        "bytes": 12914,
+        "sha256": "882e7850b9b49ad692a2d11618701afe0f4e83727ea78df114be2029923c725a"
+      },
+      "src/inferrsi/data/skill_library/methods/optimization-search-policy/metadata.json": {
+        "bytes": 1403,
+        "sha256": "9f1fd8a2af52bc964021e21f16ecbdb891e08f2057370e700d67f86605d333db"
+      },
+      "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/INTEGRATION_NOTES.md": {
+        "bytes": 3597,
+        "sha256": "157e74ba180426b279b548a729769f152bf18f9a78ffca4ebfcaf26b7d0c5292"
+      },
+      "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/MECHANISMS.md": {
+        "bytes": 5875,
+        "sha256": "18bb2f5d7e52773c516420a142edfb531d0b9c822a2934d60f3234ac1cb9d40e"
+      },
+      "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/NEXT_GATE_LEADS.md": {
+        "bytes": 941,
+        "sha256": "16a264c7159f9c2c700ce8ba46e63f3e636b277c987da4970f6d2fa085fcb053"
+      },
+      "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/knowledge/adaptation-interfaces.md": {
+        "bytes": 3893,
+        "sha256": "5e761957cb5ecd93926e6ffc13deee7c830c6e8f0ee3c3846e9dd3b50dc74446"
+      },
+      "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/references/source-notes.md": {
+        "bytes": 13370,
+        "sha256": "54d83c3713045d59a48821bcf93598374ba4a630e3891ad39498f919317c4fb8"
+      },
+      "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/updates/optimization-search-policy.md": {
+        "bytes": 4238,
+        "sha256": "d3db44fa9a6831b86a02c638b323ed943e69fe256866213f10d68c72d762879e"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/SKILL.md": {
+        "bytes": 15324,
+        "sha256": "c5f78961c23e3c8d659f5675b522518ab06e1f2a8ea8bc74c5d428ffb41693d5"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/metadata.json": {
+        "bytes": 3669,
+        "sha256": "60b5b24ead590b0fb222398fba47822bf6723690da683e256a450bb5c0eeb806"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/INTEGRATION_NOTES.md": {
+        "bytes": 7349,
+        "sha256": "083e462e0a204dce599362d46712fcfb392d053c5cda56216e22e28d4b8fa0fa"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/MECHANISMS.md": {
+        "bytes": 4003,
+        "sha256": "742c249082ceca230326f4407a40130fe624553b796ae5a157a8a4c15a7e1398"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/README.md": {
+        "bytes": 1204,
+        "sha256": "17b2d1c7c14124d59ac8a0d5ad4200ad2803d991e784a1a98cd97188cc364999"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/READSET.md": {
+        "bytes": 3937,
+        "sha256": "66459ce393858b2b2160f1e2973373bfc4b9902a12644631934b4ebe3140296a"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/STATIC_CHECKS.md": {
+        "bytes": 1111,
+        "sha256": "cb0aa7d148352ebb7b1678fc42a557a099575054716a341dfa25b6a998029064"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/cases/decision-cases.md": {
+        "bytes": 5491,
+        "sha256": "781a675a6533e0a18c044aaaee0d010da7013b6504654b8b8292a35176db801e"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/govern-optimization-experiments/procedure.md": {
+        "bytes": 3823,
+        "sha256": "ce036a4d8a0908da204515c47e2ff9623548d09de264fa65bd58fedce52a78a0"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/govern-optimization-experiments/references/source-notes.md": {
+        "bytes": 21659,
+        "sha256": "f9a8254e45473ddb2b73e576957b6d3db83c62b99dff1922f4d1e606e569559e"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/knowledge/bounds-predictions-and-decisions.md": {
+        "bytes": 5297,
+        "sha256": "5780d6eb5257612ec45deb94f85f334a6ecabf7e466d1678706041f6549bda79"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/redirect-cross-layer-optimization/procedure.md": {
+        "bytes": 3648,
+        "sha256": "6593a63fce9c474b515663a560c94b7df048dc5f6e77a95b8e54e4a64cd10a09"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/redirect-cross-layer-optimization/references/source-notes.md": {
+        "bytes": 26469,
+        "sha256": "6ff3a7bc31f1172db92aedaf7df1f1aac688bf769227b583a4cd82bfad52da67"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/agentic-kernel.md": {
+        "bytes": 41353,
+        "sha256": "7190573b85f94a4391bb0d887ccbd37a31b3f08bb0531114ad7afd72fe1567a7"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/flashvector.md": {
+        "bytes": 9830,
+        "sha256": "3926369acd88ff3c370bee5718e74dd15f3219c3e4f6311095e30e917bffa4cf"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/inferencebench.md": {
+        "bytes": 9294,
+        "sha256": "21f47111197a0b7f8d6384ea918d139dc8318a62ca3313ce060048271f50b639"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/prior-coverage.md": {
+        "bytes": 3254,
+        "sha256": "35d47ac078591221fcb666219bcd824f9400fc17aafecdc0499d8336f7178238"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/vidur.md": {
+        "bytes": 7736,
+        "sha256": "d38602af5219f5feeb49e6cd6bf53d66098c77fad141dae676f1d8a6743cf25d"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/roofline-contract.md": {
+        "bytes": 7091,
+        "sha256": "419ec4bcaec45ac579a133bce083d4f48779f26e664d7a86f720fa65f343b1d1"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/roofline-example.json": {
+        "bytes": 2015,
+        "sha256": "597cd3205df043cff3ae9d15924c77934e8d8eb300c00c4840f6bb2921f90faf"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/scripts/roofline.py": {
+        "bytes": 9004,
+        "sha256": "524de48a6764401695fe1444f1fa3d38f363069b0cd2fba4715a86ca4b28a5d4"
+      },
+      "src/inferrsi/data/skill_library/methods/plan-activation-residency/SKILL.md": {
+        "bytes": 3507,
+        "sha256": "542d8a1e6963f5019986e1fbd79d3a247a045a51487c09313026087ac457c15e"
+      },
+      "src/inferrsi/data/skill_library/methods/plan-activation-residency/metadata.json": {
+        "bytes": 478,
+        "sha256": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7"
+      },
+      "src/inferrsi/data/skill_library/methods/plan-activation-residency/references/source-notes.md": {
+        "bytes": 13907,
+        "sha256": "5ff52655cfd06fcf970ed2a7d4d4fe1c31115b102af2d41497de0c74848ed2b3"
+      },
+      "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/SKILL.md": {
+        "bytes": 13800,
+        "sha256": "7e20d3390f2501263e5e298e026c47cac16c234f2397833131e15ae5d34c0b97"
+      },
+      "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/metadata.json": {
+        "bytes": 1903,
+        "sha256": "d0e4d6950250e4cec43d82e23324c9301a925d4c22be30b4d2713d790cb6504b"
+      },
+      "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/references/capability-presence-audit.md": {
+        "bytes": 11461,
+        "sha256": "c21eb4a6aabe811a49a6540a34c54dbcfdcf7efae3052a79b54034c7838026a5"
+      },
+      "src/inferrsi/data/skill_library/methods/skill-evolution-playbook/SKILL.md": {
+        "bytes": 9448,
+        "sha256": "9cd37f21fe0d862069e4798edca8a04d59f8b840e55d521901594a63dddd9ba3"
+      },
+      "src/inferrsi/data/skill_library/methods/skill-evolution-playbook/metadata.json": {
+        "bytes": 527,
+        "sha256": "92b36338b9257e2e41b95e4370694d7de52ed84bc7971056b3bfced07b0993d8"
+      },
+      "src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md": {
+        "bytes": 6865,
+        "sha256": "64bd2c6b795d966045d1d3cca55976d3d49b08b06df7c7e06a70c2824cf55b7e"
+      },
+      "src/inferrsi/data/skill_library/methods/training-input-pipeline/metadata.json": {
+        "bytes": 587,
+        "sha256": "80659099fce6a6ec05e9084b75452c877a5812e30f60a78f31ef8c7efcdc1bac"
+      },
+      "src/inferrsi/data/skill_library/methods/training-input-pipeline/references/sources.md": {
+        "bytes": 2637,
+        "sha256": "60e0f3e88655d8e1d5eb4797f39ca4370deda01614f265b8d2ea4559d8ac020f"
+      },
+      "src/inferrsi/data/skill_library/methods/training-resource-budget/SKILL.md": {
+        "bytes": 7200,
+        "sha256": "bdfa5a4ef37d9f303c3a23cd6f98c21137b97ac5ed58e0ed5fac6d71d33e52ef"
+      },
+      "src/inferrsi/data/skill_library/methods/training-resource-budget/metadata.json": {
+        "bytes": 976,
+        "sha256": "d42b614af9cee6ee0d4a25dc5d57c7b98438feed1e807a7c9f1755109a86e9cf"
+      },
+      "src/inferrsi/data/skill_library/methods/training-resource-budget/references/mindspeed-historical.md": {
+        "bytes": 3384,
+        "sha256": "eb97f935ed30517a8fc9edec5e9cddc04f5d679d156c16d75bd29c534593360a"
+      },
+      "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/INTEGRATION_NOTES.md": {
+        "bytes": 7368,
+        "sha256": "c55c6ad2131ef788866411dcc7056f221420ebff30c95c6bc919b09a4818cb73"
+      },
+      "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/cases/paper-grounded-cases.md": {
+        "bytes": 5649,
+        "sha256": "19bd25b1a944adb2351ebc6b1d795150ea033041a44f6cc261e881677812a101"
+      },
+      "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/knowledge/offload-delta.md": {
+        "bytes": 2026,
+        "sha256": "f8740fc5c8bbbe5ac13ff1cfec7ca499569d0b46e8cd56cf1f5a44c3a5e72242"
+      },
+      "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/notes/training-memory-scheduling.md": {
+        "bytes": 18660,
+        "sha256": "9aa6d3e01452dff95c8a8940ff62ead8c146ce215c9175b78c9f2ef508a7b2de"
+      },
+      "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/references/source-readset.md": {
+        "bytes": 22946,
+        "sha256": "2a5c166767f56fa47e7eba7386f1f36da8cb73702e5b5bbe14d6a456ca9ebfae"
+      },
+      "src/inferrsi/data/skill_library/methods/training-resource-budget/references/sources.md": {
+        "bytes": 2752,
+        "sha256": "6d248f58af8b350fc66c22898988d6f451ac60d03bcd805cd6e84df5e55174ac"
+      },
+      "src/inferrsi/data/skill_library/methods/training-resource-budget/references/stas-decision-notes.md": {
+        "bytes": 4086,
+        "sha256": "f1a993e6c4ca9a035a2effcaf5500081f227b0cc56cc885a6b22fda0041fc744"
+      },
+      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md": {
+        "bytes": 25969,
+        "sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213"
+      },
+      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json": {
+        "bytes": 2418,
+        "sha256": "798cbff6f10b7c24b188080ebf233b188b0e98ef2c6fdb5e00e1fe3eff727c0f"
+      },
+      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md": {
+        "bytes": 8656,
+        "sha256": "5a39f14b102b0cee304541e011aa31e44e6b13527d1bee9987c3f3b66c064d3d"
+      },
+      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md": {
+        "bytes": 6392,
+        "sha256": "6c640f28ee3f0253a17b81461c865cb9f947367002cb75bf72550c4572d02e97"
+      },
+      "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/SKILL.md": {
+        "bytes": 7181,
+        "sha256": "a46110891eca860915e314d6921a22856d631453b22b17837fe30262d6780535"
+      },
+      "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/metadata.json": {
+        "bytes": 2175,
+        "sha256": "4e6100dfb9fc7bc00b19915f4fd38ad509704f2a609c658adf4d632d4e05b6c1"
+      },
+      "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/references/afd-plugin-LICENSE.md": {
+        "bytes": 12747,
+        "sha256": "0e8566ea362276878e1f38f97da49964dbea301c15d4aa9d727f97a4c39fd943"
+      },
+      "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/references/source-notes.md": {
+        "bytes": 17240,
+        "sha256": "5073d87450cb60d1b245308dff8807e73749c07de976d50f2fa26694fad81a72"
+      },
+      "src/inferrsi/data/skill_library/repositories/comfy-ui/LICENSE": {
+        "bytes": 35149,
+        "sha256": "3972dc9744f6499f0f9b2dbf76696f2ae7ad8af9b23dde66d6af86c9dfb36986"
+      },
+      "src/inferrsi/data/skill_library/repositories/comfy-ui/SKILL.md": {
+        "bytes": 4507,
+        "sha256": "c993bd6ff8e9b5f82f39f175ec8242341136e4ee0808af414e1d94ad3e196c5e"
+      },
+      "src/inferrsi/data/skill_library/repositories/comfy-ui/metadata.json": {
+        "bytes": 2032,
+        "sha256": "74742851a002a51c208362cdcd2a59fdab5d1734d8492b1f93fbec3cf21e0841"
+      },
+      "src/inferrsi/data/skill_library/repositories/comfy-ui/references/graph-cache-invalidation.md": {
+        "bytes": 7679,
+        "sha256": "614c9a4e6943d24d7d7be7d632682ce8b1dcc239124a37d9dc1d487113772eac"
+      },
+      "src/inferrsi/data/skill_library/repositories/comfy-ui/references/model-residency.md": {
+        "bytes": 7191,
+        "sha256": "014c68386e7cdfdb7b4f566836e8b2ff61bd5a5a506e7b4723f83fe80c5d4de7"
+      },
+      "src/inferrsi/data/skill_library/repositories/comfy-ui/references/repo-provenance.md": {
+        "bytes": 5648,
+        "sha256": "ca996ef515af42a3ce304747fa3f12a31c00533deaace5ba2ce21aefa0a62e2e"
+      },
+      "src/inferrsi/data/skill_library/repositories/comfy-ui/references/validation-boundaries.md": {
+        "bytes": 3523,
+        "sha256": "1913dedbeb558d872d6a0e015cc11d822910c3cb259a236db29fb60ad8a64263"
+      },
+      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md": {
+        "bytes": 12294,
+        "sha256": "4703beb6e5194ef2ff2c0b5ec69f64e35417a4bc0dfb50a10c80a920b77ef3b5"
+      },
+      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/metadata.json": {
+        "bytes": 2333,
+        "sha256": "014be24d0e669f4acbc0717dc323eec784026cf185186619d5a19d6d84496131"
+      },
+      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/source-notes.md": {
+        "bytes": 12036,
+        "sha256": "fd917bd5ac80bee9c9f4525f8e0300f6d67dc144e2a16632af42a67f3d53d098"
+      },
+      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/validation.md": {
+        "bytes": 5402,
+        "sha256": "60e71e70c5519bcdf6c171eec2be2f342e8801bedd746d6005f11adc646dfaff"
+      },
+      "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/SKILL.md": {
+        "bytes": 11202,
+        "sha256": "f157562d1edc905cb4fc07246567e3c5b844efb9fbf5fe39f0d40553a1b1a307"
+      },
+      "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/metadata.json": {
+        "bytes": 2156,
+        "sha256": "4d68bf4c1767eaee12ade963bbd551d0e61d1bcd81a984bb02fd3aae46965919"
+      },
+      "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/references/source-notes.md": {
+        "bytes": 11633,
+        "sha256": "6fa955022f16cd598f44fd6521901d1cb868b06cec545cabe2b82769a641b5d6"
+      },
+      "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/references/validation.md": {
+        "bytes": 7196,
+        "sha256": "673d93f6700052974b340fab25b3438e0e6344960569ac9abec7d7720f662288"
+      },
+      "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/LICENSE": {
+        "bytes": 1077,
+        "sha256": "663e51edcd9a3a8f858018ce6924ad2895a24c269c26c39ad75325bcd0f7e5bc"
+      },
+      "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/SKILL.md": {
+        "bytes": 5432,
+        "sha256": "fdb4d9483c6efa836f9e59617da605fa9c2186e11c2041b05d5c81850ba0475d"
+      },
+      "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/metadata.json": {
+        "bytes": 4923,
+        "sha256": "fe626a77af3f084f36fd73ffdb6f67976e7dceeafb94ba22078ef03c3a96cfae"
+      },
+      "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/THIRD_PARTY_NOTICES.md": {
+        "bytes": 1774,
+        "sha256": "a0d77db7d2aca0eda811ab3c386cc32b3143f2c2609a042cf4642954f0f6fb09"
+      },
+      "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/minimax-h3-apple.md": {
+        "bytes": 12059,
+        "sha256": "57ce01754ba96c378f13a91305ef3d418b926f99b2753f0af99315572bce063a"
+      },
+      "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/minimax-h3-weight-inventory.json": {
+        "bytes": 10664,
+        "sha256": "fbaacaeec340d3ce893556ee864405a0746af4d13530a38cb3d5aa9d60929271"
+      },
+      "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/phase-residency.md": {
+        "bytes": 9408,
+        "sha256": "2836a550eb2be7f5a862f51ec7b135ebd9715bf3c4faa84a7ecbc9d9c11f5d8d"
+      },
+      "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/precision-and-streaming.md": {
+        "bytes": 9440,
+        "sha256": "7e166c89193f4c7aed852ec95346a2de2b5008dac002e2e1862be1b3e57f6093"
+      },
+      "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/source-identity.json": {
+        "bytes": 7747,
+        "sha256": "730f8dd23648a3f8f47b9726cc613d893a79d1817d515eec609a9b06d8a9ff90"
+      },
+      "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/source-notes.md": {
+        "bytes": 9448,
+        "sha256": "7304bf9369efc157dc7aa40686dd828cc14db45c49442e908b3d3cc4b329eb86"
+      },
+      "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/validation.md": {
+        "bytes": 7880,
+        "sha256": "0754873a758e048480938314894e25b504ae0c269549312e7f8e6020d83fe294"
+      },
+      "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/SKILL.md": {
+        "bytes": 9338,
+        "sha256": "95e81b6bc096a954fa73139ea5bc001d1efa6c99747161da62f012b494e5c013"
+      },
+      "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/metadata.json": {
+        "bytes": 1718,
+        "sha256": "1d30b31dc8352fc6a8d7b8abbef8118c840e47e5b90795dae0c758bf9717278c"
+      },
+      "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/references/source-notes.md": {
+        "bytes": 15234,
+        "sha256": "3d12a401e9decb55a2f4f3b2bfa0d360ef5110f63a0469dfe285654e78054190"
+      },
+      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md": {
+        "bytes": 10812,
+        "sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27"
+      },
+      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/metadata.json": {
+        "bytes": 572,
+        "sha256": "c221189b8d85fac3bd41ef1c7cd560f1aebc8cfa0381c182d1d571a7169be829"
+      },
+      "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/SKILL.md": {
+        "bytes": 3788,
+        "sha256": "bf1177cb9985c12a9bf11edb64771018e1e314c4713920de8fd4a989ff70115f"
+      },
+      "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/metadata.json": {
+        "bytes": 1299,
+        "sha256": "d86cecb615e93ba0f6be09d1ead102cd5a17068c07e1be2549ac5cd8f7519279"
+      },
+      "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/cache-stop.md": {
+        "bytes": 2026,
+        "sha256": "bba6b5c73261065309cbcf11b1e591eef48916f0c79ceaaf74d98a9fa96b0992"
+      },
+      "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/generation-batch.md": {
+        "bytes": 1733,
+        "sha256": "f16d3c57d45e63bcaffd3871aa41c76972ee11f7dbf1e12b45d9b170303cad9b"
+      },
+      "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/source-map.md": {
+        "bytes": 6358,
+        "sha256": "91d26181f246307f2113f94b12ac9587d689b654dcd216e9f14b11cf5ebb650c"
+      },
+      "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/validation.md": {
+        "bytes": 1665,
+        "sha256": "73b5dd347017baba519ecc85415aead9993c1751d116b9ba4f07689f58c5f783"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang-omni/LICENSE": {
+        "bytes": 11352,
+        "sha256": "a54c8553df5d4313d293d1aef3bf2f13952bda8b7c7a1fdd729b2bbe27ee6015"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang-omni/SKILL.md": {
+        "bytes": 2304,
+        "sha256": "d75f5b2881b240ec2d9d4f5cc8e3f7d53e4378cb01732a2fb9699da2901ba38c"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang-omni/references/repo-provenance.md": {
+        "bytes": 1778,
+        "sha256": "5e8dbe168a8885e209d4b4532944ffa7fead658b23fd6588054d57b22476f63b"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang-omni/references/repo-routing-metadata.json": {
+        "bytes": 511,
+        "sha256": "6820707cb9bd877d1635928cf5a0efe38e4fe37dc04036dc70d53e479ed9bf7c"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang-omni/references/snapshot.md": {
+        "bytes": 3488,
+        "sha256": "0fb5d776349de845dc314f16ce73a06c9caf5db84dc1c293327d3b3ca1337c80"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-architecture/SKILL.md": {
+        "bytes": 2355,
+        "sha256": "766b2d66d2056dd779abd9e492cb3af6e25899298f44964f565509f130a7cb24"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-architecture/references/runtime.md": {
+        "bytes": 7734,
+        "sha256": "36bba9874fe054acc5217f9c265f5ae309c83943e0d46228370ae1bcdc69421e"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-extension/SKILL.md": {
+        "bytes": 2193,
+        "sha256": "9af38bca840d0004fc2ed049141476fc40fd40dcfa843f38a404a7a314c19425"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-extension/references/integration.md": {
+        "bytes": 5775,
+        "sha256": "6ce9383ca17bacff60a180f8fb8c43ea47eaa37117fb8da6051ad86135d228ef"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-serving/SKILL.md": {
+        "bytes": 2335,
+        "sha256": "990b6bdebfe22612de49dbebcb16810e8abef94d70691aa883b7937c85210eb6"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-serving/references/recipes-and-tuning.md": {
+        "bytes": 8918,
+        "sha256": "94afb762bbe185f58606154b3f0a5af56e7089cb84ea94e26cfcb9f93c95439f"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang/LICENSE": {
+        "bytes": 11346,
+        "sha256": "1495e1e757ef4d0925a2350563cf5754bb23c51701a8ec4fb3c5cdcbedae6747"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang/SKILL.md": {
+        "bytes": 5466,
+        "sha256": "ae72d1f0cc750699bf79d7a138a97219209bca7b7e5f19cd427d306245eaf841"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/LICENSE": {
+        "bytes": 11346,
+        "sha256": "1495e1e757ef4d0925a2350563cf5754bb23c51701a8ec4fb3c5cdcbedae6747"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/SKILL.md": {
+        "bytes": 5738,
+        "sha256": "b0a4ba4751087126143ef86983798deb0ed00e24b6ff06bf6ef76a9df6e274f2"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/metadata.json": {
+        "bytes": 2844,
+        "sha256": "6ed57dfb4f25ecb34b81f2188b055e05d172ba2581e6a3b605b814b2f733a4d6"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/references/source-notes.md": {
+        "bytes": 19512,
+        "sha256": "d0fcfd1177ae2f351041c5b2abba7f5121cec133006d9e4dca8f74ddb567332d"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang/references/architecture-and-memory.md": {
+        "bytes": 7735,
+        "sha256": "798e747022b7ec135cc4861c2c9ac44fb14b503bb77a0068ee7fbea494b764a7"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang/references/native-candidates.md": {
+        "bytes": 3961,
+        "sha256": "abafcc9a34df196a6365145554887b8d37afb36c6e236c92e50a4a06c8a4ec93"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang/references/parallelism-and-extension.md": {
+        "bytes": 6324,
+        "sha256": "e3be7f4b51264ac68682a64d264e4d96857d5bb9f9d99cae5f327c56984a9e6b"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang/references/repo-provenance.md": {
+        "bytes": 2274,
+        "sha256": "534d95ae8d0b29bcbc8f6707b3d38fb964e31f5b66ba13517a9a6b8f309bbb38"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang/references/repo-routing-metadata.json": {
+        "bytes": 625,
+        "sha256": "b2b879c1cc42ddc2a5b95ae24115eb891562fa9a6e96d151f4ddf6f7719fe73c"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang/references/source-map.md": {
+        "bytes": 8945,
+        "sha256": "c2adad64ceb0c305cdc4859c03012b3e0a4ab26cd0cb289bbd76e192b0cc3794"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang/references/tuning-and-validation.md": {
+        "bytes": 10832,
+        "sha256": "49b3343b3292cc098a787de9d24d968b444d0ef4f73fedb5d8183390e8f39c9c"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-parallel-extension/SKILL.md": {
+        "bytes": 2316,
+        "sha256": "18bd54ebdd629e1ba6dfa589f391c7cc3c5355785c2ae82f741b64ee8022464f"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-runtime-memory/SKILL.md": {
+        "bytes": 2154,
+        "sha256": "33b1a65aba00fca819d2b6cc42202e7708e9155e4d8f4373a2387f4ee2f2fa3e"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-tuning-validation/SKILL.md": {
+        "bytes": 2018,
+        "sha256": "744b2ef9122101d3f50cbbb40fe020e2b3d3888adbe2d86041041e7c7ce424fc"
+      },
+      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE": {
+        "bytes": 1062,
+        "sha256": "b53fa08f515cb5a6fff7b9fd8fcd0961a4b80df29d74372d25e1f9171aa042ee"
+      },
+      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md": {
+        "bytes": 5504,
+        "sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6"
+      },
+      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json": {
+        "bytes": 2202,
+        "sha256": "f7b7581a3f44ba9774480c559408fc66d6a4d6d1dbbeabe4529f1eb7d4dd5ad6"
+      },
+      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md": {
+        "bytes": 8136,
+        "sha256": "7b537cb972347dd68e303960f142892346d8b19b884e4d08016ea36a23bdaa7a"
+      },
+      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json": {
+        "bytes": 3617,
+        "sha256": "af4830370b6eddc2d176a9ae14da4ab56a90b0fe048a01a89ee051df4365d8e9"
+      },
+      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md": {
+        "bytes": 6589,
+        "sha256": "b67affbd7d237d82b7f51be49f1537e9820fe56ec70a71ee76809f9e599ca693"
+      },
+      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md": {
+        "bytes": 3394,
+        "sha256": "5dcf5fa7bf734fbe13214f57d2dfb40ff8215d9cd34e072a9e615bf103b82fc2"
+      },
+      "src/inferrsi/data/skill_library/repositories/strata/LICENSE": {
+        "bytes": 1093,
+        "sha256": "949b9b0aada7476006469e58ee4ffb0e0f68c2d9137e90a356757b61a09d37f0"
+      },
+      "src/inferrsi/data/skill_library/repositories/strata/SKILL.md": {
+        "bytes": 2813,
+        "sha256": "0e5daeb6409c916be45ecd7ce2531ea5f1c60a4a53e2b0ed25dc1e2cdc5c17ed"
+      },
+      "src/inferrsi/data/skill_library/repositories/strata/references/design-space-analysis.md": {
+        "bytes": 27736,
+        "sha256": "11ecafe71530ddd4839268ea1b24b3361a3895012a705859ca06dae29fa72630"
+      },
+      "src/inferrsi/data/skill_library/repositories/strata/references/repo-provenance.md": {
+        "bytes": 2913,
+        "sha256": "05f17d0ae996ad6c12314bdd29f3427574f6aa611ea60effd18598a0f386638b"
+      },
+      "src/inferrsi/data/skill_library/repositories/strata/references/repo-routing-metadata.json": {
+        "bytes": 332,
+        "sha256": "a10ba49b7e6565109b566c105e17551d042075a1ff649c6f371134e3b7402c74"
+      },
+      "src/inferrsi/data/skill_library/repositories/strata/references/trial-format.md": {
+        "bytes": 2291,
+        "sha256": "92d8cf38549268ae4eef382890662d03e18a7dff858ad45ee5eb6f47eb39c8c0"
+      },
+      "src/inferrsi/data/skill_library/repositories/strata/references/troubleshooting.md": {
+        "bytes": 2079,
+        "sha256": "7301cfa5a5f6cfbf9576c3845a96c6b4de8f8a6509f2330970a8e2e4669b3d5a"
+      },
+      "src/inferrsi/data/skill_library/repositories/strata/scripts/summarize_trials.py": {
+        "bytes": 5170,
+        "sha256": "6a243a08da15659ced6519f49f52a49a675bd9e952740c3ab03dd21964eefbe2"
+      },
+      "src/inferrsi/data/skill_library/repositories/strata/sub-skills/design-space/SKILL.md": {
+        "bytes": 1826,
+        "sha256": "857bea3c64db45dc035728a41523142ac65778b41f63092148b8ab936f971d76"
+      },
+      "src/inferrsi/data/skill_library/repositories/strata/sub-skills/design-space/references/design-space.md": {
+        "bytes": 24231,
+        "sha256": "0a428dcd12d5290b554e9da6795a110f94d98f128771602e1ea5f608048b135e"
+      },
+      "src/inferrsi/data/skill_library/repositories/strata/sub-skills/tune-execution/SKILL.md": {
+        "bytes": 1145,
+        "sha256": "cdf51c537e7237296d52ef6cda9f2e4c9d9ce3fc87265a369cd9eb7497c53512"
+      },
+      "src/inferrsi/data/skill_library/repositories/strata/sub-skills/tune-execution/references/tuning.md": {
+        "bytes": 11310,
+        "sha256": "7b69aa7872d60dd3fb91e41b77a632572070dcf80dd5b5f6dc1722d8540177b1"
+      },
+      "src/inferrsi/data/skill_library/repositories/strata/sub-skills/validate-engine/SKILL.md": {
+        "bytes": 4613,
+        "sha256": "90bfd975e25f18895ce0e85f6f35d24149ce026305a4d479aea4924510053460"
+      },
+      "src/inferrsi/data/skill_library/repositories/strata/sub-skills/validate-engine/references/validation.md": {
+        "bytes": 16895,
+        "sha256": "dae846f4c47c68731b96ec1d50cb008c03871c72078f1c5b0f4a9423be9ba397"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm-omni/LICENSE": {
+        "bytes": 11357,
+        "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm-omni/SKILL.md": {
+        "bytes": 4945,
+        "sha256": "df3e63dc28de2dac2aa92824fd88719159138650d96487332e0effca3f0e2a80"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm-omni/references/evidence.md": {
+        "bytes": 4689,
+        "sha256": "ae0d51ba56b34ed81fba34de8fd3685bebf2f160b188fa53719936002881d220"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm-omni/references/repo-provenance.md": {
+        "bytes": 2221,
+        "sha256": "43bbfe431434ca4ad6157ba35cbb9ce2d89e4065f427abfdd5f372f51ba0c9c9"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm-omni/references/repo-routing-metadata.json": {
+        "bytes": 342,
+        "sha256": "7784b4d98d096718c4f4cc42ce86d5ec113c469844e12de4690b0fccca59d329"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-architecture/SKILL.md": {
+        "bytes": 1336,
+        "sha256": "fb3ff50a7b9122a2545a97ef5179011237c4c4ca214dde7236ad7bf79a7c8623"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-architecture/references/architecture.md": {
+        "bytes": 5600,
+        "sha256": "db845eb4207ddeedb8ae381f1b91704a42cb53ba0ee6c7ee581510ba1a412190"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-serving/SKILL.md": {
+        "bytes": 1343,
+        "sha256": "8ab3209b838635395b22deca6aa879172016531d45c2374eb8e498e7f71bb40f"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-serving/references/serving.md": {
+        "bytes": 5049,
+        "sha256": "9bd5eff300d9755137a7600c8f7a7753bb22d09c8853e08062d7e535662ba8a3"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/SKILL.md": {
+        "bytes": 2227,
+        "sha256": "93c83bc4b799b32f50ff54b870cbbde050a790af16d9b7b062039c4987c80148"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/references/abort-acknowledgment.md": {
+        "bytes": 8547,
+        "sha256": "e95ae5b2cc410917ddf5e9066890d0bef9264104529bdca5ac12bdf56fe80ae3"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/references/validation.md": {
+        "bytes": 4632,
+        "sha256": "1542061f9c76fff2acb6529a58238301484b0e23ec0c697d2fdf129e200ea973"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/LICENSE": {
+        "bytes": 11324,
+        "sha256": "50e6751797c50dedd75ef1b8a0d9e42f5f8472e9fbce91f34718e9f97b0c780a"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/SKILL.md": {
+        "bytes": 2863,
+        "sha256": "38e522b8abeb0dc140a2194340432ed41dc8b569d33b63d5d4c59aa1092cd2d1"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/SKILL.md": {
+        "bytes": 6664,
+        "sha256": "313733a65cc48442b18dda1c4665139a689963a90e1032a2235bc6f40e3a8e06"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/metadata.json": {
+        "bytes": 781,
+        "sha256": "b43fc3a7ccae59c664fcdf9ff77962edcc07c2a3df8b59324db551642bd23e23"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/references/source-notes.md": {
+        "bytes": 13853,
+        "sha256": "3665b034594fc8cc767c408002f4e16f8803369242405027c7f37ce6c9ecebe9"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/SKILL.md": {
+        "bytes": 6565,
+        "sha256": "5fb9ea1d1373e110d9b506f69963c4857338b7dbf5e38ec9a49b59a0ff683ed0"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/metadata.json": {
+        "bytes": 1372,
+        "sha256": "78459416734f61a8faa819d9a1a3ad7b9a985e8bf8e149813c604d15be6bb92f"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/INTEGRATION_NOTES.md": {
+        "bytes": 9006,
+        "sha256": "82b24be82451ee121892f61575d0cf4ecb9920813a25ccb26cce2b45dda6d8b3"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/migration-is-not-free-memory.md": {
+        "bytes": 2145,
+        "sha256": "2ac168f2725cba7fc92dfcc67c01bf2258dff6a2dbe1ddb063bc60c175d82d0a"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/parallelism-change-with-state-cost.md": {
+        "bytes": 2069,
+        "sha256": "5accc81086e4500d66f894ef6af9951e1cbb76941f6c631aab5c5561a786d8d8"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/preemption-under-kv-pressure.md": {
+        "bytes": 2630,
+        "sha256": "462200a6d0b193ef01f32374a463e0f2838ca8edf33d9ab898def65ffd0b490a"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/knowledge/scheduling-state-boundaries.md": {
+        "bytes": 7488,
+        "sha256": "d4ef29869959b835a5dcb06ddb2fd8e6cff49bb176994dc1d46eb9a852f6228d"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/fastserve.md": {
+        "bytes": 9289,
+        "sha256": "9b4e275843000ef453d508d9c3c9c3f8ea1de5545fbd01d7abe74baa8a4931a7"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/llumnix.md": {
+        "bytes": 10181,
+        "sha256": "f252ce33ac2db017701ca1d49813c5ec1996c4c92900cf35bb6bda7fc59c0f18"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/loongserve.md": {
+        "bytes": 12422,
+        "sha256": "7bc6d1e256d461b04acc8ed856cf99108faceda9cde729ab95bcca596fe019ca"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/orca.md": {
+        "bytes": 8826,
+        "sha256": "f5a3b5870d64226408a8fcda5d69483769d823f8a5efd7f11698c8da483ab195"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/fastserve-readset.md": {
+        "bytes": 7173,
+        "sha256": "9866bd0c78b2b638998b3b632f6bae26aef032462c81abee336d309af5b07831"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/llumnix-readset.md": {
+        "bytes": 9386,
+        "sha256": "3a2819d65ad24087c08530492ef39d8ea046ed1f780d53512f1d23e2b3578487"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/loongserve-readset.md": {
+        "bytes": 12490,
+        "sha256": "f1a383e7bdbbd5a5a472d7f6e1292c8a17d4910ff736ca5a6aa422e1bee74f8c"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/orca-readset.md": {
+        "bytes": 4796,
+        "sha256": "a497ef23b7e60c0fbc15d95b7e07cd20547cfe56f37c24dfe6ed70d79ef7a3cf"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/source-notes.md": {
+        "bytes": 9575,
+        "sha256": "84033fd514773f10c0bddd9914fabb980497ba2b308f226a922c3e1888413827"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/references/evidence-map.json": {
+        "bytes": 14384,
+        "sha256": "dd8bbab5d285ed3a1a19d2f06ad84e8c444251690d8f673def3e2681765ef6c1"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/references/repo-provenance.json": {
+        "bytes": 1313,
+        "sha256": "4e705a964697712e950e02bedc41320806a41311b7f4dcf32f923d60b1d2e919"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/references/repo-provenance.md": {
+        "bytes": 3261,
+        "sha256": "6e841c38c35ff2cd6ff8139cb1a6c8f75701647897f1089a94a5c62928609941"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/references/repo-routing-metadata.json": {
+        "bytes": 432,
+        "sha256": "a20b54f682eb52311e445ac62b95d1935c001296f5cf23b7764cb04477d535c9"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/SKILL.md": {
+        "bytes": 1835,
+        "sha256": "2a945b5fddacfb90d1cd7cb2b2f767067d641061423a65e1c19afecec7f3c06a"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/references/capability-boundaries.md": {
+        "bytes": 6560,
+        "sha256": "c8081d03f04f7481080c46c7dcf728962729bc122f49f9502d50cd1e3d04bcf3"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/references/runtime-and-invariants.md": {
+        "bytes": 8265,
+        "sha256": "afa1f967545021f36bc1b59dbf6aa2c367f4d5ece4f08dce755e2017cf50dbe5"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/extend-validate/SKILL.md": {
+        "bytes": 1731,
+        "sha256": "495679e5b99341bf80daeaacd2d4c2699aba16bc16c447cbd8d2fe17f7559798"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/extend-validate/references/change-boundaries-and-validation.md": {
+        "bytes": 7624,
+        "sha256": "f7fdd2d699f09d96fbf5ca9382f9b56113c870db2688d8e1118eb8163519fc4a"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/SKILL.md": {
+        "bytes": 1812,
+        "sha256": "0ef4d2e3a96cbb277557e169a3fc8edf0f0504aad03259771b432a33ee008f78"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/references/commands-and-measurements.md": {
+        "bytes": 5667,
+        "sha256": "76ab51083b14595c49fa1c255cb7e5ebb996f14930a51ab3cbe864cc334b5f80"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/references/experiment-decisions.md": {
+        "bytes": 7644,
+        "sha256": "242593c0d386e57104b57a09bc7db97c2bbbcdb905bb68a397cfe612e07fb083"
+      },
+      "src/inferrsi/data/skill_library/skill-index.json": {
+        "bytes": 7514,
+        "sha256": "8ce35e18f7629260d67a95617b9918eecff73ed8d9370dd3d4595c725ada9aa1"
+      },
+      "src/inferrsi/data/skill_library/workflows/cite-check-downstream-pr/SKILL.md": {
+        "bytes": 1093,
+        "sha256": "485932cdf3b1fba5dc69be73e75d38f86a7b79442d26e82a296c963a7598d226"
+      },
+      "src/inferrsi/data/skill_library/workflows/claim-contract/SKILL.md": {
+        "bytes": 1075,
+        "sha256": "3af1dbbe81d91cb05fc2ef15a2ae25d7592a08558e1a5535f49c52f33a488c55"
+      },
+      "src/inferrsi/data/skill_library/workflows/competitor-scan/SKILL.md": {
+        "bytes": 1088,
+        "sha256": "0ede58ace65a9324a538b4816add1684f230ba536d4364fe5a5dd2c0bbed3776"
+      },
+      "src/inferrsi/data/skill_library/workflows/draft-acceptance-issue/SKILL.md": {
+        "bytes": 1230,
+        "sha256": "ce983acfb7d7d3272775e7cf00c92e4ddadd97511b602575a76e5ab7ade01b56"
+      },
+      "src/inferrsi/data/skill_library/workflows/inference-engine-research/SKILL.md": {
+        "bytes": 5681,
+        "sha256": "364beed2ca617e71f4f955bbd024740d411b94eeb9557d7e055d4463cc22caba"
+      },
+      "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/cross-repository-mechanisms.md": {
+        "bytes": 17050,
+        "sha256": "28efa1b25f548763180b70d8d0ddebb7d4e7227df189f871fd233fe1d4881303"
+      },
+      "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/import.md": {
+        "bytes": 3463,
+        "sha256": "c8a57db94768cc6cd4193932def75ba67ec453e56ee0e54f176fb82f1af53091"
+      },
+      "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/model-aware-preparation.md": {
+        "bytes": 9056,
+        "sha256": "787c017aade68ee1133b979d636fa24062302cef6174c001171c963947ee8da4"
+      },
+      "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/overview.md": {
+        "bytes": 8099,
+        "sha256": "7ed0a966239b9481b82dd349245cfbfe966c030fc916771c91fd0692464ffc56"
+      },
+      "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/repository-source-scope.md": {
+        "bytes": 11967,
+        "sha256": "85202ca209171c50573c2fc209acb1e937d8211e1652e4ca9c6cc4bce04d03eb"
+      },
+      "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/source-manifest.json": {
+        "bytes": 1613,
+        "sha256": "0ce1678803d1ede6b485fb6b4b1c88aba909056ee6e3b5af691899ceb2056240"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/SKILL.md": {
+        "bytes": 6502,
+        "sha256": "74cd863df2e80dc2bda1713d24207809173a3116f30325e3487e1e3517cfec1e"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/metadata.json": {
+        "bytes": 3236,
+        "sha256": "12fc9272bd5fbf02dd9b897d8ca40984b72ae54f20492ace5304711bbc4d71c3"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/COMPARISON.md": {
+        "bytes": 4267,
+        "sha256": "38587678a96581f5c2dc697d982127343565504bffa191352811dffcc165aa23"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/INTEGRATION_NOTES.md": {
+        "bytes": 5557,
+        "sha256": "fbc4a96b6f6c477706eed83eaa4fcc29a14f881637629d6c004601ed049f2523"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/QA.md": {
+        "bytes": 1853,
+        "sha256": "3f3785ec39a6938a5d1f2e874e99153fe92c0e666dd3e78a499c8b47d891346a"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/README.md": {
+        "bytes": 2151,
+        "sha256": "221785e91c45588f82b3caab6d52f1ea1d102b16139c20a0802921e2789e6d28"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/UNREAD_LEADS.md": {
+        "bytes": 1038,
+        "sha256": "6c1d87ebe491d27666703149eb05ad582c7e5aff77f0ac5c11a7b89c946f96ee"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/cases/methodology-counterexamples.md": {
+        "bytes": 5213,
+        "sha256": "f9fc21753ea41eef77ca020d78d031e945e5dd835b5d09c5ff0e53749ceb2ca4"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/knowledge/evidence-routing.md": {
+        "bytes": 5282,
+        "sha256": "406d5d74c3fef4c1a23d495492e21a084c3c46495a1a84b385703ed104358874"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/learning-to-commit.md": {
+        "bytes": 2847,
+        "sha256": "9c2d1d738e896d05c00f53e6adf14c9df2501ecc3f0020e71b5a1c2192547086"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/protocol-details.md": {
+        "bytes": 6520,
+        "sha256": "41298c6bde3a28f0712fa06a18bbc4ac59aea6d17cc9eb1318744f4121e3bab6"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/readset.md": {
+        "bytes": 7086,
+        "sha256": "6f953d20d8958c68339c9ceeae348604d0fea3e9c6ccb9b64a63d3b210426a22"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/repo-to-skill.md": {
+        "bytes": 3151,
+        "sha256": "5ebdac2a2bbbe6614f9d4096bc9895bd685acdf76414e2c3c1fb7b50d57c738d"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/skillopt.md": {
+        "bytes": 3040,
+        "sha256": "148075daa67e19f4b4351710498438d17c279a9bce56b2d01cab26ff3be33e99"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/source-manifest.json": {
+        "bytes": 2431,
+        "sha256": "716cd49ed01824ee9f6b6fc55ebf2b24fdd5d5a3e677f6e28e34a472616bbe25"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/trace2skill.md": {
+        "bytes": 2917,
+        "sha256": "230d13b77ba819b5e3edee147523e8cad6bfce95afccc140fb4946d6c8747a2e"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/updates/bounded-updates.md": {
+        "bytes": 7630,
+        "sha256": "d02c5bd0109846d1e1987de074810321cb4af277cd3f66ab2ac268aadd5ee142"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/skill-factory-design-20261014.md": {
+        "bytes": 9465,
+        "sha256": "fb8ce1ad83b7a6ddc657e89ee2eebb710873b213cfd85bdd6d823c8691b3cb9f"
+      },
+      "src/inferrsi/data/skill_library/workflows/rsi-evidence/SKILL.md": {
+        "bytes": 3680,
+        "sha256": "a55610874b658b86a39a1a049efb4e4eab0140f6579ed74dd7718a4ca8cedaca"
+      }
+    },
+    "sha256": "6e19365a0ad8e2951af0f3106f93aa545b4ad00b6e0f0317a20f32065b6a05fb"
+  },
+  "environment": {
+    "filesystem_cache": "not controlled; process cold is not OS-cache cold",
+    "implementation": "CPython",
+    "logical_cpu_count": 8,
+    "machine": "AMD64",
+    "perf_counter_resolution_seconds": 1e-07,
+    "python_executable_sha256": "737a7e3b71e3578f8432acc7dd88c452e593622c544bc13da4789d69c63da5ae",
+    "python_optimization": 0,
+    "python_version": "3.12.6",
+    "release": "10",
+    "site_initializers_disabled": true,
+    "system": "Windows"
+  },
+  "git_head": "407bb244651bec1c68f4b3e94b0c697b233c4a09",
+  "knowledge_record_count": 108,
+  "library": {
+    "bytes": 2294958,
+    "file_count": 342,
+    "sha256": "be8726296e6186601692dc2d95d99bfbba76f0b4c758de3a88a9b2a523eaa37e"
+  },
+  "modules": {
+    "inferrsi": {
+      "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+      "source": "src/inferrsi/__init__.py"
+    },
+    "inferrsi.catalog": {
+      "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+      "source": "src/inferrsi/catalog.py"
+    },
+    "inferrsi.cli": {
+      "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+      "source": "src/inferrsi/cli.py"
+    },
+    "inferrsi.knowledge_operations": {
+      "sha256": "c71bb2063ab8d919ca5907ac900a3930ec60730515c614fee7375d9efb77173a",
+      "source": "src/inferrsi/knowledge_operations.py"
+    }
+  },
+  "packaged_skill_count": 60,
+  "python_sources": {
+    "bytes": 274624,
+    "file_count": 53,
+    "files": {
+      "src/inferrsi/__init__.py": {
+        "bytes": 102,
+        "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47"
+      },
+      "src/inferrsi/__main__.py": {
+        "bytes": 48,
+        "sha256": "935a1c1166b0c1ea35a82256345000bf2c73ded718d77773bc27a71ecce28f7d"
+      },
+      "src/inferrsi/_skill_selection.py": {
+        "bytes": 5790,
+        "sha256": "4640399f4013425589f6702e2db844820619ba35967d1c3519978eb307168440"
+      },
+      "src/inferrsi/arguments.py": {
+        "bytes": 12200,
+        "sha256": "4456600cf4a362cf069594fe36db46932cf5cb14ddd37683aaf3538f5cf76b4d"
+      },
+      "src/inferrsi/catalog.py": {
+        "bytes": 6855,
+        "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5"
+      },
+      "src/inferrsi/claim.py": {
+        "bytes": 805,
+        "sha256": "5f24c4fa91b48f0bb4d4554531607763fc8fc108e90f332816f06cfb520b5cc1"
+      },
+      "src/inferrsi/cli.py": {
+        "bytes": 14205,
+        "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88"
+      },
+      "src/inferrsi/contracts.py": {
+        "bytes": 119,
+        "sha256": "03000cb9246ce8c0967147533f7ad25df918ba2674839801453ffb870f496490"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/scripts/roofline.py": {
+        "bytes": 9004,
+        "sha256": "524de48a6764401695fe1444f1fa3d38f363069b0cd2fba4715a86ca4b28a5d4"
+      },
+      "src/inferrsi/data/skill_library/repositories/strata/scripts/summarize_trials.py": {
+        "bytes": 5170,
+        "sha256": "6a243a08da15659ced6519f49f52a49a675bd9e952740c3ab03dd21964eefbe2"
+      },
+      "src/inferrsi/evaluation.py": {
+        "bytes": 183,
+        "sha256": "482c40710d428206c59fe3696703248cd86c4b3c7d19ad41017c1369bd6d780e"
+      },
+      "src/inferrsi/evidence_paths.py": {
+        "bytes": 859,
+        "sha256": "cc46231e0b8334ef5742defdda20e0093e22b30c86af59df1fb3ae4b5caa70fa"
+      },
+      "src/inferrsi/execution.py": {
+        "bytes": 755,
+        "sha256": "f43d8cbb4d43a72a7ff8a10b29a5fb9cf4cb22af54dd35943283c843fba0da92"
+      },
+      "src/inferrsi/harness/__init__.py": {
+        "bytes": 177,
+        "sha256": "1006e4713003f44bfb32b2e385a918e1652b9e9833801e4555674a71e57fb0d5"
+      },
+      "src/inferrsi/harness/evaluation.py": {
+        "bytes": 4160,
+        "sha256": "9b00b2e93959f16b7b76bb4b257df7dcf48520a06fd6adae96e9a165a05e6e80"
+      },
+      "src/inferrsi/harness/improvement.py": {
+        "bytes": 5777,
+        "sha256": "3faa0f5928a707a98f35f45d1ca18e226c898941e6a07b3829eebeb53e70aeb8"
+      },
+      "src/inferrsi/harness/reporting.py": {
+        "bytes": 3566,
+        "sha256": "38c58f02c44e8c8aeacf46527e142be0e3142a83e4b2693246d0ca76c0ddd7c3"
+      },
+      "src/inferrsi/harness/runner.py": {
+        "bytes": 6567,
+        "sha256": "a566307b9b586655941b627f428b5121ed38d7ac80bccd1e1799e443fb0de911"
+      },
+      "src/inferrsi/harness/scenarios/__init__.py": {
+        "bytes": 79,
+        "sha256": "4e97913c806077f0cadfeb29d986d2fa75b55ab8766cb86e6e3e9a195a9a1f5a"
+      },
+      "src/inferrsi/harness/scenarios/external.py": {
+        "bytes": 1016,
+        "sha256": "84d5f6920501a9d882cc4fffbc721d2051fdbf7c90694bde9cbfba3c2b866304"
+      },
+      "src/inferrsi/harness/scenarios/media.py": {
+        "bytes": 2914,
+        "sha256": "67d979a3ed6f13a8e3e1e40a304f78e5a3e6c2556b3aa8a11c41156745d7050a"
+      },
+      "src/inferrsi/harness/skill_evaluation.py": {
+        "bytes": 11093,
+        "sha256": "b710b8e53758757efd68c703449eb4bab7eb7ec50ae3dfc52a010cdebd8c3bf5"
+      },
+      "src/inferrsi/infra/__init__.py": {
+        "bytes": 85,
+        "sha256": "8779897733779266841b0f79b84b1dc43f0fa57c013855c906c809829ecfc77c"
+      },
+      "src/inferrsi/infra/artifacts.py": {
+        "bytes": 1040,
+        "sha256": "e72e1bebdbb88e42de430238489252386851f9e6645ed7780f2aad2e90338120"
+      },
+      "src/inferrsi/infra/campaign.py": {
+        "bytes": 1028,
+        "sha256": "be505fba1592611bd84ceaca55afbac027205413a2639baefa0c3bb343230209"
+      },
+      "src/inferrsi/infra/commands.py": {
+        "bytes": 811,
+        "sha256": "1890f115b3a2c5c6832173bf60432ad595bd6bf1c69783ecd460c6f8961194f1"
+      },
+      "src/inferrsi/infra/environment.py": {
+        "bytes": 2067,
+        "sha256": "60e46c2aefc216db202b41f44542e7b893e08e0721c667b9ccc50411d1ac84ee"
+      },
+      "src/inferrsi/infra/executors/__init__.py": {
+        "bytes": 64,
+        "sha256": "327c3cd5c821c832ff260a41dba0ee1e439cdabefc9edfb9cad1cd38d14bc87a"
+      },
+      "src/inferrsi/infra/executors/local.py": {
+        "bytes": 3654,
+        "sha256": "37bcbca9f3261c8c58a8bf806c9da40720b18d0676c82e8ed24c3901ac9bbe77"
+      },
+      "src/inferrsi/infra/http.py": {
+        "bytes": 634,
+        "sha256": "f65d6c2b55b6a89694caaf6dcdf7d74f417e9ced103eca1b3f218e305c6b6cb2"
+      },
+      "src/inferrsi/infra/inspection.py": {
+        "bytes": 2415,
+        "sha256": "d1c66deb0cced529bca996d2f88b60f51fa3417ff1cae365c650fec808946ecd"
+      },
+      "src/inferrsi/infra/platforms/__init__.py": {
+        "bytes": 1494,
+        "sha256": "9e88e8b9a396ed90c221e1585135c0a1b58da41c1bb73ed9237c53b6684abfed"
+      },
+      "src/inferrsi/infra/platforms/cuda.py": {
+        "bytes": 3309,
+        "sha256": "9e0e4276e1b2984605493b14b82fba3e86f6635204719c318302e624f66f8b6f"
+      },
+      "src/inferrsi/infra/platforms/mlx.py": {
+        "bytes": 1971,
+        "sha256": "f3b81470db28bd6183ccca2e763c5e079686bba6cdfe6c08ebe2cdf64216f2e3"
+      },
+      "src/inferrsi/infra/processes.py": {
+        "bytes": 8151,
+        "sha256": "3b96ebc7a50913e78e86ff3dcba74c8ee8d03d547d9591ca6488f390ac1e6189"
+      },
+      "src/inferrsi/infra/targets/__init__.py": {
+        "bytes": 186,
+        "sha256": "1f3ea75dded7c6cc967754c7b6d4f30ef00da8191a749938aabc3c6bb8f86eb3"
+      },
+      "src/inferrsi/infra/targets/git.py": {
+        "bytes": 2910,
+        "sha256": "6327c4f126086f310b6de309fc7659c9759da28ed76efaeeb94a3d2fba29a0c2"
+      },
+      "src/inferrsi/inspection.py": {
+        "bytes": 692,
+        "sha256": "afa44bf4cdfbb58a318c21f9b292ca142d1bbb22750e35a4cc8244da5350ebb3"
+      },
+      "src/inferrsi/knowledge_operations.py": {
+        "bytes": 27624,
+        "sha256": "c71bb2063ab8d919ca5907ac900a3930ec60730515c614fee7375d9efb77173a"
+      },
+      "src/inferrsi/knowledge_store.py": {
+        "bytes": 13700,
+        "sha256": "5e794e28906ce4687aff7f14a507e267184640fc8a4fac19e0ced279ce6a6cb6"
+      },
+      "src/inferrsi/model/__init__.py": {
+        "bytes": 77,
+        "sha256": "3c1dba4f5159fd9fbc8d4d96e2a83c8441491c1b1dc7b42031a3fc1782ed9d9f"
+      },
+      "src/inferrsi/model/anthropic_agent.py": {
+        "bytes": 24696,
+        "sha256": "b0ef9cd542544292eeb90ab05ef410be9ed6e5d1a40d729890f49562c28c12f0"
+      },
+      "src/inferrsi/model/protocols/__init__.py": {
+        "bytes": 85,
+        "sha256": "5a2f54483b164de04e309dbebb87a700b213682b0f44a50d89a752dd71a4ce47"
+      },
+      "src/inferrsi/model/protocols/image.py": {
+        "bytes": 2047,
+        "sha256": "251a1f8f885da2cd8e9d3baba766c525afe293091606b4e8668c43ba2cb7cb5f"
+      },
+      "src/inferrsi/model/protocols/metrics.py": {
+        "bytes": 610,
+        "sha256": "17db16d6ef15b41fb60d8f09507893accc43f9cb3ad686b8dc4e9839713ebc56"
+      },
+      "src/inferrsi/model/validators/__init__.py": {
+        "bytes": 65,
+        "sha256": "1d4b6ffa968e283219ddcc8a09de9280238a8d65ae2dc8cd2b60889877a8eb0b"
+      },
+      "src/inferrsi/model/validators/flare.py": {
+        "bytes": 19975,
+        "sha256": "3696b0609c772f2222b448833608565a033123d7e92b11fbc545dd356fec590f"
+      },
+      "src/inferrsi/model/validators/image.py": {
+        "bytes": 3204,
+        "sha256": "79470b6b06a22d53ec6c40636a0b3d821a2e6ee2f7bb7aed3dac7241504add1a"
+      },
+      "src/inferrsi/reporting.py": {
+        "bytes": 127,
+        "sha256": "1f771890ab3e69b1aa9246b548a7825596bdf904f6c395dd218208084266853d"
+      },
+      "src/inferrsi/research.py": {
+        "bytes": 5213,
+        "sha256": "9bfe59dd17b5476c18ce50d671711053c62800521c50d6b710e8140a2b4dca04"
+      },
+      "src/inferrsi/runspec.py": {
+        "bytes": 4398,
+        "sha256": "dca0952425caf42095ef0c7e2feb8ef4740cc422ecb61e777f203962d9b87709"
+      },
+      "src/inferrsi/skill_catalog.py": {
+        "bytes": 9938,
+        "sha256": "65f5a1f3f6894009abb98ec2fec30a63cc1bb2f6ce6aeb01f9f7a3dce0439f2d"
+      },
+      "src/inferrsi/summary_chart.py": {
+        "bytes": 40910,
+        "sha256": "d29023d60f6be804c1b01c9eb73765d4824d6c3bb870fe22c99a24544bc1d8bc"
+      }
+    },
+    "sha256": "d6edbaf975cf3bc00902d6aa7cf2bb955dc58b3e37c74f59bb226c5f0d2fa261"
+  },
+  "tracked_source_dirty": false
+}
diff --git a/evidence/knowledge-retrieval-20261027/results/context-delivery-diagnostic.json b/evidence/knowledge-retrieval-20261027/results/context-delivery-diagnostic.json
new file mode 100644
index 00000000..555418e3
--- /dev/null
+++ b/evidence/knowledge-retrieval-20261027/results/context-delivery-diagnostic.json
@@ -0,0 +1,473 @@
+{
+  "schema_version": 1,
+  "source": "quality.json",
+  "source_sha256": "5263406bb621b34ce7cdec67f0bd3719197f6642496eb419abca3e792e9ee836",
+  "analysis": "Post-hoc diagnostic, not a frozen gate or change to item-ranking metrics.",
+  "definition": "At least one returned contexts[].ref matches an independently labelled package identity. API response presence only; agent consumption was not measured.",
+  "summary": {
+    "A": {
+      "answerable": 20,
+      "labelled_owner_context_present": 19
+    },
+    "B": {
+      "answerable": 20,
+      "labelled_owner_context_present": 19
+    }
+  },
+  "per_case": [
+    {
+      "side": "A",
+      "case": "dev01",
+      "labelled_refs": [
+        "package:inferrsi.common.skills.dcp_async_save_performance"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 12636,
+      "total_context_bytes": 64465,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "A",
+      "case": "dev02",
+      "labelled_refs": [
+        "package:inferrsi.common.skills.dcp_async_save_performance"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 12636,
+      "total_context_bytes": 12636,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "A",
+      "case": "dev03",
+      "labelled_refs": [
+        "package:inferrsi.common.skills.weight_loading_optimization"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 41017,
+      "total_context_bytes": 62379,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "A",
+      "case": "dev04",
+      "labelled_refs": [
+        "package:inferrsi.common.skills.weight_loading_optimization"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 41017,
+      "total_context_bytes": 63529,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "A",
+      "case": "dev05",
+      "labelled_refs": [
+        "package:inferrsi.common.skills.storage_io_measurement"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 8813,
+      "total_context_bytes": 51533,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "A",
+      "case": "dev06",
+      "labelled_refs": [
+        "package:inferrsi.common.skills.storage_io_measurement"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 8813,
+      "total_context_bytes": 35244,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "A",
+      "case": "dev07",
+      "labelled_refs": [
+        "package:inferrsi.ascend.skills.ascendc_tiling_design"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 11860,
+      "total_context_bytes": 64663,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "A",
+      "case": "dev08",
+      "labelled_refs": [
+        "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 17459,
+      "total_context_bytes": 17459,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "A",
+      "case": "dev09",
+      "labelled_refs": [
+        "package:inferrsi.common.skills.kv_tier_management"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 10812,
+      "total_context_bytes": 61835,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "A",
+      "case": "dev10",
+      "labelled_refs": [
+        "package:inferrsi.common.skills.checkpoint_recovery_consistency"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 14400,
+      "total_context_bytes": 14400,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "A",
+      "case": "dev11",
+      "labelled_refs": [
+        "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 56835,
+      "total_context_bytes": 56835,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "A",
+      "case": "dev12",
+      "labelled_refs": [
+        "package:inferrsi.common.skills.os_process_placement"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 16944,
+      "total_context_bytes": 65309,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "A",
+      "case": "dev13",
+      "labelled_refs": [
+        "package:inferrsi.common.skills.plan_activation_residency"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 17414,
+      "total_context_bytes": 45716,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "A",
+      "case": "dev14",
+      "labelled_refs": [
+        "package:inferrsi.common.skills.claim_contract"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 1075,
+      "total_context_bytes": 40799,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "A",
+      "case": "withheld01",
+      "labelled_refs": [
+        "package:inferrsi.common.skills.training_input_pipeline"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 9502,
+      "total_context_bytes": 9502,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "A",
+      "case": "withheld02",
+      "labelled_refs": [
+        "package:inferrsi.common.skills.calibrate_serving_simulation"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 12299,
+      "total_context_bytes": 62288,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "A",
+      "case": "withheld03",
+      "labelled_refs": [
+        "package:inferrsi.common.skills.assess_serving_preemption"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 36807,
+      "total_context_bytes": 47619,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "A",
+      "case": "withheld04",
+      "labelled_refs": [
+        "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 19749,
+      "total_context_bytes": 63132,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "A",
+      "case": "withheld05",
+      "labelled_refs": [
+        "package:inferrsi.common.skills.evaluate_elastic_serving_placement"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 51463,
+      "total_context_bytes": 51463,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "A",
+      "case": "withheld06",
+      "labelled_refs": [
+        "package:inferrsi.common.skills.performance_ceiling_analysis"
+      ],
+      "labelled_owner_context_present": false,
+      "labelled_context_bytes": 0,
+      "total_context_bytes": 51702,
+      "labelled_context_omissions": [
+        {
+          "available_bytes": 65536,
+          "reason": "context_limit",
+          "ref": "package:inferrsi.common.skills.performance_ceiling_analysis",
+          "required_bytes": 170782
+        }
+      ]
+    },
+    {
+      "side": "B",
+      "case": "dev01",
+      "labelled_refs": [
+        "package:inferrsi.common.skills.dcp_async_save_performance"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 12636,
+      "total_context_bytes": 64465,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "B",
+      "case": "dev02",
+      "labelled_refs": [
+        "package:inferrsi.common.skills.dcp_async_save_performance"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 12636,
+      "total_context_bytes": 12636,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "B",
+      "case": "dev03",
+      "labelled_refs": [
+        "package:inferrsi.common.skills.weight_loading_optimization"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 41017,
+      "total_context_bytes": 62379,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "B",
+      "case": "dev04",
+      "labelled_refs": [
+        "package:inferrsi.common.skills.weight_loading_optimization"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 41017,
+      "total_context_bytes": 63529,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "B",
+      "case": "dev05",
+      "labelled_refs": [
+        "package:inferrsi.common.skills.storage_io_measurement"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 8813,
+      "total_context_bytes": 51533,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "B",
+      "case": "dev06",
+      "labelled_refs": [
+        "package:inferrsi.common.skills.storage_io_measurement"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 8813,
+      "total_context_bytes": 35244,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "B",
+      "case": "dev07",
+      "labelled_refs": [
+        "package:inferrsi.ascend.skills.ascendc_tiling_design"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 11860,
+      "total_context_bytes": 64663,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "B",
+      "case": "dev08",
+      "labelled_refs": [
+        "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 17459,
+      "total_context_bytes": 17459,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "B",
+      "case": "dev09",
+      "labelled_refs": [
+        "package:inferrsi.common.skills.kv_tier_management"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 10812,
+      "total_context_bytes": 61835,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "B",
+      "case": "dev10",
+      "labelled_refs": [
+        "package:inferrsi.common.skills.checkpoint_recovery_consistency"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 14400,
+      "total_context_bytes": 14400,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "B",
+      "case": "dev11",
+      "labelled_refs": [
+        "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 56835,
+      "total_context_bytes": 56835,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "B",
+      "case": "dev12",
+      "labelled_refs": [
+        "package:inferrsi.common.skills.os_process_placement"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 16944,
+      "total_context_bytes": 65309,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "B",
+      "case": "dev13",
+      "labelled_refs": [
+        "package:inferrsi.common.skills.plan_activation_residency"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 17414,
+      "total_context_bytes": 45716,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "B",
+      "case": "dev14",
+      "labelled_refs": [
+        "package:inferrsi.common.skills.claim_contract"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 1075,
+      "total_context_bytes": 40799,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "B",
+      "case": "withheld01",
+      "labelled_refs": [
+        "package:inferrsi.common.skills.training_input_pipeline"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 9502,
+      "total_context_bytes": 9502,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "B",
+      "case": "withheld02",
+      "labelled_refs": [
+        "package:inferrsi.common.skills.calibrate_serving_simulation"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 12299,
+      "total_context_bytes": 62288,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "B",
+      "case": "withheld03",
+      "labelled_refs": [
+        "package:inferrsi.common.skills.assess_serving_preemption"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 36807,
+      "total_context_bytes": 47619,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "B",
+      "case": "withheld04",
+      "labelled_refs": [
+        "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 19749,
+      "total_context_bytes": 63132,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "B",
+      "case": "withheld05",
+      "labelled_refs": [
+        "package:inferrsi.common.skills.evaluate_elastic_serving_placement"
+      ],
+      "labelled_owner_context_present": true,
+      "labelled_context_bytes": 51463,
+      "total_context_bytes": 51463,
+      "labelled_context_omissions": []
+    },
+    {
+      "side": "B",
+      "case": "withheld06",
+      "labelled_refs": [
+        "package:inferrsi.common.skills.performance_ceiling_analysis"
+      ],
+      "labelled_owner_context_present": false,
+      "labelled_context_bytes": 0,
+      "total_context_bytes": 51702,
+      "labelled_context_omissions": [
+        {
+          "available_bytes": 65536,
+          "reason": "context_limit",
+          "ref": "package:inferrsi.common.skills.performance_ceiling_analysis",
+          "required_bytes": 170782
+        }
+      ]
+    }
+  ]
+}
diff --git a/evidence/knowledge-retrieval-20261027/results/dry-run-benchmark.json b/evidence/knowledge-retrieval-20261027/results/dry-run-benchmark.json
new file mode 100644
index 00000000..dba0efc8
--- /dev/null
+++ b/evidence/knowledge-retrieval-20261027/results/dry-run-benchmark.json
@@ -0,0 +1,5386 @@
+{
+  "started_at_utc": "2026-10-27T11:13:19.366966+00:00",
+  "dry_run": true,
+  "verdict": "INCONCLUSIVE",
+  "freeze_content_sha256": "2a90de774129b06597c3c99e5b06d05bf735026829deb84376d06a985cdcc276",
+  "identities": {
+    "A": {
+      "corpus": {
+        "bytes": 2883003,
+        "file_count": 458,
+        "files": {
+          "src/inferrsi/data/knowledge_packs/README.md": {
+            "bytes": 14172,
+            "sha256": "b877e67d9cccf0dcbe57419eeb3ff9374e5bd03ff8eec63adef1025c446140e9"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/diffusers/cuda/diffusion/group-offload-transfer-overlap.json": {
+            "bytes": 3277,
+            "sha256": "9bec3d912192efa2f2c853d3fb70bdae21a75875630a02294fcd084b18c30414"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/flashattention/cuda/attention/generation-specific-dispatch.json": {
+            "bytes": 3085,
+            "sha256": "90290d7bc80b1626863e0520c91bb196d93ebfe08190a450e9de59c6505ae8ad"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/flashinfer/cuda/llm/paged-decode-dispatch.json": {
+            "bytes": 3745,
+            "sha256": "0f2986494fcc9b751df71e598fd61caa68e28bc4b0ac401e4cc763bc3f75283b"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/cross-stream-dependency.json": {
+            "bytes": 3809,
+            "sha256": "b829315e55adf8e11f0bd581d5f45dd130d7eb3f81ce7de532f63c7b9c27cbb5"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/evaluation-completion.json": {
+            "bytes": 3841,
+            "sha256": "19d2e506ac4f50895776a83a062059aa7a052175c8e4b21dd0d46e1a5a6716a1"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/post-detachment-buffer-lifetime.json": {
+            "bytes": 4663,
+            "sha256": "18d78c0e542dacb42ca33e5264d92aef80fa06d163fd16d4feafd5ed0827ab99"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/pytorch/cuda/compiler/cudagraph-shapes-and-lifetimes.json": {
+            "bytes": 3623,
+            "sha256": "7fef4122ff794967e61f55d02a4a6b558d7e498d9513607f5ca86db6386928a8"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/transformers/common/llm/kv-cache-strategy-contracts.json": {
+            "bytes": 3072,
+            "sha256": "a2df26052112a9f3fc6242b2bd636ab13991442db57e56646168d6c52032d38d"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/triton/common/kernel/bandwidth-fusion-resource-limits.json": {
+            "bytes": 2941,
+            "sha256": "6e9673d1a600a6ad34820e4325f2a6417f44f0f96d8ed0bcf21466d998f19ee4"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/cache_dit/common/diffusion/residual-cache-calibration.json": {
+            "bytes": 3841,
+            "sha256": "bb9509ad0225a91b52bc6787477581d05f0ecd5e44fa59f08f718eea36a2ac45"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/bounded-windowed-replay.json": {
+            "bytes": 10959,
+            "sha256": "43613225e9ccdbd34c0f7e3e0767887b9bbfd0f38b2c21684677d99e31a432f0"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/speculative-verification-budget.json": {
+            "bytes": 10177,
+            "sha256": "a729e1fc00be44e1480a5dc65bbcdd58df8599681df74dbc651a5a49f27e9028"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/zero-overhead-deferred-scheduling.json": {
+            "bytes": 8191,
+            "sha256": "ed5426af145f066f6183bd316c4073c46f324f5ee83aa9bd616387c0d0e7149c"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/lmdeploy/cuda/llm/kv-quantization-capacity-cost.json": {
+            "bytes": 3366,
+            "sha256": "873997b38981ccc03ff73e6fb84e96022f293a019843efd8c4aa50f27878ac9a"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/codec-constants-and-state.json": {
+            "bytes": 4250,
+            "sha256": "1cecd7905fef5af3ef8952a4d803900a7df91d662ebae34db47cc0dbf352d4e7"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/fixed-customvoice-prompt.json": {
+            "bytes": 3660,
+            "sha256": "ea5e8fdf54cb731b255b73aa0cf47ef53fac103baa2bd5246be45f95d54cb593"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/graph-batching-state-ownership.json": {
+            "bytes": 3950,
+            "sha256": "130741a9f7ac9d366784d6c4dd42860262235caff0c277de1e8a8f8ec933b37f"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/sglang/cuda/llm/hierarchical-prefix-cache.json": {
+            "bytes": 3641,
+            "sha256": "b50ca58df28d4343545d07adf0da23f8f181297b30aba7870e5690630e7d7315"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/tensorrt_llm/cuda/llm/prefill-decode-disaggregation.json": {
+            "bytes": 2967,
+            "sha256": "349856dd98628f94987e014889cd1b15d0e3ae6ab6c576cdde917f28fbafb102"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/xdit/cuda/diffusion/parallelism-freshness-and-topology.json": {
+            "bytes": 3236,
+            "sha256": "3674b5916cb7b384951bf3e449b6796b161878daabcc1d7e7f6d6f8ff861cc17"
+          },
+          "src/inferrsi/data/repo_packs/afd_plugin/pack.json": {
+            "bytes": 502,
+            "sha256": "00042abeabdc9c0e24f1a4edaba47e233df398aef0ad4e181fc70191d383da5f"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-aclgraph-capture-replay.json": {
+            "bytes": 4311,
+            "sha256": "327414443ea9537ae320aa405912c338351bc06cefdc0b99352e1eea36f61dc3"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-device-health.json": {
+            "bytes": 4115,
+            "sha256": "633fa5fdcea1e1e9edda8a218d6a3fde0d12637848348e1e2fe7d1b0bd744a16"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-distributed-parallel.json": {
+            "bytes": 5841,
+            "sha256": "709ca0d061e4d6917fe8fa049fe93165e71dfecf96c7bf264ced25346381f603"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-graph-mode.json": {
+            "bytes": 4135,
+            "sha256": "bb82ab3d628b91be957e557545810ef5c20790be7d43296096f20a21f45ec391"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-model-adaptation.json": {
+            "bytes": 4137,
+            "sha256": "7ea1ab0f0044afff935a91071b36ac09bc46b5ef044fe2a68ab0608256ddd853"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-profiling.json": {
+            "bytes": 4186,
+            "sha256": "0d27058fd8d0ef8b524b7a60d1cb76950622746c94b5a03915e64e6edb4ab3d4"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-tuning.json": {
+            "bytes": 4439,
+            "sha256": "fcb49d71f59d8b39c1eaa03d24e95e2cbfe5f9d7f6e0f22df03ee0d60f6a568a"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-optimization-casebook.json": {
+            "bytes": 4371,
+            "sha256": "05783630bfaf3c7ef99b848865f4bbab50d0b52e80ad04fcd028c624d7638e9b"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-performance-analysis.json": {
+            "bytes": 7200,
+            "sha256": "b6eea897c12d0c6ac9d172089535c82c25eb71c09d7a4972e64812cc4682b5c9"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascendc-tiling-design.json": {
+            "bytes": 4505,
+            "sha256": "e69b6e89dd342ba55ae54b5fac14aa1b91ee2a361da40cb669c287ce519f4517"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/cann-attribution-overlap.json": {
+            "bytes": 4915,
+            "sha256": "8ac01b60f5cd37fd594d6cb5119e4e6525199d017296fc8779dc68a576f5525c"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/torchnpu-device-memory-lifecycle.json": {
+            "bytes": 4349,
+            "sha256": "7a79d0eb2c50381d40086e6db72d4bdd2cec5d2261bfb00cf73a60701a94da2a"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/torchnpu-pinned-host-memory-lifecycle.json": {
+            "bytes": 4431,
+            "sha256": "8bb34dcfbebb48667a94434b036a1a408b970af9890b047a3de325f364ed5f7c"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/assess-serving-preemption.json": {
+            "bytes": 4198,
+            "sha256": "e3d2a1a72d0d980f31fc85deee27db26bc0daa0c709e8ae99e42deb814a4a7f4"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/assess-split-backward-schedule.json": {
+            "bytes": 3915,
+            "sha256": "a8561794786eb0a897d831f00c92b4fdd3b0492316c1b1d8ad13c1f8f75ac815"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-h3-runtime-residency.json": {
+            "bytes": 12169,
+            "sha256": "602dd625874bd4a89861d5bb25881cc63e4b5a7fdedc2867089f0952943f6eff"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-llama-runtime-state.json": {
+            "bytes": 5816,
+            "sha256": "64f6ab8a95172378350323cead08bc6c201c624b711e6242938a6a21a8e83ad7"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-sd-cpp-residency.json": {
+            "bytes": 8349,
+            "sha256": "95ed97b717255be3b817e5c69debd6d41197559dbf21771507d5da79134a72e1"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-sglang-diffusion-lifetimes.json": {
+            "bytes": 7276,
+            "sha256": "5e5889223878c6cf42a49c6f09c60e9d9add88cb7e7d0df050c4e75d88a05aa6"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-vllm-kv-lifetimes.json": {
+            "bytes": 4275,
+            "sha256": "00424005a5eeb8b15f239edd2cb2c71047787e3ecf55924860cb7a78d137b586"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/bound-windowed-attention-replay.json": {
+            "bytes": 7877,
+            "sha256": "98b2291147a796ef59cf20a075ff662998ee6a529f85b49e9c6fb62fe41fad9f"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/budget-speculative-verification.json": {
+            "bytes": 7490,
+            "sha256": "66448f59460c9dffac46e74160388fab650cf8bbbd5bdc95c1465b70daf0c2dc"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/calibrate-serving-simulation.json": {
+            "bytes": 3992,
+            "sha256": "055b70e0b86f65156775b519e4b8cbf06a956a8993b06a66175167cedb78f2cb"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/cgroup-resource-controls.json": {
+            "bytes": 4801,
+            "sha256": "ad406b6dd5d9415bd28d448e64269a388b50e11f9eb7f5a9adb6b0a6246fd303"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/checkpoint-recovery-consistency.json": {
+            "bytes": 4740,
+            "sha256": "ccb85314a831bde58f8cc515580d001b174c11dca0bbcbc6221303acf629046d"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/choose-parallel-training-plan.json": {
+            "bytes": 3901,
+            "sha256": "40afcb41c59988fc1af8fdd00d2340800c54e96847dd0de16ecddf7ac221d1b4"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/cite-check-downstream-pr.json": {
+            "bytes": 2533,
+            "sha256": "dabff14684d5ae9a2b05a9743ff5b76d00a90fdb29319588aba67f5087aad706"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/claim-contract.json": {
+            "bytes": 2469,
+            "sha256": "78f15f6533434fc29adf1df651f7a15eb8ba81733f8b02c710d90102541d707b"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/comfy-ui.json": {
+            "bytes": 6919,
+            "sha256": "1496807d958df055f720ca93e6dcf7f7e8f506f3f12c94c9a2166f789b00058d"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/competitor-scan.json": {
+            "bytes": 2449,
+            "sha256": "c2e97e49e5185710cdfaa08ac1c2b1c19b958e68ebfca885d5aee0ed8c9f8c8b"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dcp-async-save-performance.json": {
+            "bytes": 4538,
+            "sha256": "359a5acf8304a78fc074688676f9a0be62f56f463d0960a544e409f49843bdaf"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dense-performance-evidence.json": {
+            "bytes": 4227,
+            "sha256": "d87f93a9dc9ddc583e3929baaac85f7a425666dd6dd71665933f18344ccec834"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/diagnose-vllm-admission-pressure.json": {
+            "bytes": 9789,
+            "sha256": "a1de40def775b2e840c27330b67e6f37c3c244db1731fab2bd3e8a787143210e"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/diffusers-invocation-state.json": {
+            "bytes": 6109,
+            "sha256": "914d2ffe9776ba931bed780a2a270c1525e3c6cbaf0735b8531064d3180f3233"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/draft-acceptance-issue.json": {
+            "bytes": 2549,
+            "sha256": "6aaa63dbfe6cda982ccd3a4e368527b80fd2873672c000c66d06d28e40a2bf14"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/evaluate-elastic-serving-placement.json": {
+            "bytes": 4334,
+            "sha256": "fab1fa814e789ec2d86587fcf7de7c38174464b9e9ba0cf53ef3fb470053e7c4"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/generation-cache-lifecycle.json": {
+            "bytes": 5972,
+            "sha256": "e5679cdecb722bbd17280913913da8473052ad09e9e7a13b3a9b95463a139ef6"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/host-memory-management.json": {
+            "bytes": 6530,
+            "sha256": "a36aea21169281866b2468860d3ffeb912e61edd571308b02f82939f17f694a8"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/host-overhead-optimization.json": {
+            "bytes": 4243,
+            "sha256": "05fd7f054966b96892980d585352bc1c3fcd47f439cb34268fdf07feee58a998"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/inference-benchmark-design.json": {
+            "bytes": 10735,
+            "sha256": "091bb086c6f0123dbb5db78968c13f846cad135e8969982efe557644b7c7613b"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/inference-optimization-casebook.json": {
+            "bytes": 6531,
+            "sha256": "c512025d7a59934c91665e24752d9393ca9b6a9dfb624c72789294defae93c0e"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/kernel-writing-guide.json": {
+            "bytes": 4545,
+            "sha256": "f1f0ca1d11d3550c9b8c5fdbed8f6cd09b877ac14341f973d58dca02b5d8995d"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/kv-tier-management.json": {
+            "bytes": 4609,
+            "sha256": "6141f61fb6b0d147b704ab7a2627e9babdb53a90e15b659152a3d565967afddf"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/lazy-evaluation-stream-lifetimes.json": {
+            "bytes": 5934,
+            "sha256": "34f47009a8a33f8907bd78f110b95fcd62409bef1b38e1055acb46b2eab285ff"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/npu-skill-factory.json": {
+            "bytes": 11868,
+            "sha256": "24c1d3bff288e86b99ea674e3252ec6f54d9ba944274aef6a7666197afd6788a"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/optimization-search-policy.json": {
+            "bytes": 7101,
+            "sha256": "d7bf71d5491765f62eec02fe5013dc1a5ef56e1516892da48593c6165a86366d"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/os-process-placement.json": {
+            "bytes": 6304,
+            "sha256": "e6e8fe5d7d66f3e98f05762d94441726dfa757ce7fbffd2b35402b3f327f8fb3"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/performance-ceiling-analysis.json": {
+            "bytes": 13695,
+            "sha256": "47c034d96026e1c2a02848d9c9c3caf4024a5033664bfd089ad30d9f23a80deb"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/plan-activation-residency.json": {
+            "bytes": 3871,
+            "sha256": "436e364020cc854fd52a10ee76e4593c7519bb0546c8aa5238fcfd89158e8ab0"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/repository-evolution-analysis.json": {
+            "bytes": 5865,
+            "sha256": "e0af402c3320ebd9bd36c774cb52c0708721a86ca446c05f6e94306888838e30"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/rsi-evidence.json": {
+            "bytes": 2475,
+            "sha256": "c8b4bb5a8c88009ac5b1526e82639715381c940d3c3eb295f0333cae6c2d3ca9"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/skill-evolution-playbook.json": {
+            "bytes": 4091,
+            "sha256": "0e063c2e7686de0c27c061e1aac0eb2d8c4fa83246e32ddeeb0e28126c176425"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/storage-io-measurement.json": {
+            "bytes": 4783,
+            "sha256": "80e589e3a0b67d3a045e03feeadc45e720ceb277e6124c917eec04863eb67419"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/training-input-pipeline.json": {
+            "bytes": 4296,
+            "sha256": "ae696bea8d2b7bfdaa29b6e5bc0bd861abcc0e42e7d9bec6ae20c6e7d6f27669"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/training-resource-budget.json": {
+            "bytes": 6944,
+            "sha256": "ea5e5e864e3aa34bd00be67c8e741877ae01480f039652073b65b71dfcd54c97"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/weight-loading-optimization.json": {
+            "bytes": 8139,
+            "sha256": "7f2d1fbc6b1cfd80c3d6ed8df107d4ca9b97655626ee078c345445a303b1b367"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/audit-afd-p2p-exchanges.json": {
+            "bytes": 6640,
+            "sha256": "9ac45aa784b430ad1ca9ffddecce8dbae9740cfe9c98c10b330e01d350015a38"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/diagnose-attention-bottlenecks.json": {
+            "bytes": 7602,
+            "sha256": "e525d75578de1cc164e6553d01efbd0a014e74646200580f59b1161c030413ad"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/host-overhead-profiling.json": {
+            "bytes": 3965,
+            "sha256": "e9e6ae8c5b9986e83cade3bd029baae112859e01aa6c96f741a264f3bee44adc"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/reason-hopper-kernels.json": {
+            "bytes": 15086,
+            "sha256": "758a649e651d8f9635ac2df45d18043e156c8d54b8f8dbc95f8e68ebcb866a28"
+          },
+          "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/parallelism-and-extension.json": {
+            "bytes": 3399,
+            "sha256": "1253f5f01bcf1498703c7f34960f0c50580df5ba8f140af557a903e296ad81d3"
+          },
+          "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/runtime-radix-hicache.json": {
+            "bytes": 3317,
+            "sha256": "5a0b2aed51649af214cdc1034f57650b7ce195c934a17452ee67a35131f80cdd"
+          },
+          "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/tuning-and-validation.json": {
+            "bytes": 3385,
+            "sha256": "ef4d4f7182da4433d59d72dda0d21b4aeafadd73260be11ab8d4cb36c20fafc6"
+          },
+          "src/inferrsi/data/repo_packs/sglang/pack.json": {
+            "bytes": 620,
+            "sha256": "bce2e0c42796a9ef499d459c37b56d1d56f4b19344841359c534fca29c0359ce"
+          },
+          "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/audio/serving-and-streaming-quality.json": {
+            "bytes": 3440,
+            "sha256": "97fc1d5a11a39a0634dae9e9602a4ec47a4d2348e121076c2d5c836cf0692114"
+          },
+          "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/omni/extension-lifecycle-validation.json": {
+            "bytes": 3482,
+            "sha256": "7da8dc5d2655099de2bcd5f0fa922d9e1bcb0bfd2f571f026a004a7f402a75b6"
+          },
+          "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/omni/pipeline-state-and-transport.json": {
+            "bytes": 3404,
+            "sha256": "0c57692772019504264e39bf0444a69e20cff86f6e112b072c53cf7a4299cc82"
+          },
+          "src/inferrsi/data/repo_packs/sglang_omni/pack.json": {
+            "bytes": 630,
+            "sha256": "8e70336bc2aac0c74732a026337813652b5d5c6e1fc60f5571e9442345077b96"
+          },
+          "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/architecture-search-space.json": {
+            "bytes": 3294,
+            "sha256": "2c5ad72d5bab4f7082fc3be1532c347444cde38f5a84639fe16b22134b68685e"
+          },
+          "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/controlled-execution-tuning.json": {
+            "bytes": 3255,
+            "sha256": "6a2bf611a432213919e1c237687507ceb0ab753cccad0e934dc77f86604805e9"
+          },
+          "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/engine-validation-contract.json": {
+            "bytes": 3295,
+            "sha256": "ef89775bc82965fe87a832c7ec8a5cf55cb9eba4ecfc6279caea9ce3b0410c41"
+          },
+          "src/inferrsi/data/repo_packs/strata/pack.json": {
+            "bytes": 617,
+            "sha256": "643b32defb8db9aa230f624856f4fb4b595b5c23db58d9453091f5420cff71c3"
+          },
+          "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/bounded-serving-measurement.json": {
+            "bytes": 3239,
+            "sha256": "fb00ddac14ef6b06f48da9ba335529bd55d5d83c2a2fe331f0aa1ca16afceb37"
+          },
+          "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/extension-correctness-gates.json": {
+            "bytes": 3343,
+            "sha256": "4ea7083d277df205ea4c361bee87a8207951596e9ef6cfa136cc7d2aeab02f46"
+          },
+          "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/runtime-and-kv-ownership.json": {
+            "bytes": 3280,
+            "sha256": "e6142c1065754716a5f26f42119722908dc8fff63e667c986d57f1227b8610bd"
+          },
+          "src/inferrsi/data/repo_packs/vllm/pack.json": {
+            "bytes": 508,
+            "sha256": "1705d96710249c4514c15d8d011b43c1a1ae7c15fbc9ddf11ef2d0e6b2ab7d07"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/minimax-h3-reference-kv-reuse.json": {
+            "bytes": 4047,
+            "sha256": "80b853fc594d387ab48bed04dca245e0a22b8cdaf7c9b853551ab3bafa302e84"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/minimax-h3-sp-head-buckets.json": {
+            "bytes": 3742,
+            "sha256": "c4d29d1bceb5e2cf86151bf53c9dafb3d83428fb1e9d2a848d852bb8824bf857"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/qwen-image-21-eager-t2i.json": {
+            "bytes": 5268,
+            "sha256": "ef107d43cc448219eda4d64ab961b61cda6350e20d30d94de1ff7c4ffff7e775"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/audio/streaming-serving-measurement.json": {
+            "bytes": 3389,
+            "sha256": "2680f32f28907f6abc7a7e1e3ecab4f46ee412f40f4898ed2f371ffadf47ae4d"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/measurement-and-profiling.json": {
+            "bytes": 4472,
+            "sha256": "c8b25583cd160024c7472a20debea34cb6d4bc79714b4bad2953555adc9f794e"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/production-integration-gates.json": {
+            "bytes": 6112,
+            "sha256": "649b49ff64d3e9e88bd620027dabf9d126ca69bbe91b448f28e845ecbe1f8e2f"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/step-state-and-output-validation.json": {
+            "bytes": 5090,
+            "sha256": "e1ffe28c4fa54af739cdef2fedf4755bba18e078f8380d4fde8138ea712e5778"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/upstream-review-gates.json": {
+            "bytes": 7614,
+            "sha256": "5127709bb85ce31008d19468e1b4fcb5454a71776f69f991841cd9ef7558f9c4"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/omni/stage-ownership-and-search.json": {
+            "bytes": 3427,
+            "sha256": "e75d6c2829ae22881893f5da2ad7edba099fd64ccfb0fc93d30473e06e9d4d80"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/audio/concurrency-scaling-state-and-overlap.json": {
+            "bytes": 25324,
+            "sha256": "cb47fabb232db81ed525cbb54371becc7e4a28b2bae9d8492f60dfb6f31563e8"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-adaln-schedule-cache.json": {
+            "bytes": 4253,
+            "sha256": "ddbba3bb72b5789938595a82d3280e861f75f6ed54f05a1d2b7dcb1089d30f63"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-cuda-mxfp8.json": {
+            "bytes": 4468,
+            "sha256": "e4c538f03f4ab90a4acc8e3d8721bbd226aec1ee9bde45757a7c27d3d970821d"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-fp32-modulation.json": {
+            "bytes": 3625,
+            "sha256": "abb8d94a1b025f0e2c8e085275e96ee85d68aeb27cb12d0918e4647f8255b693"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-qk-norm-rope.json": {
+            "bytes": 3706,
+            "sha256": "4c1c4acfb9cc0baa90bc28fed1903020ea1ea38755b84315711795db45e48bb6"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-ref2va-teacache.json": {
+            "bytes": 3835,
+            "sha256": "8aeb3f99ac4227cf8b3a5cece13d938b19082e265a557eddae27b7e495bcc22e"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-vae-stacked-tiling.json": {
+            "bytes": 3508,
+            "sha256": "8cc18eacbbfd8fc4d3a6b4152c4b584d559fc9b15377da56ef072e542d807c93"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/pack.json": {
+            "bytes": 566,
+            "sha256": "58c971e9dcf6ac15752629ed4d9426aeb5b5d857fc37b5bcf921aa2e81c0cdd5"
+          },
+          "src/inferrsi/data/repo_packs/vllm_rlt/pack.json": {
+            "bytes": 649,
+            "sha256": "6097041b8d026740060e527f49b1b4e288315b1c3227f7583805b86d8fa415f7"
+          },
+          "src/inferrsi/data/skill_library/README.md": {
+            "bytes": 18819,
+            "sha256": "ae583f3578425df1dc91ba32b8b6543c0469c9a817ebea873a77051d6b13e1ce"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/SKILL.md": {
+            "bytes": 5494,
+            "sha256": "7581b56336d07213a87d01477793ae1a8acf0b7cf876fce36c9a761ee32c1ff4"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/metadata.json": {
+            "bytes": 1429,
+            "sha256": "68ac7f64b7ab947fc2db664bc3990f102c1cc5539b42350aa5818d0bbf17d86c"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/ownership.md": {
+            "bytes": 2081,
+            "sha256": "40688cc44183027efbf2fca8d3548c299cdf2e33fef54f76ab3549353eb9bd1b"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/source-map.md": {
+            "bytes": 4635,
+            "sha256": "10562a75683bd9bd07ec2e37a1cbd0f93f2534d6f85b55a566adb2d97485215f"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/validation.md": {
+            "bytes": 1477,
+            "sha256": "5a0b854b87d091d7d505b0bdd5c03330052507a851ab6edf35b93a6675f2d4f7"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/SKILL.md": {
+            "bytes": 4010,
+            "sha256": "a95569bcc0da6e715fd74783a6a1e0489e92b680c582f15dafbf98df589d505a"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/metadata.json": {
+            "bytes": 1519,
+            "sha256": "9c2293c8956508de987c886c7224d673d22912e7e915b026b05abdd287e97401"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/execution-lifetime.md": {
+            "bytes": 2147,
+            "sha256": "c4bc41e08aca68f661325f61e6f0db1760a079dc362f295c847bdc374fee3bf6"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/source-map.md": {
+            "bytes": 5918,
+            "sha256": "798b9cb89ed6b885eb47464ca9f137d2802a54571b9327e631d7742d3e5abd5d"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/validation.md": {
+            "bytes": 1608,
+            "sha256": "ebfc00bf7d6ab0a1dc415953ae84d56ccdc713d5614aff85e433c4a843793bff"
+          },
+          "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md": {
+            "bytes": 7322,
+            "sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845"
+          },
+          "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json": {
+            "bytes": 587,
+            "sha256": "db7ad368394ccbfdeff6a16e3423c300d51305b1f95ff5c3936df91738a8cc35"
+          },
+          "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md": {
+            "bytes": 5314,
+            "sha256": "22afe4f4d77f441096cc657d55d763473bfcbea8fb531c8e493fe49ab84955f7"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/SKILL.md": {
+            "bytes": 6745,
+            "sha256": "d494fc7e539cbb7388189bd7b3f04bb5c4983fdd7a82484348aad1cee7f741f2"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/metadata.json": {
+            "bytes": 558,
+            "sha256": "71c68431663ed2336cccc985a354adab6b13b4cfd6123a18a5d8d53c28268509"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/references/source-notes.md": {
+            "bytes": 13004,
+            "sha256": "6ffbfb68cc7e04923b4fbca2a73a7555b7c3b45a6f40819bb5f3213ef05cf88d"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/SKILL.md": {
+            "bytes": 5977,
+            "sha256": "03829039860ece35519f2e4c339a48b5014ec25e07a2c150a3efb6a3a4e35b4a"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/metadata.json": {
+            "bytes": 559,
+            "sha256": "80d80c15514919826fa72d6f6cf26b76db8accd85d21187fc239232ca454e7bf"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/SKILL.md": {
+            "bytes": 14703,
+            "sha256": "4ce9dcadd1ae3c7a8cc633336760d0a3c382f354f1b7028e3a10d28fa0d0c9fa"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/metadata.json": {
+            "bytes": 1403,
+            "sha256": "b6cb25879b1bca5f5af08b48776f2e664cf5dcd11a0efb1c4f56879e2afa9006"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/references/awesome-ascend-communication-tests.md": {
+            "bytes": 9388,
+            "sha256": "0cb27a3d24409171bb414d7ebf0c49ad62be5f46482f29624695496ba9eb3e63"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/SKILL.md": {
+            "bytes": 5839,
+            "sha256": "9203cc5a14a1d65df5fbf275734b93be023e06527459b2f79864f28a338cb44a"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/metadata.json": {
+            "bytes": 536,
+            "sha256": "72b41e7845915d34b09219624e82976386f3125816293540c2c38143fe6b477f"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/SKILL.md": {
+            "bytes": 6440,
+            "sha256": "8fdf46fb2743b67a66bc069d22a45305a222debfb606777bc451b4b8f31d9cc6"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/metadata.json": {
+            "bytes": 536,
+            "sha256": "be681a3206e5e183b4c9670c9d7784bcffce18c891449e5b15f94ca5f1324b9a"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/SKILL.md": {
+            "bytes": 12988,
+            "sha256": "401981cf7828edcab0c9ea882f7ecd83badfa3c2a95006e912cf99d67b3d6a12"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/metadata.json": {
+            "bytes": 760,
+            "sha256": "efd37bb23acb500dad4ec8ebe7ddfe5c3f196374052825f923ac85da8445f06b"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/SKILL.md": {
+            "bytes": 12236,
+            "sha256": "38b151bc6b4e9734c6fa41efb7a30343559400c0c75bbdd26bd8a968b8ea4592"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/metadata.json": {
+            "bytes": 557,
+            "sha256": "04d2002fc8a0a7becf86b0b73daaf2da8716ab466d7f988da43d41a10f26680e"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/SKILL.md": {
+            "bytes": 9172,
+            "sha256": "647f5e9a288f2c4e1617ba53df59d9d827df749eb4fe86182d21a9c41175caf8"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/metadata.json": {
+            "bytes": 536,
+            "sha256": "f425f7c1abbd616387273e8932c6bd02f4b1da012f099286ad4171e0f6037b22"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/SKILL.md": {
+            "bytes": 9880,
+            "sha256": "a057b7f86a24f34a8f84abbbb4910d58d4e50ff47060e1c404600017e3febe89"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/metadata.json": {
+            "bytes": 2297,
+            "sha256": "9e705625c8308013c287576b0247da6f6116d817a4a535cbecff80abd6ea35ee"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-CANN-2.0-LICENSE.md": {
+            "bytes": 5262,
+            "sha256": "89b1c114df62cfd183004f38d18d07f186436c3b37203dd69c3e84becf9d2282"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-stage-evidence.md": {
+            "bytes": 11789,
+            "sha256": "f56e4125325c9c1b9235a1d64fd732861c7686444492422daff44ba8dd648be4"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md": {
+            "bytes": 11860,
+            "sha256": "93aa1b775ced304cebfe6c4dec0503ddb2260ecb3cb01edc17ce329446c5ce2b"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/metadata.json": {
+            "bytes": 552,
+            "sha256": "f6d94f104cb70969024b2f4542078a240e200712c66279eca8b94f60bee1dd71"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md": {
+            "bytes": 8972,
+            "sha256": "0e6f094dc554abea12e07def50f20b233b47127779a8aefad595b4ab1efd24c2"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/metadata.json": {
+            "bytes": 569,
+            "sha256": "acccc8eb86de0df6caf78a9b3c6db76a2a686adc4a5ba0029a4d62d50ff9a8f8"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md": {
+            "bytes": 5989,
+            "sha256": "8428df7157af024fa17cfe0b3818ed54bd9f468306dda46eb8d4f0edf4f3f230"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/metadata.json": {
+            "bytes": 579,
+            "sha256": "c650dce88c07b636d68e063dbda4d319f6f7f3ecbcdd5e47f393cbfc3af7c71d"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/references/source-notes.md": {
+            "bytes": 11470,
+            "sha256": "80cba4d3a884e7ee2618326cf74b9f9c3a34e4d9cf443707f6bbe673f2d5ba65"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/SKILL.md": {
+            "bytes": 6159,
+            "sha256": "23e93ea7c98f99743509fdc04139d78ed2e7123c63e80b366c6beaba7eb70d0d"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/metadata.json": {
+            "bytes": 581,
+            "sha256": "97836a38465106d816a9ace894a7b277d6d330f9accade8f8493ce6f0fb57ef9"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/references/source-notes.md": {
+            "bytes": 13493,
+            "sha256": "435193590b4a2ba32e2e4c6fa9d7e8f889528d412cbbb48e35ec5433f7dab384"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/SKILL.md": {
+            "bytes": 7137,
+            "sha256": "a25738dbb3b9daa15c5827cb536a20fe13f7c8b9dd14aa3fb7cfa7e132879431"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/metadata.json": {
+            "bytes": 1007,
+            "sha256": "3abe626a6e4c73c943429f609ce2c62aa77aae7e1726475ebbbdfbe1794f6f7d"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/mechanisms.md": {
+            "bytes": 17376,
+            "sha256": "fe08b2292fa587be3874774ddf41063b84c579a0085df422ec6710d8aa90b8b7"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 4024,
+            "sha256": "371b01ee201ef51fecbfd6da892d517aaba295930b9395025f03f34e21c030a1"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/NEXT_GATE_LEADS.md": {
+            "bytes": 1623,
+            "sha256": "35b8b58ae5db02b2df51ccbe5cdf870a7383d17f345f9ee5eb7b0a4146dfc74b"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/README.md": {
+            "bytes": 2888,
+            "sha256": "e0b1ec1b39c10067f5ddce58320851ad089568c4b2296a24777389097cb5b388"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/io-reduction-can-plateau.md": {
+            "bytes": 2671,
+            "sha256": "5cc35f514224ac6e6efd2da6db221bba468b2f4db66807295e67b240bcdfdba5"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/overlap-and-fp8-counter-signals.md": {
+            "bytes": 3692,
+            "sha256": "80c8a60a6df55f3d84fab23e03567d8b38e23bcdbd63cc09881e46db1e2888ab"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/partitioning-after-fusion.md": {
+            "bytes": 3132,
+            "sha256": "4d4cabe0991eab554dcfbd4fa9c8199e61fa4767709b0c4cd646e2588bb398a2"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/knowledge/attention-cost-model-boundaries.md": {
+            "bytes": 5223,
+            "sha256": "5ed8f888443e38ccb72545b6025447aab7b92936f78c151fed680e582126239b"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/sources.md": {
+            "bytes": 9069,
+            "sha256": "53436e0de4e6dda2d29426591bf61f8526378cdc6f9ec7967b88b1bcd4c61e6f"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/host-overhead-profiling/SKILL.md": {
+            "bytes": 11199,
+            "sha256": "326e566fbe30d3ba04e4b50c3e87aad9df037f0ac4ad6fe129165ae97dc5f91d"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/host-overhead-profiling/metadata.json": {
+            "bytes": 527,
+            "sha256": "76e34c1c64aba473a9e126b032a776417caa0f913f6ac3cf4dba3a5d123b1ee0"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/SKILL.md": {
+            "bytes": 4905,
+            "sha256": "cec4e3aa1eb64c31c54b5f5ca40af2b61b7919d94be4174968581d5f752c0538"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/metadata.json": {
+            "bytes": 7453,
+            "sha256": "7520da1cc97412631a600957ecb18aa5a7a4461ea21e76c237330180193ed898"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/fp8.md": {
+            "bytes": 6287,
+            "sha256": "31d9c211c11c2224007abbdcd4059913c4a54833fe2d2f926411c5238fcd3501"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/cutlass-bsd.md": {
+            "bytes": 1547,
+            "sha256": "42fec630f410aa308f70a51a89fadcd19586fa620f9831a32bee528a9a10000e"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/deepgemm-mit.md": {
+            "bytes": 1065,
+            "sha256": "6c73a30c14a2938abaa6afe98ebb19e6d81dac5bd4ace0cb05c61c20afa37b85"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/kernelwiki-kda-mit.md": {
+            "bytes": 1063,
+            "sha256": "84aab57158a914f448b38d8be29dd842b634a08c192d844f416a256beed4ae94"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/notices.md": {
+            "bytes": 3240,
+            "sha256": "876ec016644bf5331b3f257985e126c0e61980e3bfd82569b7b0c3f6b2c6c64c"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/resources-and-scheduling.md": {
+            "bytes": 8919,
+            "sha256": "1ce25b426736072435a1ead9a2a8e1f56df5d0abb8198e8c4fdcb34321db4964"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/source-inventory.json": {
+            "bytes": 120506,
+            "sha256": "1dc09c895c970a0c9708158bcc6dd65cdb736274fd2c35d44542c70407d3849d"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/sources.md": {
+            "bytes": 9004,
+            "sha256": "6b74b9e774f2a80db0860510d9e6940d36284430f29c1e124f115a9eaf25833d"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/synchronization.md": {
+            "bytes": 6843,
+            "sha256": "4322f6b3c0e4368bf8c192b4ec752529767f50d19c0c38dad8fdf9cd781e41d9"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/SKILL.md": {
+            "bytes": 7348,
+            "sha256": "fd1f5a7e6a55714730044155f5b38c1d700383d4e16c3e91963bbfbe1c773af5"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/metadata.json": {
+            "bytes": 572,
+            "sha256": "19e40763fedc20990651bb989c32a7693306c5c54af46c3d4be77fd4f87b6c03"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/SKILL.md": {
+            "bytes": 23213,
+            "sha256": "2c807f3df2da2f533d25183de5771f3d412058d16410994292ef8f48c49f9b3a"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/metadata.json": {
+            "bytes": 1684,
+            "sha256": "9ac6af135f0b6d76a76bc201b6bbd8ac9afe5201dd1e2ea82c0ac4826de401da"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-LICENSE.md": {
+            "bytes": 11357,
+            "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-native-offload-readset.json": {
+            "bytes": 6267,
+            "sha256": "cc6c2011be746002ba3ff96ccefab1809863748ddd17eb102b75e7b0e134fe55"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-native-offload.md": {
+            "bytes": 8966,
+            "sha256": "6c2a72b5ba3708a01d229e0576570837a617e6a42dbb0b5312a26a65a6c31542"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/SKILL.md": {
+            "bytes": 16072,
+            "sha256": "36ef88826a32ac3147b869001004ded17a8d5a491e25bfe105005ba39ea9d1eb"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/metadata.json": {
+            "bytes": 539,
+            "sha256": "ae71c389dfdb1a19d267feecb4c6f1fbb4c9afad501ff6d8fab7e38b29bcc37a"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/references/deepinfer-zero-overhead-scheduling.md": {
+            "bytes": 10894,
+            "sha256": "f452bf40132c9195d739f1c86b3aa18857138c7ed0f1861af5853d93b9955924"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/os-process-placement/SKILL.md": {
+            "bytes": 10327,
+            "sha256": "2003492e4d89fde0443b180d74933e764bd1979435e109c8d26e20eb8f6e5053"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/os-process-placement/metadata.json": {
+            "bytes": 1399,
+            "sha256": "ec02995c075f0a540921f57b4846f2cafe334eb90f44389f17c3c50457c3e9c5"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/os-process-placement/references/llama-numa-thread-readset.md": {
+            "bytes": 6617,
+            "sha256": "748b254b3b043a5ddb0f480eb4a566854d27eedd7b111499863e9bf0a5a6951e"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md": {
+            "bytes": 8813,
+            "sha256": "63b883c10a8d3e7d918abfc7175b72e2b401ff1e02697f94b8382ec1c8f369b7"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/metadata.json": {
+            "bytes": 558,
+            "sha256": "0a41c7b00978332b54ee1e4a7c1ade88ecde3e961128f24d1b4f7708caff5ff5"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md": {
+            "bytes": 6402,
+            "sha256": "0a5f48d9352c5bf460775ef47f9136829978811f4a16e84983e299782cae305d"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-serving-preemption/metadata.json": {
+            "bytes": 537,
+            "sha256": "77ce5ba16245cbd80d46d0abdf80aaf26fc0924fa2d2a5624ff430de68a1daf0"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-serving-preemption/references/source-notes.md": {
+            "bytes": 30405,
+            "sha256": "f29fcecfba0490008a354c11b42403983fe165e56ff907ad61cc6040d2b6407f"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/SKILL.md": {
+            "bytes": 3181,
+            "sha256": "1e64f28ea71672fb53c56f6108b87864738c6d03d6dcd32eedaef9e0625e96e3"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/metadata.json": {
+            "bytes": 478,
+            "sha256": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/references/source-notes.md": {
+            "bytes": 10548,
+            "sha256": "61b1609f71c37b61d136b8c4d39f793de87cffb1cf31c8ec737a1c3fa6fb1b4e"
+          },
+          "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/SKILL.md": {
+            "bytes": 3884,
+            "sha256": "ac773c6f1c110a664dfca28a3ae9129b10c07614e60ca3640dea776220e7b352"
+          },
+          "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/metadata.json": {
+            "bytes": 648,
+            "sha256": "261ce9f644ea6eebb90c830f41fb8b320d2b401456e5bd8d494095a44b8620d0"
+          },
+          "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/references/source-notes.md": {
+            "bytes": 8415,
+            "sha256": "9696d3c0ef96b6611787446bd6f96d4d6255ca8d107105de8c1d9af83edb9fa5"
+          },
+          "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/SKILL.md": {
+            "bytes": 6416,
+            "sha256": "6b7db1eb2780dec27858cb7aa0a3c093c5b5dfa4c30261aebfeaa255213ff688"
+          },
+          "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/metadata.json": {
+            "bytes": 582,
+            "sha256": "4dc8659806539b4784ea9348a01d4194ac70d0bb297a916f90ff6419efa5c328"
+          },
+          "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/references/dlrover-persistence-replicas.md": {
+            "bytes": 8398,
+            "sha256": "055f4357f79026fcbcc766bbe77db610376ec353623f8898f5d14a30ee69e7bc"
+          },
+          "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/references/sources.md": {
+            "bytes": 7984,
+            "sha256": "9425b1aab963a2c8d35f3db897bd15ee69699bf5b0d0c8425c3c8bba3e9a0233"
+          },
+          "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/SKILL.md": {
+            "bytes": 3068,
+            "sha256": "25d04aed9782b2a8ff1186828becdc7fce93c4418039ceaef9b8dcc5ec9bcd24"
+          },
+          "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/metadata.json": {
+            "bytes": 478,
+            "sha256": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7"
+          },
+          "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/references/source-notes.md": {
+            "bytes": 8521,
+            "sha256": "9f9184d46b19312138a6e805e439e43dc0594a0cc6b46675255a582dc946ca5b"
+          },
+          "src/inferrsi/data/skill_library/methods/dense-performance-evidence/SKILL.md": {
+            "bytes": 8305,
+            "sha256": "53cc1867a13f54d1006d50517dea8848869ca50df3f0cd610b2424cc8d18c933"
+          },
+          "src/inferrsi/data/skill_library/methods/dense-performance-evidence/metadata.json": {
+            "bytes": 707,
+            "sha256": "a0d1e3b17140295510a764045d1b222e19be3a3406e6d1e7dddaad49ff474490"
+          },
+          "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/SKILL.md": {
+            "bytes": 6667,
+            "sha256": "82d0099d3c33706f46c657d4fa7f76d3d486e019b36760d1de001921328a60cc"
+          },
+          "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/metadata.json": {
+            "bytes": 547,
+            "sha256": "37379a4d4a058ca46da2b20a920dc01be9ec11e557c32b2332702ec76027f4b5"
+          },
+          "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/references/source-notes.md": {
+            "bytes": 44796,
+            "sha256": "ea416831fd7d0874e06d0a76fb0c7196df53072b717673a0cdf635db6bfd917b"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-benchmark-design/SKILL.md": {
+            "bytes": 8974,
+            "sha256": "19ebc1f567aadf49f7a534a18fdb629c72c0e4ce043d360983136491cb72bfab"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-benchmark-design/metadata.json": {
+            "bytes": 5923,
+            "sha256": "18310402a8ade03fd48ac687f79ba9d447871e90337de84c0e1bc15c0b827306"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-benchmark-design/references/benchmark-commands.md": {
+            "bytes": 8466,
+            "sha256": "0898424e0bff3a6246e1eeeda6abd6c5bd974c047fe32f3bbd5ab9d31ce17cd3"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-benchmark-design/references/task-benchmark-selection.md": {
+            "bytes": 25086,
+            "sha256": "0d4bb080a754e26846e81680023db0dff5e6a4c96be4327c411c3b72ac4ef8e2"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/SKILL.md": {
+            "bytes": 6644,
+            "sha256": "89a6a82c692f7f5192f4d7f4c3a5ff08a4df3b1c4bdaa0aafb7e04292d8aa5b4"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/admission-drain-occupancy.json": {
+            "bytes": 1354,
+            "sha256": "3319e23bbe2bd40e2d45fb89dc87bdc92d4df2b5ddd8086fcf84813b43885b7c"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/admission-drain-occupancy.md": {
+            "bytes": 3134,
+            "sha256": "e88bbac5d253a8e7e3a3d5e21047f23f2970c11ac23887d4303ab94fca52dd27"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/drama3-rejected-directions.json": {
+            "bytes": 1459,
+            "sha256": "2cc0a526fe0509551607a03e7626dd3d7844e643c4afb194837f75cddead8b12"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/drama3-rejected-directions.md": {
+            "bytes": 3557,
+            "sha256": "6f143bc363a3290fc008c08e57cb075f1776d1c8cca8a6a59297d5b1fbbf61bf"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-ascend-rejected-directions.json": {
+            "bytes": 1293,
+            "sha256": "b3d538f79c56724ccb2895cbf50cf2e6b0fec3979be62a684a8230cd932dff55"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-ascend-rejected-directions.md": {
+            "bytes": 2564,
+            "sha256": "be396983812262515899a3e2ba4f01b23ff92e1b836881882c7580372b8ae552"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-b300-rejected-directions.json": {
+            "bytes": 1196,
+            "sha256": "146ea1564a46ed03092c7dc5796b318fc05b04b1821105ee00eb7fa7fa1333e2"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-b300-rejected-directions.md": {
+            "bytes": 2656,
+            "sha256": "cce2930f544a0f78f21797fba4d8ded298e4a81f76c34d7977b0e31d361bc738"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/metadata.json": {
+            "bytes": 535,
+            "sha256": "fb55067fb45970319787068a3206110641cceaf0ccbabd706be6d0b3d2afc2ed"
+          },
+          "src/inferrsi/data/skill_library/methods/kernel-writing-guide/SKILL.md": {
+            "bytes": 10029,
+            "sha256": "01fb1712d7eb988c5e412f02340d2e48b4f8564fcf4ab232dc5897869d0d50a0"
+          },
+          "src/inferrsi/data/skill_library/methods/kernel-writing-guide/metadata.json": {
+            "bytes": 653,
+            "sha256": "0189d184337662056fc6952c7beffc1bf6154c4fa173ae3e3e463bd22d70f36c"
+          },
+          "src/inferrsi/data/skill_library/methods/kernel-writing-guide/references/akg-affine-tail-guards.md": {
+            "bytes": 11941,
+            "sha256": "113deb2db10b171db3425bb3f1792bee763706b794c8cc2e222b4cc0d0114118"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/SKILL.md": {
+            "bytes": 12914,
+            "sha256": "882e7850b9b49ad692a2d11618701afe0f4e83727ea78df114be2029923c725a"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/metadata.json": {
+            "bytes": 1403,
+            "sha256": "9f1fd8a2af52bc964021e21f16ecbdb891e08f2057370e700d67f86605d333db"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 3597,
+            "sha256": "157e74ba180426b279b548a729769f152bf18f9a78ffca4ebfcaf26b7d0c5292"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/MECHANISMS.md": {
+            "bytes": 5875,
+            "sha256": "18bb2f5d7e52773c516420a142edfb531d0b9c822a2934d60f3234ac1cb9d40e"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/NEXT_GATE_LEADS.md": {
+            "bytes": 941,
+            "sha256": "16a264c7159f9c2c700ce8ba46e63f3e636b277c987da4970f6d2fa085fcb053"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/knowledge/adaptation-interfaces.md": {
+            "bytes": 3893,
+            "sha256": "5e761957cb5ecd93926e6ffc13deee7c830c6e8f0ee3c3846e9dd3b50dc74446"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/references/source-notes.md": {
+            "bytes": 13370,
+            "sha256": "54d83c3713045d59a48821bcf93598374ba4a630e3891ad39498f919317c4fb8"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/updates/optimization-search-policy.md": {
+            "bytes": 4238,
+            "sha256": "d3db44fa9a6831b86a02c638b323ed943e69fe256866213f10d68c72d762879e"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/SKILL.md": {
+            "bytes": 15324,
+            "sha256": "c5f78961c23e3c8d659f5675b522518ab06e1f2a8ea8bc74c5d428ffb41693d5"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/metadata.json": {
+            "bytes": 3669,
+            "sha256": "60b5b24ead590b0fb222398fba47822bf6723690da683e256a450bb5c0eeb806"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 7349,
+            "sha256": "083e462e0a204dce599362d46712fcfb392d053c5cda56216e22e28d4b8fa0fa"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/MECHANISMS.md": {
+            "bytes": 4003,
+            "sha256": "742c249082ceca230326f4407a40130fe624553b796ae5a157a8a4c15a7e1398"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/README.md": {
+            "bytes": 1204,
+            "sha256": "17b2d1c7c14124d59ac8a0d5ad4200ad2803d991e784a1a98cd97188cc364999"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/READSET.md": {
+            "bytes": 3937,
+            "sha256": "66459ce393858b2b2160f1e2973373bfc4b9902a12644631934b4ebe3140296a"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/STATIC_CHECKS.md": {
+            "bytes": 1111,
+            "sha256": "cb0aa7d148352ebb7b1678fc42a557a099575054716a341dfa25b6a998029064"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/cases/decision-cases.md": {
+            "bytes": 5491,
+            "sha256": "781a675a6533e0a18c044aaaee0d010da7013b6504654b8b8292a35176db801e"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/govern-optimization-experiments/procedure.md": {
+            "bytes": 3823,
+            "sha256": "ce036a4d8a0908da204515c47e2ff9623548d09de264fa65bd58fedce52a78a0"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/govern-optimization-experiments/references/source-notes.md": {
+            "bytes": 21659,
+            "sha256": "f9a8254e45473ddb2b73e576957b6d3db83c62b99dff1922f4d1e606e569559e"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/knowledge/bounds-predictions-and-decisions.md": {
+            "bytes": 5297,
+            "sha256": "5780d6eb5257612ec45deb94f85f334a6ecabf7e466d1678706041f6549bda79"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/redirect-cross-layer-optimization/procedure.md": {
+            "bytes": 3648,
+            "sha256": "6593a63fce9c474b515663a560c94b7df048dc5f6e77a95b8e54e4a64cd10a09"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/redirect-cross-layer-optimization/references/source-notes.md": {
+            "bytes": 26469,
+            "sha256": "6ff3a7bc31f1172db92aedaf7df1f1aac688bf769227b583a4cd82bfad52da67"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/agentic-kernel.md": {
+            "bytes": 41353,
+            "sha256": "7190573b85f94a4391bb0d887ccbd37a31b3f08bb0531114ad7afd72fe1567a7"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/flashvector.md": {
+            "bytes": 9830,
+            "sha256": "3926369acd88ff3c370bee5718e74dd15f3219c3e4f6311095e30e917bffa4cf"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/inferencebench.md": {
+            "bytes": 9294,
+            "sha256": "21f47111197a0b7f8d6384ea918d139dc8318a62ca3313ce060048271f50b639"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/prior-coverage.md": {
+            "bytes": 3254,
+            "sha256": "35d47ac078591221fcb666219bcd824f9400fc17aafecdc0499d8336f7178238"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/vidur.md": {
+            "bytes": 7736,
+            "sha256": "d38602af5219f5feeb49e6cd6bf53d66098c77fad141dae676f1d8a6743cf25d"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/roofline-contract.md": {
+            "bytes": 7091,
+            "sha256": "419ec4bcaec45ac579a133bce083d4f48779f26e664d7a86f720fa65f343b1d1"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/roofline-example.json": {
+            "bytes": 2015,
+            "sha256": "597cd3205df043cff3ae9d15924c77934e8d8eb300c00c4840f6bb2921f90faf"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/scripts/roofline.py": {
+            "bytes": 9004,
+            "sha256": "524de48a6764401695fe1444f1fa3d38f363069b0cd2fba4715a86ca4b28a5d4"
+          },
+          "src/inferrsi/data/skill_library/methods/plan-activation-residency/SKILL.md": {
+            "bytes": 3507,
+            "sha256": "542d8a1e6963f5019986e1fbd79d3a247a045a51487c09313026087ac457c15e"
+          },
+          "src/inferrsi/data/skill_library/methods/plan-activation-residency/metadata.json": {
+            "bytes": 478,
+            "sha256": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7"
+          },
+          "src/inferrsi/data/skill_library/methods/plan-activation-residency/references/source-notes.md": {
+            "bytes": 13907,
+            "sha256": "5ff52655cfd06fcf970ed2a7d4d4fe1c31115b102af2d41497de0c74848ed2b3"
+          },
+          "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/SKILL.md": {
+            "bytes": 13800,
+            "sha256": "7e20d3390f2501263e5e298e026c47cac16c234f2397833131e15ae5d34c0b97"
+          },
+          "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/metadata.json": {
+            "bytes": 1903,
+            "sha256": "d0e4d6950250e4cec43d82e23324c9301a925d4c22be30b4d2713d790cb6504b"
+          },
+          "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/references/capability-presence-audit.md": {
+            "bytes": 11461,
+            "sha256": "c21eb4a6aabe811a49a6540a34c54dbcfdcf7efae3052a79b54034c7838026a5"
+          },
+          "src/inferrsi/data/skill_library/methods/skill-evolution-playbook/SKILL.md": {
+            "bytes": 9448,
+            "sha256": "9cd37f21fe0d862069e4798edca8a04d59f8b840e55d521901594a63dddd9ba3"
+          },
+          "src/inferrsi/data/skill_library/methods/skill-evolution-playbook/metadata.json": {
+            "bytes": 527,
+            "sha256": "92b36338b9257e2e41b95e4370694d7de52ed84bc7971056b3bfced07b0993d8"
+          },
+          "src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md": {
+            "bytes": 6865,
+            "sha256": "64bd2c6b795d966045d1d3cca55976d3d49b08b06df7c7e06a70c2824cf55b7e"
+          },
+          "src/inferrsi/data/skill_library/methods/training-input-pipeline/metadata.json": {
+            "bytes": 587,
+            "sha256": "80659099fce6a6ec05e9084b75452c877a5812e30f60a78f31ef8c7efcdc1bac"
+          },
+          "src/inferrsi/data/skill_library/methods/training-input-pipeline/references/sources.md": {
+            "bytes": 2637,
+            "sha256": "60e0f3e88655d8e1d5eb4797f39ca4370deda01614f265b8d2ea4559d8ac020f"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/SKILL.md": {
+            "bytes": 7200,
+            "sha256": "bdfa5a4ef37d9f303c3a23cd6f98c21137b97ac5ed58e0ed5fac6d71d33e52ef"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/metadata.json": {
+            "bytes": 976,
+            "sha256": "d42b614af9cee6ee0d4a25dc5d57c7b98438feed1e807a7c9f1755109a86e9cf"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/mindspeed-historical.md": {
+            "bytes": 3384,
+            "sha256": "eb97f935ed30517a8fc9edec5e9cddc04f5d679d156c16d75bd29c534593360a"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 7368,
+            "sha256": "c55c6ad2131ef788866411dcc7056f221420ebff30c95c6bc919b09a4818cb73"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/cases/paper-grounded-cases.md": {
+            "bytes": 5649,
+            "sha256": "19bd25b1a944adb2351ebc6b1d795150ea033041a44f6cc261e881677812a101"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/knowledge/offload-delta.md": {
+            "bytes": 2026,
+            "sha256": "f8740fc5c8bbbe5ac13ff1cfec7ca499569d0b46e8cd56cf1f5a44c3a5e72242"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/notes/training-memory-scheduling.md": {
+            "bytes": 18660,
+            "sha256": "9aa6d3e01452dff95c8a8940ff62ead8c146ce215c9175b78c9f2ef508a7b2de"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/references/source-readset.md": {
+            "bytes": 22946,
+            "sha256": "2a5c166767f56fa47e7eba7386f1f36da8cb73702e5b5bbe14d6a456ca9ebfae"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/sources.md": {
+            "bytes": 2752,
+            "sha256": "6d248f58af8b350fc66c22898988d6f451ac60d03bcd805cd6e84df5e55174ac"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/stas-decision-notes.md": {
+            "bytes": 4086,
+            "sha256": "f1a993e6c4ca9a035a2effcaf5500081f227b0cc56cc885a6b22fda0041fc744"
+          },
+          "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md": {
+            "bytes": 25969,
+            "sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213"
+          },
+          "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json": {
+            "bytes": 2418,
+            "sha256": "798cbff6f10b7c24b188080ebf233b188b0e98ef2c6fdb5e00e1fe3eff727c0f"
+          },
+          "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md": {
+            "bytes": 8656,
+            "sha256": "5a39f14b102b0cee304541e011aa31e44e6b13527d1bee9987c3f3b66c064d3d"
+          },
+          "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md": {
+            "bytes": 6392,
+            "sha256": "6c640f28ee3f0253a17b81461c865cb9f947367002cb75bf72550c4572d02e97"
+          },
+          "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/SKILL.md": {
+            "bytes": 7181,
+            "sha256": "a46110891eca860915e314d6921a22856d631453b22b17837fe30262d6780535"
+          },
+          "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/metadata.json": {
+            "bytes": 2175,
+            "sha256": "4e6100dfb9fc7bc00b19915f4fd38ad509704f2a609c658adf4d632d4e05b6c1"
+          },
+          "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/references/afd-plugin-LICENSE.md": {
+            "bytes": 12747,
+            "sha256": "0e8566ea362276878e1f38f97da49964dbea301c15d4aa9d727f97a4c39fd943"
+          },
+          "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/references/source-notes.md": {
+            "bytes": 17240,
+            "sha256": "5073d87450cb60d1b245308dff8807e73749c07de976d50f2fa26694fad81a72"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/LICENSE": {
+            "bytes": 35149,
+            "sha256": "3972dc9744f6499f0f9b2dbf76696f2ae7ad8af9b23dde66d6af86c9dfb36986"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/SKILL.md": {
+            "bytes": 4507,
+            "sha256": "c993bd6ff8e9b5f82f39f175ec8242341136e4ee0808af414e1d94ad3e196c5e"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/metadata.json": {
+            "bytes": 2032,
+            "sha256": "74742851a002a51c208362cdcd2a59fdab5d1734d8492b1f93fbec3cf21e0841"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/references/graph-cache-invalidation.md": {
+            "bytes": 7679,
+            "sha256": "614c9a4e6943d24d7d7be7d632682ce8b1dcc239124a37d9dc1d487113772eac"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/references/model-residency.md": {
+            "bytes": 7191,
+            "sha256": "014c68386e7cdfdb7b4f566836e8b2ff61bd5a5a506e7b4723f83fe80c5d4de7"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/references/repo-provenance.md": {
+            "bytes": 5648,
+            "sha256": "ca996ef515af42a3ce304747fa3f12a31c00533deaace5ba2ce21aefa0a62e2e"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/references/validation-boundaries.md": {
+            "bytes": 3523,
+            "sha256": "1913dedbeb558d872d6a0e015cc11d822910c3cb259a236db29fb60ad8a64263"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md": {
+            "bytes": 12294,
+            "sha256": "4703beb6e5194ef2ff2c0b5ec69f64e35417a4bc0dfb50a10c80a920b77ef3b5"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/metadata.json": {
+            "bytes": 2333,
+            "sha256": "014be24d0e669f4acbc0717dc323eec784026cf185186619d5a19d6d84496131"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/source-notes.md": {
+            "bytes": 12036,
+            "sha256": "fd917bd5ac80bee9c9f4525f8e0300f6d67dc144e2a16632af42a67f3d53d098"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/validation.md": {
+            "bytes": 5402,
+            "sha256": "60e71e70c5519bcdf6c171eec2be2f342e8801bedd746d6005f11adc646dfaff"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/SKILL.md": {
+            "bytes": 11202,
+            "sha256": "f157562d1edc905cb4fc07246567e3c5b844efb9fbf5fe39f0d40553a1b1a307"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/metadata.json": {
+            "bytes": 2156,
+            "sha256": "4d68bf4c1767eaee12ade963bbd551d0e61d1bcd81a984bb02fd3aae46965919"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/references/source-notes.md": {
+            "bytes": 11633,
+            "sha256": "6fa955022f16cd598f44fd6521901d1cb868b06cec545cabe2b82769a641b5d6"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/references/validation.md": {
+            "bytes": 7196,
+            "sha256": "673d93f6700052974b340fab25b3438e0e6344960569ac9abec7d7720f662288"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/LICENSE": {
+            "bytes": 1077,
+            "sha256": "663e51edcd9a3a8f858018ce6924ad2895a24c269c26c39ad75325bcd0f7e5bc"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/SKILL.md": {
+            "bytes": 5432,
+            "sha256": "fdb4d9483c6efa836f9e59617da605fa9c2186e11c2041b05d5c81850ba0475d"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/metadata.json": {
+            "bytes": 4923,
+            "sha256": "fe626a77af3f084f36fd73ffdb6f67976e7dceeafb94ba22078ef03c3a96cfae"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/THIRD_PARTY_NOTICES.md": {
+            "bytes": 1774,
+            "sha256": "a0d77db7d2aca0eda811ab3c386cc32b3143f2c2609a042cf4642954f0f6fb09"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/minimax-h3-apple.md": {
+            "bytes": 12059,
+            "sha256": "57ce01754ba96c378f13a91305ef3d418b926f99b2753f0af99315572bce063a"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/minimax-h3-weight-inventory.json": {
+            "bytes": 10664,
+            "sha256": "fbaacaeec340d3ce893556ee864405a0746af4d13530a38cb3d5aa9d60929271"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/phase-residency.md": {
+            "bytes": 9408,
+            "sha256": "2836a550eb2be7f5a862f51ec7b135ebd9715bf3c4faa84a7ecbc9d9c11f5d8d"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/precision-and-streaming.md": {
+            "bytes": 9440,
+            "sha256": "7e166c89193f4c7aed852ec95346a2de2b5008dac002e2e1862be1b3e57f6093"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/source-identity.json": {
+            "bytes": 7747,
+            "sha256": "730f8dd23648a3f8f47b9726cc613d893a79d1817d515eec609a9b06d8a9ff90"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/source-notes.md": {
+            "bytes": 9448,
+            "sha256": "7304bf9369efc157dc7aa40686dd828cc14db45c49442e908b3d3cc4b329eb86"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/validation.md": {
+            "bytes": 7880,
+            "sha256": "0754873a758e048480938314894e25b504ae0c269549312e7f8e6020d83fe294"
+          },
+          "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/SKILL.md": {
+            "bytes": 9338,
+            "sha256": "95e81b6bc096a954fa73139ea5bc001d1efa6c99747161da62f012b494e5c013"
+          },
+          "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/metadata.json": {
+            "bytes": 1718,
+            "sha256": "1d30b31dc8352fc6a8d7b8abbef8118c840e47e5b90795dae0c758bf9717278c"
+          },
+          "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/references/source-notes.md": {
+            "bytes": 15234,
+            "sha256": "3d12a401e9decb55a2f4f3b2bfa0d360ef5110f63a0469dfe285654e78054190"
+          },
+          "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md": {
+            "bytes": 10812,
+            "sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27"
+          },
+          "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/metadata.json": {
+            "bytes": 572,
+            "sha256": "c221189b8d85fac3bd41ef1c7cd560f1aebc8cfa0381c182d1d571a7169be829"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/SKILL.md": {
+            "bytes": 3788,
+            "sha256": "bf1177cb9985c12a9bf11edb64771018e1e314c4713920de8fd4a989ff70115f"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/metadata.json": {
+            "bytes": 1299,
+            "sha256": "d86cecb615e93ba0f6be09d1ead102cd5a17068c07e1be2549ac5cd8f7519279"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/cache-stop.md": {
+            "bytes": 2026,
+            "sha256": "bba6b5c73261065309cbcf11b1e591eef48916f0c79ceaaf74d98a9fa96b0992"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/generation-batch.md": {
+            "bytes": 1733,
+            "sha256": "f16d3c57d45e63bcaffd3871aa41c76972ee11f7dbf1e12b45d9b170303cad9b"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/source-map.md": {
+            "bytes": 6358,
+            "sha256": "91d26181f246307f2113f94b12ac9587d689b654dcd216e9f14b11cf5ebb650c"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/validation.md": {
+            "bytes": 1665,
+            "sha256": "73b5dd347017baba519ecc85415aead9993c1751d116b9ba4f07689f58c5f783"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/LICENSE": {
+            "bytes": 11352,
+            "sha256": "a54c8553df5d4313d293d1aef3bf2f13952bda8b7c7a1fdd729b2bbe27ee6015"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/SKILL.md": {
+            "bytes": 2304,
+            "sha256": "d75f5b2881b240ec2d9d4f5cc8e3f7d53e4378cb01732a2fb9699da2901ba38c"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/references/repo-provenance.md": {
+            "bytes": 1778,
+            "sha256": "5e8dbe168a8885e209d4b4532944ffa7fead658b23fd6588054d57b22476f63b"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/references/repo-routing-metadata.json": {
+            "bytes": 511,
+            "sha256": "6820707cb9bd877d1635928cf5a0efe38e4fe37dc04036dc70d53e479ed9bf7c"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/references/snapshot.md": {
+            "bytes": 3488,
+            "sha256": "0fb5d776349de845dc314f16ce73a06c9caf5db84dc1c293327d3b3ca1337c80"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-architecture/SKILL.md": {
+            "bytes": 2355,
+            "sha256": "766b2d66d2056dd779abd9e492cb3af6e25899298f44964f565509f130a7cb24"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-architecture/references/runtime.md": {
+            "bytes": 7734,
+            "sha256": "36bba9874fe054acc5217f9c265f5ae309c83943e0d46228370ae1bcdc69421e"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-extension/SKILL.md": {
+            "bytes": 2193,
+            "sha256": "9af38bca840d0004fc2ed049141476fc40fd40dcfa843f38a404a7a314c19425"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-extension/references/integration.md": {
+            "bytes": 5775,
+            "sha256": "6ce9383ca17bacff60a180f8fb8c43ea47eaa37117fb8da6051ad86135d228ef"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-serving/SKILL.md": {
+            "bytes": 2335,
+            "sha256": "990b6bdebfe22612de49dbebcb16810e8abef94d70691aa883b7937c85210eb6"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-serving/references/recipes-and-tuning.md": {
+            "bytes": 8918,
+            "sha256": "94afb762bbe185f58606154b3f0a5af56e7089cb84ea94e26cfcb9f93c95439f"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/LICENSE": {
+            "bytes": 11346,
+            "sha256": "1495e1e757ef4d0925a2350563cf5754bb23c51701a8ec4fb3c5cdcbedae6747"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/SKILL.md": {
+            "bytes": 5466,
+            "sha256": "ae72d1f0cc750699bf79d7a138a97219209bca7b7e5f19cd427d306245eaf841"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/LICENSE": {
+            "bytes": 11346,
+            "sha256": "1495e1e757ef4d0925a2350563cf5754bb23c51701a8ec4fb3c5cdcbedae6747"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/SKILL.md": {
+            "bytes": 5738,
+            "sha256": "b0a4ba4751087126143ef86983798deb0ed00e24b6ff06bf6ef76a9df6e274f2"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/metadata.json": {
+            "bytes": 2844,
+            "sha256": "6ed57dfb4f25ecb34b81f2188b055e05d172ba2581e6a3b605b814b2f733a4d6"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/references/source-notes.md": {
+            "bytes": 19512,
+            "sha256": "d0fcfd1177ae2f351041c5b2abba7f5121cec133006d9e4dca8f74ddb567332d"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/architecture-and-memory.md": {
+            "bytes": 7735,
+            "sha256": "798e747022b7ec135cc4861c2c9ac44fb14b503bb77a0068ee7fbea494b764a7"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/native-candidates.md": {
+            "bytes": 3961,
+            "sha256": "abafcc9a34df196a6365145554887b8d37afb36c6e236c92e50a4a06c8a4ec93"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/parallelism-and-extension.md": {
+            "bytes": 6324,
+            "sha256": "e3be7f4b51264ac68682a64d264e4d96857d5bb9f9d99cae5f327c56984a9e6b"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/repo-provenance.md": {
+            "bytes": 2274,
+            "sha256": "534d95ae8d0b29bcbc8f6707b3d38fb964e31f5b66ba13517a9a6b8f309bbb38"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/repo-routing-metadata.json": {
+            "bytes": 625,
+            "sha256": "b2b879c1cc42ddc2a5b95ae24115eb891562fa9a6e96d151f4ddf6f7719fe73c"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/source-map.md": {
+            "bytes": 8945,
+            "sha256": "c2adad64ceb0c305cdc4859c03012b3e0a4ab26cd0cb289bbd76e192b0cc3794"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/tuning-and-validation.md": {
+            "bytes": 10832,
+            "sha256": "49b3343b3292cc098a787de9d24d968b444d0ef4f73fedb5d8183390e8f39c9c"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-parallel-extension/SKILL.md": {
+            "bytes": 2316,
+            "sha256": "18bd54ebdd629e1ba6dfa589f391c7cc3c5355785c2ae82f741b64ee8022464f"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-runtime-memory/SKILL.md": {
+            "bytes": 2154,
+            "sha256": "33b1a65aba00fca819d2b6cc42202e7708e9155e4d8f4373a2387f4ee2f2fa3e"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-tuning-validation/SKILL.md": {
+            "bytes": 2018,
+            "sha256": "744b2ef9122101d3f50cbbb40fe020e2b3d3888adbe2d86041041e7c7ce424fc"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE": {
+            "bytes": 1062,
+            "sha256": "b53fa08f515cb5a6fff7b9fd8fcd0961a4b80df29d74372d25e1f9171aa042ee"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md": {
+            "bytes": 5504,
+            "sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json": {
+            "bytes": 2202,
+            "sha256": "f7b7581a3f44ba9774480c559408fc66d6a4d6d1dbbeabe4529f1eb7d4dd5ad6"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md": {
+            "bytes": 8136,
+            "sha256": "7b537cb972347dd68e303960f142892346d8b19b884e4d08016ea36a23bdaa7a"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json": {
+            "bytes": 3617,
+            "sha256": "af4830370b6eddc2d176a9ae14da4ab56a90b0fe048a01a89ee051df4365d8e9"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md": {
+            "bytes": 6589,
+            "sha256": "b67affbd7d237d82b7f51be49f1537e9820fe56ec70a71ee76809f9e599ca693"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md": {
+            "bytes": 3394,
+            "sha256": "5dcf5fa7bf734fbe13214f57d2dfb40ff8215d9cd34e072a9e615bf103b82fc2"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/LICENSE": {
+            "bytes": 1093,
+            "sha256": "949b9b0aada7476006469e58ee4ffb0e0f68c2d9137e90a356757b61a09d37f0"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/SKILL.md": {
+            "bytes": 2813,
+            "sha256": "0e5daeb6409c916be45ecd7ce2531ea5f1c60a4a53e2b0ed25dc1e2cdc5c17ed"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/design-space-analysis.md": {
+            "bytes": 27736,
+            "sha256": "11ecafe71530ddd4839268ea1b24b3361a3895012a705859ca06dae29fa72630"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/repo-provenance.md": {
+            "bytes": 2913,
+            "sha256": "05f17d0ae996ad6c12314bdd29f3427574f6aa611ea60effd18598a0f386638b"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/repo-routing-metadata.json": {
+            "bytes": 332,
+            "sha256": "a10ba49b7e6565109b566c105e17551d042075a1ff649c6f371134e3b7402c74"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/trial-format.md": {
+            "bytes": 2291,
+            "sha256": "92d8cf38549268ae4eef382890662d03e18a7dff858ad45ee5eb6f47eb39c8c0"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/troubleshooting.md": {
+            "bytes": 2079,
+            "sha256": "7301cfa5a5f6cfbf9576c3845a96c6b4de8f8a6509f2330970a8e2e4669b3d5a"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/scripts/summarize_trials.py": {
+            "bytes": 5170,
+            "sha256": "6a243a08da15659ced6519f49f52a49a675bd9e952740c3ab03dd21964eefbe2"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/design-space/SKILL.md": {
+            "bytes": 1826,
+            "sha256": "857bea3c64db45dc035728a41523142ac65778b41f63092148b8ab936f971d76"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/design-space/references/design-space.md": {
+            "bytes": 24231,
+            "sha256": "0a428dcd12d5290b554e9da6795a110f94d98f128771602e1ea5f608048b135e"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/tune-execution/SKILL.md": {
+            "bytes": 1145,
+            "sha256": "cdf51c537e7237296d52ef6cda9f2e4c9d9ce3fc87265a369cd9eb7497c53512"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/tune-execution/references/tuning.md": {
+            "bytes": 11310,
+            "sha256": "7b69aa7872d60dd3fb91e41b77a632572070dcf80dd5b5f6dc1722d8540177b1"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/validate-engine/SKILL.md": {
+            "bytes": 4613,
+            "sha256": "90bfd975e25f18895ce0e85f6f35d24149ce026305a4d479aea4924510053460"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/validate-engine/references/validation.md": {
+            "bytes": 16895,
+            "sha256": "dae846f4c47c68731b96ec1d50cb008c03871c72078f1c5b0f4a9423be9ba397"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/LICENSE": {
+            "bytes": 11357,
+            "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/SKILL.md": {
+            "bytes": 4945,
+            "sha256": "df3e63dc28de2dac2aa92824fd88719159138650d96487332e0effca3f0e2a80"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/references/evidence.md": {
+            "bytes": 4689,
+            "sha256": "ae0d51ba56b34ed81fba34de8fd3685bebf2f160b188fa53719936002881d220"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/references/repo-provenance.md": {
+            "bytes": 2221,
+            "sha256": "43bbfe431434ca4ad6157ba35cbb9ce2d89e4065f427abfdd5f372f51ba0c9c9"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/references/repo-routing-metadata.json": {
+            "bytes": 342,
+            "sha256": "7784b4d98d096718c4f4cc42ce86d5ec113c469844e12de4690b0fccca59d329"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-architecture/SKILL.md": {
+            "bytes": 1336,
+            "sha256": "fb3ff50a7b9122a2545a97ef5179011237c4c4ca214dde7236ad7bf79a7c8623"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-architecture/references/architecture.md": {
+            "bytes": 5600,
+            "sha256": "db845eb4207ddeedb8ae381f1b91704a42cb53ba0ee6c7ee581510ba1a412190"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-serving/SKILL.md": {
+            "bytes": 1343,
+            "sha256": "8ab3209b838635395b22deca6aa879172016531d45c2374eb8e498e7f71bb40f"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-serving/references/serving.md": {
+            "bytes": 5049,
+            "sha256": "9bd5eff300d9755137a7600c8f7a7753bb22d09c8853e08062d7e535662ba8a3"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/SKILL.md": {
+            "bytes": 2227,
+            "sha256": "93c83bc4b799b32f50ff54b870cbbde050a790af16d9b7b062039c4987c80148"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/references/abort-acknowledgment.md": {
+            "bytes": 8547,
+            "sha256": "e95ae5b2cc410917ddf5e9066890d0bef9264104529bdca5ac12bdf56fe80ae3"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/references/validation.md": {
+            "bytes": 4632,
+            "sha256": "1542061f9c76fff2acb6529a58238301484b0e23ec0c697d2fdf129e200ea973"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/LICENSE": {
+            "bytes": 11324,
+            "sha256": "50e6751797c50dedd75ef1b8a0d9e42f5f8472e9fbce91f34718e9f97b0c780a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/SKILL.md": {
+            "bytes": 2863,
+            "sha256": "38e522b8abeb0dc140a2194340432ed41dc8b569d33b63d5d4c59aa1092cd2d1"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/SKILL.md": {
+            "bytes": 6664,
+            "sha256": "313733a65cc48442b18dda1c4665139a689963a90e1032a2235bc6f40e3a8e06"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/metadata.json": {
+            "bytes": 781,
+            "sha256": "b43fc3a7ccae59c664fcdf9ff77962edcc07c2a3df8b59324db551642bd23e23"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/references/source-notes.md": {
+            "bytes": 13853,
+            "sha256": "3665b034594fc8cc767c408002f4e16f8803369242405027c7f37ce6c9ecebe9"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/SKILL.md": {
+            "bytes": 6565,
+            "sha256": "5fb9ea1d1373e110d9b506f69963c4857338b7dbf5e38ec9a49b59a0ff683ed0"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/metadata.json": {
+            "bytes": 1372,
+            "sha256": "78459416734f61a8faa819d9a1a3ad7b9a985e8bf8e149813c604d15be6bb92f"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 9006,
+            "sha256": "82b24be82451ee121892f61575d0cf4ecb9920813a25ccb26cce2b45dda6d8b3"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/migration-is-not-free-memory.md": {
+            "bytes": 2145,
+            "sha256": "2ac168f2725cba7fc92dfcc67c01bf2258dff6a2dbe1ddb063bc60c175d82d0a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/parallelism-change-with-state-cost.md": {
+            "bytes": 2069,
+            "sha256": "5accc81086e4500d66f894ef6af9951e1cbb76941f6c631aab5c5561a786d8d8"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/preemption-under-kv-pressure.md": {
+            "bytes": 2630,
+            "sha256": "462200a6d0b193ef01f32374a463e0f2838ca8edf33d9ab898def65ffd0b490a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/knowledge/scheduling-state-boundaries.md": {
+            "bytes": 7488,
+            "sha256": "d4ef29869959b835a5dcb06ddb2fd8e6cff49bb176994dc1d46eb9a852f6228d"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/fastserve.md": {
+            "bytes": 9289,
+            "sha256": "9b4e275843000ef453d508d9c3c9c3f8ea1de5545fbd01d7abe74baa8a4931a7"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/llumnix.md": {
+            "bytes": 10181,
+            "sha256": "f252ce33ac2db017701ca1d49813c5ec1996c4c92900cf35bb6bda7fc59c0f18"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/loongserve.md": {
+            "bytes": 12422,
+            "sha256": "7bc6d1e256d461b04acc8ed856cf99108faceda9cde729ab95bcca596fe019ca"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/orca.md": {
+            "bytes": 8826,
+            "sha256": "f5a3b5870d64226408a8fcda5d69483769d823f8a5efd7f11698c8da483ab195"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/fastserve-readset.md": {
+            "bytes": 7173,
+            "sha256": "9866bd0c78b2b638998b3b632f6bae26aef032462c81abee336d309af5b07831"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/llumnix-readset.md": {
+            "bytes": 9386,
+            "sha256": "3a2819d65ad24087c08530492ef39d8ea046ed1f780d53512f1d23e2b3578487"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/loongserve-readset.md": {
+            "bytes": 12490,
+            "sha256": "f1a383e7bdbbd5a5a472d7f6e1292c8a17d4910ff736ca5a6aa422e1bee74f8c"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/orca-readset.md": {
+            "bytes": 4796,
+            "sha256": "a497ef23b7e60c0fbc15d95b7e07cd20547cfe56f37c24dfe6ed70d79ef7a3cf"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/source-notes.md": {
+            "bytes": 9575,
+            "sha256": "84033fd514773f10c0bddd9914fabb980497ba2b308f226a922c3e1888413827"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/references/evidence-map.json": {
+            "bytes": 14384,
+            "sha256": "dd8bbab5d285ed3a1a19d2f06ad84e8c444251690d8f673def3e2681765ef6c1"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/references/repo-provenance.json": {
+            "bytes": 1313,
+            "sha256": "4e705a964697712e950e02bedc41320806a41311b7f4dcf32f923d60b1d2e919"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/references/repo-provenance.md": {
+            "bytes": 3261,
+            "sha256": "6e841c38c35ff2cd6ff8139cb1a6c8f75701647897f1089a94a5c62928609941"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/references/repo-routing-metadata.json": {
+            "bytes": 432,
+            "sha256": "a20b54f682eb52311e445ac62b95d1935c001296f5cf23b7764cb04477d535c9"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/SKILL.md": {
+            "bytes": 1835,
+            "sha256": "2a945b5fddacfb90d1cd7cb2b2f767067d641061423a65e1c19afecec7f3c06a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/references/capability-boundaries.md": {
+            "bytes": 6560,
+            "sha256": "c8081d03f04f7481080c46c7dcf728962729bc122f49f9502d50cd1e3d04bcf3"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/references/runtime-and-invariants.md": {
+            "bytes": 8265,
+            "sha256": "afa1f967545021f36bc1b59dbf6aa2c367f4d5ece4f08dce755e2017cf50dbe5"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/extend-validate/SKILL.md": {
+            "bytes": 1731,
+            "sha256": "495679e5b99341bf80daeaacd2d4c2699aba16bc16c447cbd8d2fe17f7559798"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/extend-validate/references/change-boundaries-and-validation.md": {
+            "bytes": 7624,
+            "sha256": "f7fdd2d699f09d96fbf5ca9382f9b56113c870db2688d8e1118eb8163519fc4a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/SKILL.md": {
+            "bytes": 1812,
+            "sha256": "0ef4d2e3a96cbb277557e169a3fc8edf0f0504aad03259771b432a33ee008f78"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/references/commands-and-measurements.md": {
+            "bytes": 5667,
+            "sha256": "76ab51083b14595c49fa1c255cb7e5ebb996f14930a51ab3cbe864cc334b5f80"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/references/experiment-decisions.md": {
+            "bytes": 7644,
+            "sha256": "242593c0d386e57104b57a09bc7db97c2bbbcdb905bb68a397cfe612e07fb083"
+          },
+          "src/inferrsi/data/skill_library/skill-index.json": {
+            "bytes": 7514,
+            "sha256": "8ce35e18f7629260d67a95617b9918eecff73ed8d9370dd3d4595c725ada9aa1"
+          },
+          "src/inferrsi/data/skill_library/workflows/cite-check-downstream-pr/SKILL.md": {
+            "bytes": 1093,
+            "sha256": "485932cdf3b1fba5dc69be73e75d38f86a7b79442d26e82a296c963a7598d226"
+          },
+          "src/inferrsi/data/skill_library/workflows/claim-contract/SKILL.md": {
+            "bytes": 1075,
+            "sha256": "3af1dbbe81d91cb05fc2ef15a2ae25d7592a08558e1a5535f49c52f33a488c55"
+          },
+          "src/inferrsi/data/skill_library/workflows/competitor-scan/SKILL.md": {
+            "bytes": 1088,
+            "sha256": "0ede58ace65a9324a538b4816add1684f230ba536d4364fe5a5dd2c0bbed3776"
+          },
+          "src/inferrsi/data/skill_library/workflows/draft-acceptance-issue/SKILL.md": {
+            "bytes": 1230,
+            "sha256": "ce983acfb7d7d3272775e7cf00c92e4ddadd97511b602575a76e5ab7ade01b56"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/SKILL.md": {
+            "bytes": 5681,
+            "sha256": "364beed2ca617e71f4f955bbd024740d411b94eeb9557d7e055d4463cc22caba"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/cross-repository-mechanisms.md": {
+            "bytes": 17050,
+            "sha256": "28efa1b25f548763180b70d8d0ddebb7d4e7227df189f871fd233fe1d4881303"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/import.md": {
+            "bytes": 3463,
+            "sha256": "c8a57db94768cc6cd4193932def75ba67ec453e56ee0e54f176fb82f1af53091"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/model-aware-preparation.md": {
+            "bytes": 9056,
+            "sha256": "787c017aade68ee1133b979d636fa24062302cef6174c001171c963947ee8da4"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/overview.md": {
+            "bytes": 8099,
+            "sha256": "7ed0a966239b9481b82dd349245cfbfe966c030fc916771c91fd0692464ffc56"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/repository-source-scope.md": {
+            "bytes": 11967,
+            "sha256": "85202ca209171c50573c2fc209acb1e937d8211e1652e4ca9c6cc4bce04d03eb"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/source-manifest.json": {
+            "bytes": 1613,
+            "sha256": "0ce1678803d1ede6b485fb6b4b1c88aba909056ee6e3b5af691899ceb2056240"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/SKILL.md": {
+            "bytes": 6502,
+            "sha256": "74cd863df2e80dc2bda1713d24207809173a3116f30325e3487e1e3517cfec1e"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/metadata.json": {
+            "bytes": 3236,
+            "sha256": "12fc9272bd5fbf02dd9b897d8ca40984b72ae54f20492ace5304711bbc4d71c3"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/COMPARISON.md": {
+            "bytes": 4267,
+            "sha256": "38587678a96581f5c2dc697d982127343565504bffa191352811dffcc165aa23"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 5557,
+            "sha256": "fbc4a96b6f6c477706eed83eaa4fcc29a14f881637629d6c004601ed049f2523"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/QA.md": {
+            "bytes": 1853,
+            "sha256": "3f3785ec39a6938a5d1f2e874e99153fe92c0e666dd3e78a499c8b47d891346a"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/README.md": {
+            "bytes": 2151,
+            "sha256": "221785e91c45588f82b3caab6d52f1ea1d102b16139c20a0802921e2789e6d28"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/UNREAD_LEADS.md": {
+            "bytes": 1038,
+            "sha256": "6c1d87ebe491d27666703149eb05ad582c7e5aff77f0ac5c11a7b89c946f96ee"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/cases/methodology-counterexamples.md": {
+            "bytes": 5213,
+            "sha256": "f9fc21753ea41eef77ca020d78d031e945e5dd835b5d09c5ff0e53749ceb2ca4"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/knowledge/evidence-routing.md": {
+            "bytes": 5282,
+            "sha256": "406d5d74c3fef4c1a23d495492e21a084c3c46495a1a84b385703ed104358874"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/learning-to-commit.md": {
+            "bytes": 2847,
+            "sha256": "9c2d1d738e896d05c00f53e6adf14c9df2501ecc3f0020e71b5a1c2192547086"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/protocol-details.md": {
+            "bytes": 6520,
+            "sha256": "41298c6bde3a28f0712fa06a18bbc4ac59aea6d17cc9eb1318744f4121e3bab6"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/readset.md": {
+            "bytes": 7086,
+            "sha256": "6f953d20d8958c68339c9ceeae348604d0fea3e9c6ccb9b64a63d3b210426a22"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/repo-to-skill.md": {
+            "bytes": 3151,
+            "sha256": "5ebdac2a2bbbe6614f9d4096bc9895bd685acdf76414e2c3c1fb7b50d57c738d"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/skillopt.md": {
+            "bytes": 3040,
+            "sha256": "148075daa67e19f4b4351710498438d17c279a9bce56b2d01cab26ff3be33e99"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/source-manifest.json": {
+            "bytes": 2431,
+            "sha256": "716cd49ed01824ee9f6b6fc55ebf2b24fdd5d5a3e677f6e28e34a472616bbe25"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/trace2skill.md": {
+            "bytes": 2917,
+            "sha256": "230d13b77ba819b5e3edee147523e8cad6bfce95afccc140fb4946d6c8747a2e"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/updates/bounded-updates.md": {
+            "bytes": 7630,
+            "sha256": "d02c5bd0109846d1e1987de074810321cb4af277cd3f66ab2ac268aadd5ee142"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/skill-factory-design-20261014.md": {
+            "bytes": 9465,
+            "sha256": "fb8ce1ad83b7a6ddc657e89ee2eebb710873b213cfd85bdd6d823c8691b3cb9f"
+          },
+          "src/inferrsi/data/skill_library/workflows/rsi-evidence/SKILL.md": {
+            "bytes": 3680,
+            "sha256": "a55610874b658b86a39a1a049efb4e4eab0140f6579ed74dd7718a4ca8cedaca"
+          }
+        },
+        "sha256": "6e19365a0ad8e2951af0f3106f93aa545b4ad00b6e0f0317a20f32065b6a05fb"
+      },
+      "environment": {
+        "filesystem_cache": "not controlled; process cold is not OS-cache cold",
+        "implementation": "CPython",
+        "logical_cpu_count": 8,
+        "machine": "AMD64",
+        "perf_counter_resolution_seconds": 1e-07,
+        "python_executable_sha256": "737a7e3b71e3578f8432acc7dd88c452e593622c544bc13da4789d69c63da5ae",
+        "python_version": "3.12.6",
+        "release": "10",
+        "site_initializers_disabled": true,
+        "system": "Windows"
+      },
+      "git_head": "4817f92121df73fd9cd5fb4a319e3c56a68b4a5b",
+      "knowledge_record_count": 108,
+      "library": {
+        "bytes": 2294958,
+        "file_count": 342,
+        "sha256": "be8726296e6186601692dc2d95d99bfbba76f0b4c758de3a88a9b2a523eaa37e"
+      },
+      "modules": {
+        "inferrsi": {
+          "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+          "source": "src/inferrsi/__init__.py"
+        },
+        "inferrsi.catalog": {
+          "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+          "source": "src/inferrsi/catalog.py"
+        },
+        "inferrsi.cli": {
+          "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+          "source": "src/inferrsi/cli.py"
+        },
+        "inferrsi.knowledge_operations": {
+          "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+          "source": "src/inferrsi/knowledge_operations.py"
+        }
+      },
+      "packaged_skill_count": 60,
+      "python_sources": {
+        "bytes": 274444,
+        "file_count": 53,
+        "files": {
+          "src/inferrsi/__init__.py": {
+            "bytes": 102,
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47"
+          },
+          "src/inferrsi/__main__.py": {
+            "bytes": 48,
+            "sha256": "935a1c1166b0c1ea35a82256345000bf2c73ded718d77773bc27a71ecce28f7d"
+          },
+          "src/inferrsi/_skill_selection.py": {
+            "bytes": 5790,
+            "sha256": "4640399f4013425589f6702e2db844820619ba35967d1c3519978eb307168440"
+          },
+          "src/inferrsi/arguments.py": {
+            "bytes": 12200,
+            "sha256": "4456600cf4a362cf069594fe36db46932cf5cb14ddd37683aaf3538f5cf76b4d"
+          },
+          "src/inferrsi/catalog.py": {
+            "bytes": 6855,
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5"
+          },
+          "src/inferrsi/claim.py": {
+            "bytes": 805,
+            "sha256": "5f24c4fa91b48f0bb4d4554531607763fc8fc108e90f332816f06cfb520b5cc1"
+          },
+          "src/inferrsi/cli.py": {
+            "bytes": 14205,
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88"
+          },
+          "src/inferrsi/contracts.py": {
+            "bytes": 119,
+            "sha256": "03000cb9246ce8c0967147533f7ad25df918ba2674839801453ffb870f496490"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/scripts/roofline.py": {
+            "bytes": 9004,
+            "sha256": "524de48a6764401695fe1444f1fa3d38f363069b0cd2fba4715a86ca4b28a5d4"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/scripts/summarize_trials.py": {
+            "bytes": 5170,
+            "sha256": "6a243a08da15659ced6519f49f52a49a675bd9e952740c3ab03dd21964eefbe2"
+          },
+          "src/inferrsi/evaluation.py": {
+            "bytes": 183,
+            "sha256": "482c40710d428206c59fe3696703248cd86c4b3c7d19ad41017c1369bd6d780e"
+          },
+          "src/inferrsi/evidence_paths.py": {
+            "bytes": 859,
+            "sha256": "cc46231e0b8334ef5742defdda20e0093e22b30c86af59df1fb3ae4b5caa70fa"
+          },
+          "src/inferrsi/execution.py": {
+            "bytes": 755,
+            "sha256": "f43d8cbb4d43a72a7ff8a10b29a5fb9cf4cb22af54dd35943283c843fba0da92"
+          },
+          "src/inferrsi/harness/__init__.py": {
+            "bytes": 177,
+            "sha256": "1006e4713003f44bfb32b2e385a918e1652b9e9833801e4555674a71e57fb0d5"
+          },
+          "src/inferrsi/harness/evaluation.py": {
+            "bytes": 4160,
+            "sha256": "9b00b2e93959f16b7b76bb4b257df7dcf48520a06fd6adae96e9a165a05e6e80"
+          },
+          "src/inferrsi/harness/improvement.py": {
+            "bytes": 5777,
+            "sha256": "3faa0f5928a707a98f35f45d1ca18e226c898941e6a07b3829eebeb53e70aeb8"
+          },
+          "src/inferrsi/harness/reporting.py": {
+            "bytes": 3566,
+            "sha256": "38c58f02c44e8c8aeacf46527e142be0e3142a83e4b2693246d0ca76c0ddd7c3"
+          },
+          "src/inferrsi/harness/runner.py": {
+            "bytes": 6567,
+            "sha256": "a566307b9b586655941b627f428b5121ed38d7ac80bccd1e1799e443fb0de911"
+          },
+          "src/inferrsi/harness/scenarios/__init__.py": {
+            "bytes": 79,
+            "sha256": "4e97913c806077f0cadfeb29d986d2fa75b55ab8766cb86e6e3e9a195a9a1f5a"
+          },
+          "src/inferrsi/harness/scenarios/external.py": {
+            "bytes": 1016,
+            "sha256": "84d5f6920501a9d882cc4fffbc721d2051fdbf7c90694bde9cbfba3c2b866304"
+          },
+          "src/inferrsi/harness/scenarios/media.py": {
+            "bytes": 2914,
+            "sha256": "67d979a3ed6f13a8e3e1e40a304f78e5a3e6c2556b3aa8a11c41156745d7050a"
+          },
+          "src/inferrsi/harness/skill_evaluation.py": {
+            "bytes": 11093,
+            "sha256": "b710b8e53758757efd68c703449eb4bab7eb7ec50ae3dfc52a010cdebd8c3bf5"
+          },
+          "src/inferrsi/infra/__init__.py": {
+            "bytes": 85,
+            "sha256": "8779897733779266841b0f79b84b1dc43f0fa57c013855c906c809829ecfc77c"
+          },
+          "src/inferrsi/infra/artifacts.py": {
+            "bytes": 1040,
+            "sha256": "e72e1bebdbb88e42de430238489252386851f9e6645ed7780f2aad2e90338120"
+          },
+          "src/inferrsi/infra/campaign.py": {
+            "bytes": 1028,
+            "sha256": "be505fba1592611bd84ceaca55afbac027205413a2639baefa0c3bb343230209"
+          },
+          "src/inferrsi/infra/commands.py": {
+            "bytes": 811,
+            "sha256": "1890f115b3a2c5c6832173bf60432ad595bd6bf1c69783ecd460c6f8961194f1"
+          },
+          "src/inferrsi/infra/environment.py": {
+            "bytes": 2067,
+            "sha256": "60e46c2aefc216db202b41f44542e7b893e08e0721c667b9ccc50411d1ac84ee"
+          },
+          "src/inferrsi/infra/executors/__init__.py": {
+            "bytes": 64,
+            "sha256": "327c3cd5c821c832ff260a41dba0ee1e439cdabefc9edfb9cad1cd38d14bc87a"
+          },
+          "src/inferrsi/infra/executors/local.py": {
+            "bytes": 3654,
+            "sha256": "37bcbca9f3261c8c58a8bf806c9da40720b18d0676c82e8ed24c3901ac9bbe77"
+          },
+          "src/inferrsi/infra/http.py": {
+            "bytes": 634,
+            "sha256": "f65d6c2b55b6a89694caaf6dcdf7d74f417e9ced103eca1b3f218e305c6b6cb2"
+          },
+          "src/inferrsi/infra/inspection.py": {
+            "bytes": 2415,
+            "sha256": "d1c66deb0cced529bca996d2f88b60f51fa3417ff1cae365c650fec808946ecd"
+          },
+          "src/inferrsi/infra/platforms/__init__.py": {
+            "bytes": 1494,
+            "sha256": "9e88e8b9a396ed90c221e1585135c0a1b58da41c1bb73ed9237c53b6684abfed"
+          },
+          "src/inferrsi/infra/platforms/cuda.py": {
+            "bytes": 3309,
+            "sha256": "9e0e4276e1b2984605493b14b82fba3e86f6635204719c318302e624f66f8b6f"
+          },
+          "src/inferrsi/infra/platforms/mlx.py": {
+            "bytes": 1971,
+            "sha256": "f3b81470db28bd6183ccca2e763c5e079686bba6cdfe6c08ebe2cdf64216f2e3"
+          },
+          "src/inferrsi/infra/processes.py": {
+            "bytes": 8151,
+            "sha256": "3b96ebc7a50913e78e86ff3dcba74c8ee8d03d547d9591ca6488f390ac1e6189"
+          },
+          "src/inferrsi/infra/targets/__init__.py": {
+            "bytes": 186,
+            "sha256": "1f3ea75dded7c6cc967754c7b6d4f30ef00da8191a749938aabc3c6bb8f86eb3"
+          },
+          "src/inferrsi/infra/targets/git.py": {
+            "bytes": 2910,
+            "sha256": "6327c4f126086f310b6de309fc7659c9759da28ed76efaeeb94a3d2fba29a0c2"
+          },
+          "src/inferrsi/inspection.py": {
+            "bytes": 692,
+            "sha256": "afa44bf4cdfbb58a318c21f9b292ca142d1bbb22750e35a4cc8244da5350ebb3"
+          },
+          "src/inferrsi/knowledge_operations.py": {
+            "bytes": 27444,
+            "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b"
+          },
+          "src/inferrsi/knowledge_store.py": {
+            "bytes": 13700,
+            "sha256": "5e794e28906ce4687aff7f14a507e267184640fc8a4fac19e0ced279ce6a6cb6"
+          },
+          "src/inferrsi/model/__init__.py": {
+            "bytes": 77,
+            "sha256": "3c1dba4f5159fd9fbc8d4d96e2a83c8441491c1b1dc7b42031a3fc1782ed9d9f"
+          },
+          "src/inferrsi/model/anthropic_agent.py": {
+            "bytes": 24696,
+            "sha256": "b0ef9cd542544292eeb90ab05ef410be9ed6e5d1a40d729890f49562c28c12f0"
+          },
+          "src/inferrsi/model/protocols/__init__.py": {
+            "bytes": 85,
+            "sha256": "5a2f54483b164de04e309dbebb87a700b213682b0f44a50d89a752dd71a4ce47"
+          },
+          "src/inferrsi/model/protocols/image.py": {
+            "bytes": 2047,
+            "sha256": "251a1f8f885da2cd8e9d3baba766c525afe293091606b4e8668c43ba2cb7cb5f"
+          },
+          "src/inferrsi/model/protocols/metrics.py": {
+            "bytes": 610,
+            "sha256": "17db16d6ef15b41fb60d8f09507893accc43f9cb3ad686b8dc4e9839713ebc56"
+          },
+          "src/inferrsi/model/validators/__init__.py": {
+            "bytes": 65,
+            "sha256": "1d4b6ffa968e283219ddcc8a09de9280238a8d65ae2dc8cd2b60889877a8eb0b"
+          },
+          "src/inferrsi/model/validators/flare.py": {
+            "bytes": 19975,
+            "sha256": "3696b0609c772f2222b448833608565a033123d7e92b11fbc545dd356fec590f"
+          },
+          "src/inferrsi/model/validators/image.py": {
+            "bytes": 3204,
+            "sha256": "79470b6b06a22d53ec6c40636a0b3d821a2e6ee2f7bb7aed3dac7241504add1a"
+          },
+          "src/inferrsi/reporting.py": {
+            "bytes": 127,
+            "sha256": "1f771890ab3e69b1aa9246b548a7825596bdf904f6c395dd218208084266853d"
+          },
+          "src/inferrsi/research.py": {
+            "bytes": 5213,
+            "sha256": "9bfe59dd17b5476c18ce50d671711053c62800521c50d6b710e8140a2b4dca04"
+          },
+          "src/inferrsi/runspec.py": {
+            "bytes": 4398,
+            "sha256": "dca0952425caf42095ef0c7e2feb8ef4740cc422ecb61e777f203962d9b87709"
+          },
+          "src/inferrsi/skill_catalog.py": {
+            "bytes": 9938,
+            "sha256": "65f5a1f3f6894009abb98ec2fec30a63cc1bb2f6ce6aeb01f9f7a3dce0439f2d"
+          },
+          "src/inferrsi/summary_chart.py": {
+            "bytes": 40910,
+            "sha256": "d29023d60f6be804c1b01c9eb73765d4824d6c3bb870fe22c99a24544bc1d8bc"
+          }
+        },
+        "sha256": "4faa529dc47f23c24f8c5b8e0110d8e413aab20df5ad61db591fb751564a6448"
+      },
+      "tracked_source_dirty": false
+    },
+    "B": {
+      "corpus": {
+        "bytes": 2883003,
+        "file_count": 458,
+        "files": {
+          "src/inferrsi/data/knowledge_packs/README.md": {
+            "bytes": 14172,
+            "sha256": "b877e67d9cccf0dcbe57419eeb3ff9374e5bd03ff8eec63adef1025c446140e9"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/diffusers/cuda/diffusion/group-offload-transfer-overlap.json": {
+            "bytes": 3277,
+            "sha256": "9bec3d912192efa2f2c853d3fb70bdae21a75875630a02294fcd084b18c30414"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/flashattention/cuda/attention/generation-specific-dispatch.json": {
+            "bytes": 3085,
+            "sha256": "90290d7bc80b1626863e0520c91bb196d93ebfe08190a450e9de59c6505ae8ad"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/flashinfer/cuda/llm/paged-decode-dispatch.json": {
+            "bytes": 3745,
+            "sha256": "0f2986494fcc9b751df71e598fd61caa68e28bc4b0ac401e4cc763bc3f75283b"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/cross-stream-dependency.json": {
+            "bytes": 3809,
+            "sha256": "b829315e55adf8e11f0bd581d5f45dd130d7eb3f81ce7de532f63c7b9c27cbb5"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/evaluation-completion.json": {
+            "bytes": 3841,
+            "sha256": "19d2e506ac4f50895776a83a062059aa7a052175c8e4b21dd0d46e1a5a6716a1"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/post-detachment-buffer-lifetime.json": {
+            "bytes": 4663,
+            "sha256": "18d78c0e542dacb42ca33e5264d92aef80fa06d163fd16d4feafd5ed0827ab99"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/pytorch/cuda/compiler/cudagraph-shapes-and-lifetimes.json": {
+            "bytes": 3623,
+            "sha256": "7fef4122ff794967e61f55d02a4a6b558d7e498d9513607f5ca86db6386928a8"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/transformers/common/llm/kv-cache-strategy-contracts.json": {
+            "bytes": 3072,
+            "sha256": "a2df26052112a9f3fc6242b2bd636ab13991442db57e56646168d6c52032d38d"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/triton/common/kernel/bandwidth-fusion-resource-limits.json": {
+            "bytes": 2941,
+            "sha256": "6e9673d1a600a6ad34820e4325f2a6417f44f0f96d8ed0bcf21466d998f19ee4"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/cache_dit/common/diffusion/residual-cache-calibration.json": {
+            "bytes": 3841,
+            "sha256": "bb9509ad0225a91b52bc6787477581d05f0ecd5e44fa59f08f718eea36a2ac45"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/bounded-windowed-replay.json": {
+            "bytes": 10959,
+            "sha256": "43613225e9ccdbd34c0f7e3e0767887b9bbfd0f38b2c21684677d99e31a432f0"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/speculative-verification-budget.json": {
+            "bytes": 10177,
+            "sha256": "a729e1fc00be44e1480a5dc65bbcdd58df8599681df74dbc651a5a49f27e9028"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/zero-overhead-deferred-scheduling.json": {
+            "bytes": 8191,
+            "sha256": "ed5426af145f066f6183bd316c4073c46f324f5ee83aa9bd616387c0d0e7149c"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/lmdeploy/cuda/llm/kv-quantization-capacity-cost.json": {
+            "bytes": 3366,
+            "sha256": "873997b38981ccc03ff73e6fb84e96022f293a019843efd8c4aa50f27878ac9a"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/codec-constants-and-state.json": {
+            "bytes": 4250,
+            "sha256": "1cecd7905fef5af3ef8952a4d803900a7df91d662ebae34db47cc0dbf352d4e7"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/fixed-customvoice-prompt.json": {
+            "bytes": 3660,
+            "sha256": "ea5e8fdf54cb731b255b73aa0cf47ef53fac103baa2bd5246be45f95d54cb593"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/graph-batching-state-ownership.json": {
+            "bytes": 3950,
+            "sha256": "130741a9f7ac9d366784d6c4dd42860262235caff0c277de1e8a8f8ec933b37f"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/sglang/cuda/llm/hierarchical-prefix-cache.json": {
+            "bytes": 3641,
+            "sha256": "b50ca58df28d4343545d07adf0da23f8f181297b30aba7870e5690630e7d7315"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/tensorrt_llm/cuda/llm/prefill-decode-disaggregation.json": {
+            "bytes": 2967,
+            "sha256": "349856dd98628f94987e014889cd1b15d0e3ae6ab6c576cdde917f28fbafb102"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/xdit/cuda/diffusion/parallelism-freshness-and-topology.json": {
+            "bytes": 3236,
+            "sha256": "3674b5916cb7b384951bf3e449b6796b161878daabcc1d7e7f6d6f8ff861cc17"
+          },
+          "src/inferrsi/data/repo_packs/afd_plugin/pack.json": {
+            "bytes": 502,
+            "sha256": "00042abeabdc9c0e24f1a4edaba47e233df398aef0ad4e181fc70191d383da5f"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-aclgraph-capture-replay.json": {
+            "bytes": 4311,
+            "sha256": "327414443ea9537ae320aa405912c338351bc06cefdc0b99352e1eea36f61dc3"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-device-health.json": {
+            "bytes": 4115,
+            "sha256": "633fa5fdcea1e1e9edda8a218d6a3fde0d12637848348e1e2fe7d1b0bd744a16"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-distributed-parallel.json": {
+            "bytes": 5841,
+            "sha256": "709ca0d061e4d6917fe8fa049fe93165e71dfecf96c7bf264ced25346381f603"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-graph-mode.json": {
+            "bytes": 4135,
+            "sha256": "bb82ab3d628b91be957e557545810ef5c20790be7d43296096f20a21f45ec391"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-model-adaptation.json": {
+            "bytes": 4137,
+            "sha256": "7ea1ab0f0044afff935a91071b36ac09bc46b5ef044fe2a68ab0608256ddd853"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-profiling.json": {
+            "bytes": 4186,
+            "sha256": "0d27058fd8d0ef8b524b7a60d1cb76950622746c94b5a03915e64e6edb4ab3d4"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-tuning.json": {
+            "bytes": 4439,
+            "sha256": "fcb49d71f59d8b39c1eaa03d24e95e2cbfe5f9d7f6e0f22df03ee0d60f6a568a"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-optimization-casebook.json": {
+            "bytes": 4371,
+            "sha256": "05783630bfaf3c7ef99b848865f4bbab50d0b52e80ad04fcd028c624d7638e9b"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-performance-analysis.json": {
+            "bytes": 7200,
+            "sha256": "b6eea897c12d0c6ac9d172089535c82c25eb71c09d7a4972e64812cc4682b5c9"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascendc-tiling-design.json": {
+            "bytes": 4505,
+            "sha256": "e69b6e89dd342ba55ae54b5fac14aa1b91ee2a361da40cb669c287ce519f4517"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/cann-attribution-overlap.json": {
+            "bytes": 4915,
+            "sha256": "8ac01b60f5cd37fd594d6cb5119e4e6525199d017296fc8779dc68a576f5525c"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/torchnpu-device-memory-lifecycle.json": {
+            "bytes": 4349,
+            "sha256": "7a79d0eb2c50381d40086e6db72d4bdd2cec5d2261bfb00cf73a60701a94da2a"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/torchnpu-pinned-host-memory-lifecycle.json": {
+            "bytes": 4431,
+            "sha256": "8bb34dcfbebb48667a94434b036a1a408b970af9890b047a3de325f364ed5f7c"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/assess-serving-preemption.json": {
+            "bytes": 4198,
+            "sha256": "e3d2a1a72d0d980f31fc85deee27db26bc0daa0c709e8ae99e42deb814a4a7f4"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/assess-split-backward-schedule.json": {
+            "bytes": 3915,
+            "sha256": "a8561794786eb0a897d831f00c92b4fdd3b0492316c1b1d8ad13c1f8f75ac815"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-h3-runtime-residency.json": {
+            "bytes": 12169,
+            "sha256": "602dd625874bd4a89861d5bb25881cc63e4b5a7fdedc2867089f0952943f6eff"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-llama-runtime-state.json": {
+            "bytes": 5816,
+            "sha256": "64f6ab8a95172378350323cead08bc6c201c624b711e6242938a6a21a8e83ad7"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-sd-cpp-residency.json": {
+            "bytes": 8349,
+            "sha256": "95ed97b717255be3b817e5c69debd6d41197559dbf21771507d5da79134a72e1"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-sglang-diffusion-lifetimes.json": {
+            "bytes": 7276,
+            "sha256": "5e5889223878c6cf42a49c6f09c60e9d9add88cb7e7d0df050c4e75d88a05aa6"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-vllm-kv-lifetimes.json": {
+            "bytes": 4275,
+            "sha256": "00424005a5eeb8b15f239edd2cb2c71047787e3ecf55924860cb7a78d137b586"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/bound-windowed-attention-replay.json": {
+            "bytes": 7877,
+            "sha256": "98b2291147a796ef59cf20a075ff662998ee6a529f85b49e9c6fb62fe41fad9f"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/budget-speculative-verification.json": {
+            "bytes": 7490,
+            "sha256": "66448f59460c9dffac46e74160388fab650cf8bbbd5bdc95c1465b70daf0c2dc"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/calibrate-serving-simulation.json": {
+            "bytes": 3992,
+            "sha256": "055b70e0b86f65156775b519e4b8cbf06a956a8993b06a66175167cedb78f2cb"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/cgroup-resource-controls.json": {
+            "bytes": 4801,
+            "sha256": "ad406b6dd5d9415bd28d448e64269a388b50e11f9eb7f5a9adb6b0a6246fd303"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/checkpoint-recovery-consistency.json": {
+            "bytes": 4740,
+            "sha256": "ccb85314a831bde58f8cc515580d001b174c11dca0bbcbc6221303acf629046d"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/choose-parallel-training-plan.json": {
+            "bytes": 3901,
+            "sha256": "40afcb41c59988fc1af8fdd00d2340800c54e96847dd0de16ecddf7ac221d1b4"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/cite-check-downstream-pr.json": {
+            "bytes": 2533,
+            "sha256": "dabff14684d5ae9a2b05a9743ff5b76d00a90fdb29319588aba67f5087aad706"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/claim-contract.json": {
+            "bytes": 2469,
+            "sha256": "78f15f6533434fc29adf1df651f7a15eb8ba81733f8b02c710d90102541d707b"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/comfy-ui.json": {
+            "bytes": 6919,
+            "sha256": "1496807d958df055f720ca93e6dcf7f7e8f506f3f12c94c9a2166f789b00058d"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/competitor-scan.json": {
+            "bytes": 2449,
+            "sha256": "c2e97e49e5185710cdfaa08ac1c2b1c19b958e68ebfca885d5aee0ed8c9f8c8b"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dcp-async-save-performance.json": {
+            "bytes": 4538,
+            "sha256": "359a5acf8304a78fc074688676f9a0be62f56f463d0960a544e409f49843bdaf"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dense-performance-evidence.json": {
+            "bytes": 4227,
+            "sha256": "d87f93a9dc9ddc583e3929baaac85f7a425666dd6dd71665933f18344ccec834"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/diagnose-vllm-admission-pressure.json": {
+            "bytes": 9789,
+            "sha256": "a1de40def775b2e840c27330b67e6f37c3c244db1731fab2bd3e8a787143210e"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/diffusers-invocation-state.json": {
+            "bytes": 6109,
+            "sha256": "914d2ffe9776ba931bed780a2a270c1525e3c6cbaf0735b8531064d3180f3233"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/draft-acceptance-issue.json": {
+            "bytes": 2549,
+            "sha256": "6aaa63dbfe6cda982ccd3a4e368527b80fd2873672c000c66d06d28e40a2bf14"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/evaluate-elastic-serving-placement.json": {
+            "bytes": 4334,
+            "sha256": "fab1fa814e789ec2d86587fcf7de7c38174464b9e9ba0cf53ef3fb470053e7c4"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/generation-cache-lifecycle.json": {
+            "bytes": 5972,
+            "sha256": "e5679cdecb722bbd17280913913da8473052ad09e9e7a13b3a9b95463a139ef6"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/host-memory-management.json": {
+            "bytes": 6530,
+            "sha256": "a36aea21169281866b2468860d3ffeb912e61edd571308b02f82939f17f694a8"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/host-overhead-optimization.json": {
+            "bytes": 4243,
+            "sha256": "05fd7f054966b96892980d585352bc1c3fcd47f439cb34268fdf07feee58a998"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/inference-benchmark-design.json": {
+            "bytes": 10735,
+            "sha256": "091bb086c6f0123dbb5db78968c13f846cad135e8969982efe557644b7c7613b"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/inference-optimization-casebook.json": {
+            "bytes": 6531,
+            "sha256": "c512025d7a59934c91665e24752d9393ca9b6a9dfb624c72789294defae93c0e"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/kernel-writing-guide.json": {
+            "bytes": 4545,
+            "sha256": "f1f0ca1d11d3550c9b8c5fdbed8f6cd09b877ac14341f973d58dca02b5d8995d"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/kv-tier-management.json": {
+            "bytes": 4609,
+            "sha256": "6141f61fb6b0d147b704ab7a2627e9babdb53a90e15b659152a3d565967afddf"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/lazy-evaluation-stream-lifetimes.json": {
+            "bytes": 5934,
+            "sha256": "34f47009a8a33f8907bd78f110b95fcd62409bef1b38e1055acb46b2eab285ff"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/npu-skill-factory.json": {
+            "bytes": 11868,
+            "sha256": "24c1d3bff288e86b99ea674e3252ec6f54d9ba944274aef6a7666197afd6788a"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/optimization-search-policy.json": {
+            "bytes": 7101,
+            "sha256": "d7bf71d5491765f62eec02fe5013dc1a5ef56e1516892da48593c6165a86366d"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/os-process-placement.json": {
+            "bytes": 6304,
+            "sha256": "e6e8fe5d7d66f3e98f05762d94441726dfa757ce7fbffd2b35402b3f327f8fb3"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/performance-ceiling-analysis.json": {
+            "bytes": 13695,
+            "sha256": "47c034d96026e1c2a02848d9c9c3caf4024a5033664bfd089ad30d9f23a80deb"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/plan-activation-residency.json": {
+            "bytes": 3871,
+            "sha256": "436e364020cc854fd52a10ee76e4593c7519bb0546c8aa5238fcfd89158e8ab0"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/repository-evolution-analysis.json": {
+            "bytes": 5865,
+            "sha256": "e0af402c3320ebd9bd36c774cb52c0708721a86ca446c05f6e94306888838e30"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/rsi-evidence.json": {
+            "bytes": 2475,
+            "sha256": "c8b4bb5a8c88009ac5b1526e82639715381c940d3c3eb295f0333cae6c2d3ca9"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/skill-evolution-playbook.json": {
+            "bytes": 4091,
+            "sha256": "0e063c2e7686de0c27c061e1aac0eb2d8c4fa83246e32ddeeb0e28126c176425"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/storage-io-measurement.json": {
+            "bytes": 4783,
+            "sha256": "80e589e3a0b67d3a045e03feeadc45e720ceb277e6124c917eec04863eb67419"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/training-input-pipeline.json": {
+            "bytes": 4296,
+            "sha256": "ae696bea8d2b7bfdaa29b6e5bc0bd861abcc0e42e7d9bec6ae20c6e7d6f27669"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/training-resource-budget.json": {
+            "bytes": 6944,
+            "sha256": "ea5e5e864e3aa34bd00be67c8e741877ae01480f039652073b65b71dfcd54c97"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/weight-loading-optimization.json": {
+            "bytes": 8139,
+            "sha256": "7f2d1fbc6b1cfd80c3d6ed8df107d4ca9b97655626ee078c345445a303b1b367"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/audit-afd-p2p-exchanges.json": {
+            "bytes": 6640,
+            "sha256": "9ac45aa784b430ad1ca9ffddecce8dbae9740cfe9c98c10b330e01d350015a38"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/diagnose-attention-bottlenecks.json": {
+            "bytes": 7602,
+            "sha256": "e525d75578de1cc164e6553d01efbd0a014e74646200580f59b1161c030413ad"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/host-overhead-profiling.json": {
+            "bytes": 3965,
+            "sha256": "e9e6ae8c5b9986e83cade3bd029baae112859e01aa6c96f741a264f3bee44adc"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/reason-hopper-kernels.json": {
+            "bytes": 15086,
+            "sha256": "758a649e651d8f9635ac2df45d18043e156c8d54b8f8dbc95f8e68ebcb866a28"
+          },
+          "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/parallelism-and-extension.json": {
+            "bytes": 3399,
+            "sha256": "1253f5f01bcf1498703c7f34960f0c50580df5ba8f140af557a903e296ad81d3"
+          },
+          "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/runtime-radix-hicache.json": {
+            "bytes": 3317,
+            "sha256": "5a0b2aed51649af214cdc1034f57650b7ce195c934a17452ee67a35131f80cdd"
+          },
+          "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/tuning-and-validation.json": {
+            "bytes": 3385,
+            "sha256": "ef4d4f7182da4433d59d72dda0d21b4aeafadd73260be11ab8d4cb36c20fafc6"
+          },
+          "src/inferrsi/data/repo_packs/sglang/pack.json": {
+            "bytes": 620,
+            "sha256": "bce2e0c42796a9ef499d459c37b56d1d56f4b19344841359c534fca29c0359ce"
+          },
+          "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/audio/serving-and-streaming-quality.json": {
+            "bytes": 3440,
+            "sha256": "97fc1d5a11a39a0634dae9e9602a4ec47a4d2348e121076c2d5c836cf0692114"
+          },
+          "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/omni/extension-lifecycle-validation.json": {
+            "bytes": 3482,
+            "sha256": "7da8dc5d2655099de2bcd5f0fa922d9e1bcb0bfd2f571f026a004a7f402a75b6"
+          },
+          "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/omni/pipeline-state-and-transport.json": {
+            "bytes": 3404,
+            "sha256": "0c57692772019504264e39bf0444a69e20cff86f6e112b072c53cf7a4299cc82"
+          },
+          "src/inferrsi/data/repo_packs/sglang_omni/pack.json": {
+            "bytes": 630,
+            "sha256": "8e70336bc2aac0c74732a026337813652b5d5c6e1fc60f5571e9442345077b96"
+          },
+          "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/architecture-search-space.json": {
+            "bytes": 3294,
+            "sha256": "2c5ad72d5bab4f7082fc3be1532c347444cde38f5a84639fe16b22134b68685e"
+          },
+          "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/controlled-execution-tuning.json": {
+            "bytes": 3255,
+            "sha256": "6a2bf611a432213919e1c237687507ceb0ab753cccad0e934dc77f86604805e9"
+          },
+          "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/engine-validation-contract.json": {
+            "bytes": 3295,
+            "sha256": "ef89775bc82965fe87a832c7ec8a5cf55cb9eba4ecfc6279caea9ce3b0410c41"
+          },
+          "src/inferrsi/data/repo_packs/strata/pack.json": {
+            "bytes": 617,
+            "sha256": "643b32defb8db9aa230f624856f4fb4b595b5c23db58d9453091f5420cff71c3"
+          },
+          "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/bounded-serving-measurement.json": {
+            "bytes": 3239,
+            "sha256": "fb00ddac14ef6b06f48da9ba335529bd55d5d83c2a2fe331f0aa1ca16afceb37"
+          },
+          "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/extension-correctness-gates.json": {
+            "bytes": 3343,
+            "sha256": "4ea7083d277df205ea4c361bee87a8207951596e9ef6cfa136cc7d2aeab02f46"
+          },
+          "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/runtime-and-kv-ownership.json": {
+            "bytes": 3280,
+            "sha256": "e6142c1065754716a5f26f42119722908dc8fff63e667c986d57f1227b8610bd"
+          },
+          "src/inferrsi/data/repo_packs/vllm/pack.json": {
+            "bytes": 508,
+            "sha256": "1705d96710249c4514c15d8d011b43c1a1ae7c15fbc9ddf11ef2d0e6b2ab7d07"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/minimax-h3-reference-kv-reuse.json": {
+            "bytes": 4047,
+            "sha256": "80b853fc594d387ab48bed04dca245e0a22b8cdaf7c9b853551ab3bafa302e84"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/minimax-h3-sp-head-buckets.json": {
+            "bytes": 3742,
+            "sha256": "c4d29d1bceb5e2cf86151bf53c9dafb3d83428fb1e9d2a848d852bb8824bf857"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/qwen-image-21-eager-t2i.json": {
+            "bytes": 5268,
+            "sha256": "ef107d43cc448219eda4d64ab961b61cda6350e20d30d94de1ff7c4ffff7e775"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/audio/streaming-serving-measurement.json": {
+            "bytes": 3389,
+            "sha256": "2680f32f28907f6abc7a7e1e3ecab4f46ee412f40f4898ed2f371ffadf47ae4d"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/measurement-and-profiling.json": {
+            "bytes": 4472,
+            "sha256": "c8b25583cd160024c7472a20debea34cb6d4bc79714b4bad2953555adc9f794e"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/production-integration-gates.json": {
+            "bytes": 6112,
+            "sha256": "649b49ff64d3e9e88bd620027dabf9d126ca69bbe91b448f28e845ecbe1f8e2f"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/step-state-and-output-validation.json": {
+            "bytes": 5090,
+            "sha256": "e1ffe28c4fa54af739cdef2fedf4755bba18e078f8380d4fde8138ea712e5778"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/upstream-review-gates.json": {
+            "bytes": 7614,
+            "sha256": "5127709bb85ce31008d19468e1b4fcb5454a71776f69f991841cd9ef7558f9c4"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/omni/stage-ownership-and-search.json": {
+            "bytes": 3427,
+            "sha256": "e75d6c2829ae22881893f5da2ad7edba099fd64ccfb0fc93d30473e06e9d4d80"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/audio/concurrency-scaling-state-and-overlap.json": {
+            "bytes": 25324,
+            "sha256": "cb47fabb232db81ed525cbb54371becc7e4a28b2bae9d8492f60dfb6f31563e8"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-adaln-schedule-cache.json": {
+            "bytes": 4253,
+            "sha256": "ddbba3bb72b5789938595a82d3280e861f75f6ed54f05a1d2b7dcb1089d30f63"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-cuda-mxfp8.json": {
+            "bytes": 4468,
+            "sha256": "e4c538f03f4ab90a4acc8e3d8721bbd226aec1ee9bde45757a7c27d3d970821d"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-fp32-modulation.json": {
+            "bytes": 3625,
+            "sha256": "abb8d94a1b025f0e2c8e085275e96ee85d68aeb27cb12d0918e4647f8255b693"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-qk-norm-rope.json": {
+            "bytes": 3706,
+            "sha256": "4c1c4acfb9cc0baa90bc28fed1903020ea1ea38755b84315711795db45e48bb6"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-ref2va-teacache.json": {
+            "bytes": 3835,
+            "sha256": "8aeb3f99ac4227cf8b3a5cece13d938b19082e265a557eddae27b7e495bcc22e"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-vae-stacked-tiling.json": {
+            "bytes": 3508,
+            "sha256": "8cc18eacbbfd8fc4d3a6b4152c4b584d559fc9b15377da56ef072e542d807c93"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/pack.json": {
+            "bytes": 566,
+            "sha256": "58c971e9dcf6ac15752629ed4d9426aeb5b5d857fc37b5bcf921aa2e81c0cdd5"
+          },
+          "src/inferrsi/data/repo_packs/vllm_rlt/pack.json": {
+            "bytes": 649,
+            "sha256": "6097041b8d026740060e527f49b1b4e288315b1c3227f7583805b86d8fa415f7"
+          },
+          "src/inferrsi/data/skill_library/README.md": {
+            "bytes": 18819,
+            "sha256": "ae583f3578425df1dc91ba32b8b6543c0469c9a817ebea873a77051d6b13e1ce"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/SKILL.md": {
+            "bytes": 5494,
+            "sha256": "7581b56336d07213a87d01477793ae1a8acf0b7cf876fce36c9a761ee32c1ff4"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/metadata.json": {
+            "bytes": 1429,
+            "sha256": "68ac7f64b7ab947fc2db664bc3990f102c1cc5539b42350aa5818d0bbf17d86c"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/ownership.md": {
+            "bytes": 2081,
+            "sha256": "40688cc44183027efbf2fca8d3548c299cdf2e33fef54f76ab3549353eb9bd1b"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/source-map.md": {
+            "bytes": 4635,
+            "sha256": "10562a75683bd9bd07ec2e37a1cbd0f93f2534d6f85b55a566adb2d97485215f"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/validation.md": {
+            "bytes": 1477,
+            "sha256": "5a0b854b87d091d7d505b0bdd5c03330052507a851ab6edf35b93a6675f2d4f7"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/SKILL.md": {
+            "bytes": 4010,
+            "sha256": "a95569bcc0da6e715fd74783a6a1e0489e92b680c582f15dafbf98df589d505a"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/metadata.json": {
+            "bytes": 1519,
+            "sha256": "9c2293c8956508de987c886c7224d673d22912e7e915b026b05abdd287e97401"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/execution-lifetime.md": {
+            "bytes": 2147,
+            "sha256": "c4bc41e08aca68f661325f61e6f0db1760a079dc362f295c847bdc374fee3bf6"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/source-map.md": {
+            "bytes": 5918,
+            "sha256": "798b9cb89ed6b885eb47464ca9f137d2802a54571b9327e631d7742d3e5abd5d"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/validation.md": {
+            "bytes": 1608,
+            "sha256": "ebfc00bf7d6ab0a1dc415953ae84d56ccdc713d5614aff85e433c4a843793bff"
+          },
+          "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md": {
+            "bytes": 7322,
+            "sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845"
+          },
+          "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json": {
+            "bytes": 587,
+            "sha256": "db7ad368394ccbfdeff6a16e3423c300d51305b1f95ff5c3936df91738a8cc35"
+          },
+          "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md": {
+            "bytes": 5314,
+            "sha256": "22afe4f4d77f441096cc657d55d763473bfcbea8fb531c8e493fe49ab84955f7"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/SKILL.md": {
+            "bytes": 6745,
+            "sha256": "d494fc7e539cbb7388189bd7b3f04bb5c4983fdd7a82484348aad1cee7f741f2"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/metadata.json": {
+            "bytes": 558,
+            "sha256": "71c68431663ed2336cccc985a354adab6b13b4cfd6123a18a5d8d53c28268509"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/references/source-notes.md": {
+            "bytes": 13004,
+            "sha256": "6ffbfb68cc7e04923b4fbca2a73a7555b7c3b45a6f40819bb5f3213ef05cf88d"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/SKILL.md": {
+            "bytes": 5977,
+            "sha256": "03829039860ece35519f2e4c339a48b5014ec25e07a2c150a3efb6a3a4e35b4a"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/metadata.json": {
+            "bytes": 559,
+            "sha256": "80d80c15514919826fa72d6f6cf26b76db8accd85d21187fc239232ca454e7bf"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/SKILL.md": {
+            "bytes": 14703,
+            "sha256": "4ce9dcadd1ae3c7a8cc633336760d0a3c382f354f1b7028e3a10d28fa0d0c9fa"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/metadata.json": {
+            "bytes": 1403,
+            "sha256": "b6cb25879b1bca5f5af08b48776f2e664cf5dcd11a0efb1c4f56879e2afa9006"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/references/awesome-ascend-communication-tests.md": {
+            "bytes": 9388,
+            "sha256": "0cb27a3d24409171bb414d7ebf0c49ad62be5f46482f29624695496ba9eb3e63"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/SKILL.md": {
+            "bytes": 5839,
+            "sha256": "9203cc5a14a1d65df5fbf275734b93be023e06527459b2f79864f28a338cb44a"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/metadata.json": {
+            "bytes": 536,
+            "sha256": "72b41e7845915d34b09219624e82976386f3125816293540c2c38143fe6b477f"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/SKILL.md": {
+            "bytes": 6440,
+            "sha256": "8fdf46fb2743b67a66bc069d22a45305a222debfb606777bc451b4b8f31d9cc6"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/metadata.json": {
+            "bytes": 536,
+            "sha256": "be681a3206e5e183b4c9670c9d7784bcffce18c891449e5b15f94ca5f1324b9a"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/SKILL.md": {
+            "bytes": 12988,
+            "sha256": "401981cf7828edcab0c9ea882f7ecd83badfa3c2a95006e912cf99d67b3d6a12"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/metadata.json": {
+            "bytes": 760,
+            "sha256": "efd37bb23acb500dad4ec8ebe7ddfe5c3f196374052825f923ac85da8445f06b"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/SKILL.md": {
+            "bytes": 12236,
+            "sha256": "38b151bc6b4e9734c6fa41efb7a30343559400c0c75bbdd26bd8a968b8ea4592"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/metadata.json": {
+            "bytes": 557,
+            "sha256": "04d2002fc8a0a7becf86b0b73daaf2da8716ab466d7f988da43d41a10f26680e"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/SKILL.md": {
+            "bytes": 9172,
+            "sha256": "647f5e9a288f2c4e1617ba53df59d9d827df749eb4fe86182d21a9c41175caf8"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/metadata.json": {
+            "bytes": 536,
+            "sha256": "f425f7c1abbd616387273e8932c6bd02f4b1da012f099286ad4171e0f6037b22"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/SKILL.md": {
+            "bytes": 9880,
+            "sha256": "a057b7f86a24f34a8f84abbbb4910d58d4e50ff47060e1c404600017e3febe89"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/metadata.json": {
+            "bytes": 2297,
+            "sha256": "9e705625c8308013c287576b0247da6f6116d817a4a535cbecff80abd6ea35ee"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-CANN-2.0-LICENSE.md": {
+            "bytes": 5262,
+            "sha256": "89b1c114df62cfd183004f38d18d07f186436c3b37203dd69c3e84becf9d2282"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-stage-evidence.md": {
+            "bytes": 11789,
+            "sha256": "f56e4125325c9c1b9235a1d64fd732861c7686444492422daff44ba8dd648be4"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md": {
+            "bytes": 11860,
+            "sha256": "93aa1b775ced304cebfe6c4dec0503ddb2260ecb3cb01edc17ce329446c5ce2b"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/metadata.json": {
+            "bytes": 552,
+            "sha256": "f6d94f104cb70969024b2f4542078a240e200712c66279eca8b94f60bee1dd71"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md": {
+            "bytes": 8972,
+            "sha256": "0e6f094dc554abea12e07def50f20b233b47127779a8aefad595b4ab1efd24c2"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/metadata.json": {
+            "bytes": 569,
+            "sha256": "acccc8eb86de0df6caf78a9b3c6db76a2a686adc4a5ba0029a4d62d50ff9a8f8"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md": {
+            "bytes": 5989,
+            "sha256": "8428df7157af024fa17cfe0b3818ed54bd9f468306dda46eb8d4f0edf4f3f230"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/metadata.json": {
+            "bytes": 579,
+            "sha256": "c650dce88c07b636d68e063dbda4d319f6f7f3ecbcdd5e47f393cbfc3af7c71d"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/references/source-notes.md": {
+            "bytes": 11470,
+            "sha256": "80cba4d3a884e7ee2618326cf74b9f9c3a34e4d9cf443707f6bbe673f2d5ba65"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/SKILL.md": {
+            "bytes": 6159,
+            "sha256": "23e93ea7c98f99743509fdc04139d78ed2e7123c63e80b366c6beaba7eb70d0d"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/metadata.json": {
+            "bytes": 581,
+            "sha256": "97836a38465106d816a9ace894a7b277d6d330f9accade8f8493ce6f0fb57ef9"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/references/source-notes.md": {
+            "bytes": 13493,
+            "sha256": "435193590b4a2ba32e2e4c6fa9d7e8f889528d412cbbb48e35ec5433f7dab384"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/SKILL.md": {
+            "bytes": 7137,
+            "sha256": "a25738dbb3b9daa15c5827cb536a20fe13f7c8b9dd14aa3fb7cfa7e132879431"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/metadata.json": {
+            "bytes": 1007,
+            "sha256": "3abe626a6e4c73c943429f609ce2c62aa77aae7e1726475ebbbdfbe1794f6f7d"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/mechanisms.md": {
+            "bytes": 17376,
+            "sha256": "fe08b2292fa587be3874774ddf41063b84c579a0085df422ec6710d8aa90b8b7"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 4024,
+            "sha256": "371b01ee201ef51fecbfd6da892d517aaba295930b9395025f03f34e21c030a1"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/NEXT_GATE_LEADS.md": {
+            "bytes": 1623,
+            "sha256": "35b8b58ae5db02b2df51ccbe5cdf870a7383d17f345f9ee5eb7b0a4146dfc74b"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/README.md": {
+            "bytes": 2888,
+            "sha256": "e0b1ec1b39c10067f5ddce58320851ad089568c4b2296a24777389097cb5b388"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/io-reduction-can-plateau.md": {
+            "bytes": 2671,
+            "sha256": "5cc35f514224ac6e6efd2da6db221bba468b2f4db66807295e67b240bcdfdba5"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/overlap-and-fp8-counter-signals.md": {
+            "bytes": 3692,
+            "sha256": "80c8a60a6df55f3d84fab23e03567d8b38e23bcdbd63cc09881e46db1e2888ab"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/partitioning-after-fusion.md": {
+            "bytes": 3132,
+            "sha256": "4d4cabe0991eab554dcfbd4fa9c8199e61fa4767709b0c4cd646e2588bb398a2"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/knowledge/attention-cost-model-boundaries.md": {
+            "bytes": 5223,
+            "sha256": "5ed8f888443e38ccb72545b6025447aab7b92936f78c151fed680e582126239b"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/sources.md": {
+            "bytes": 9069,
+            "sha256": "53436e0de4e6dda2d29426591bf61f8526378cdc6f9ec7967b88b1bcd4c61e6f"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/host-overhead-profiling/SKILL.md": {
+            "bytes": 11199,
+            "sha256": "326e566fbe30d3ba04e4b50c3e87aad9df037f0ac4ad6fe129165ae97dc5f91d"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/host-overhead-profiling/metadata.json": {
+            "bytes": 527,
+            "sha256": "76e34c1c64aba473a9e126b032a776417caa0f913f6ac3cf4dba3a5d123b1ee0"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/SKILL.md": {
+            "bytes": 4905,
+            "sha256": "cec4e3aa1eb64c31c54b5f5ca40af2b61b7919d94be4174968581d5f752c0538"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/metadata.json": {
+            "bytes": 7453,
+            "sha256": "7520da1cc97412631a600957ecb18aa5a7a4461ea21e76c237330180193ed898"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/fp8.md": {
+            "bytes": 6287,
+            "sha256": "31d9c211c11c2224007abbdcd4059913c4a54833fe2d2f926411c5238fcd3501"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/cutlass-bsd.md": {
+            "bytes": 1547,
+            "sha256": "42fec630f410aa308f70a51a89fadcd19586fa620f9831a32bee528a9a10000e"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/deepgemm-mit.md": {
+            "bytes": 1065,
+            "sha256": "6c73a30c14a2938abaa6afe98ebb19e6d81dac5bd4ace0cb05c61c20afa37b85"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/kernelwiki-kda-mit.md": {
+            "bytes": 1063,
+            "sha256": "84aab57158a914f448b38d8be29dd842b634a08c192d844f416a256beed4ae94"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/notices.md": {
+            "bytes": 3240,
+            "sha256": "876ec016644bf5331b3f257985e126c0e61980e3bfd82569b7b0c3f6b2c6c64c"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/resources-and-scheduling.md": {
+            "bytes": 8919,
+            "sha256": "1ce25b426736072435a1ead9a2a8e1f56df5d0abb8198e8c4fdcb34321db4964"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/source-inventory.json": {
+            "bytes": 120506,
+            "sha256": "1dc09c895c970a0c9708158bcc6dd65cdb736274fd2c35d44542c70407d3849d"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/sources.md": {
+            "bytes": 9004,
+            "sha256": "6b74b9e774f2a80db0860510d9e6940d36284430f29c1e124f115a9eaf25833d"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/synchronization.md": {
+            "bytes": 6843,
+            "sha256": "4322f6b3c0e4368bf8c192b4ec752529767f50d19c0c38dad8fdf9cd781e41d9"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/SKILL.md": {
+            "bytes": 7348,
+            "sha256": "fd1f5a7e6a55714730044155f5b38c1d700383d4e16c3e91963bbfbe1c773af5"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/metadata.json": {
+            "bytes": 572,
+            "sha256": "19e40763fedc20990651bb989c32a7693306c5c54af46c3d4be77fd4f87b6c03"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/SKILL.md": {
+            "bytes": 23213,
+            "sha256": "2c807f3df2da2f533d25183de5771f3d412058d16410994292ef8f48c49f9b3a"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/metadata.json": {
+            "bytes": 1684,
+            "sha256": "9ac6af135f0b6d76a76bc201b6bbd8ac9afe5201dd1e2ea82c0ac4826de401da"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-LICENSE.md": {
+            "bytes": 11357,
+            "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-native-offload-readset.json": {
+            "bytes": 6267,
+            "sha256": "cc6c2011be746002ba3ff96ccefab1809863748ddd17eb102b75e7b0e134fe55"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-native-offload.md": {
+            "bytes": 8966,
+            "sha256": "6c2a72b5ba3708a01d229e0576570837a617e6a42dbb0b5312a26a65a6c31542"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/SKILL.md": {
+            "bytes": 16072,
+            "sha256": "36ef88826a32ac3147b869001004ded17a8d5a491e25bfe105005ba39ea9d1eb"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/metadata.json": {
+            "bytes": 539,
+            "sha256": "ae71c389dfdb1a19d267feecb4c6f1fbb4c9afad501ff6d8fab7e38b29bcc37a"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/references/deepinfer-zero-overhead-scheduling.md": {
+            "bytes": 10894,
+            "sha256": "f452bf40132c9195d739f1c86b3aa18857138c7ed0f1861af5853d93b9955924"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/os-process-placement/SKILL.md": {
+            "bytes": 10327,
+            "sha256": "2003492e4d89fde0443b180d74933e764bd1979435e109c8d26e20eb8f6e5053"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/os-process-placement/metadata.json": {
+            "bytes": 1399,
+            "sha256": "ec02995c075f0a540921f57b4846f2cafe334eb90f44389f17c3c50457c3e9c5"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/os-process-placement/references/llama-numa-thread-readset.md": {
+            "bytes": 6617,
+            "sha256": "748b254b3b043a5ddb0f480eb4a566854d27eedd7b111499863e9bf0a5a6951e"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md": {
+            "bytes": 8813,
+            "sha256": "63b883c10a8d3e7d918abfc7175b72e2b401ff1e02697f94b8382ec1c8f369b7"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/metadata.json": {
+            "bytes": 558,
+            "sha256": "0a41c7b00978332b54ee1e4a7c1ade88ecde3e961128f24d1b4f7708caff5ff5"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md": {
+            "bytes": 6402,
+            "sha256": "0a5f48d9352c5bf460775ef47f9136829978811f4a16e84983e299782cae305d"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-serving-preemption/metadata.json": {
+            "bytes": 537,
+            "sha256": "77ce5ba16245cbd80d46d0abdf80aaf26fc0924fa2d2a5624ff430de68a1daf0"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-serving-preemption/references/source-notes.md": {
+            "bytes": 30405,
+            "sha256": "f29fcecfba0490008a354c11b42403983fe165e56ff907ad61cc6040d2b6407f"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/SKILL.md": {
+            "bytes": 3181,
+            "sha256": "1e64f28ea71672fb53c56f6108b87864738c6d03d6dcd32eedaef9e0625e96e3"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/metadata.json": {
+            "bytes": 478,
+            "sha256": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/references/source-notes.md": {
+            "bytes": 10548,
+            "sha256": "61b1609f71c37b61d136b8c4d39f793de87cffb1cf31c8ec737a1c3fa6fb1b4e"
+          },
+          "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/SKILL.md": {
+            "bytes": 3884,
+            "sha256": "ac773c6f1c110a664dfca28a3ae9129b10c07614e60ca3640dea776220e7b352"
+          },
+          "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/metadata.json": {
+            "bytes": 648,
+            "sha256": "261ce9f644ea6eebb90c830f41fb8b320d2b401456e5bd8d494095a44b8620d0"
+          },
+          "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/references/source-notes.md": {
+            "bytes": 8415,
+            "sha256": "9696d3c0ef96b6611787446bd6f96d4d6255ca8d107105de8c1d9af83edb9fa5"
+          },
+          "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/SKILL.md": {
+            "bytes": 6416,
+            "sha256": "6b7db1eb2780dec27858cb7aa0a3c093c5b5dfa4c30261aebfeaa255213ff688"
+          },
+          "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/metadata.json": {
+            "bytes": 582,
+            "sha256": "4dc8659806539b4784ea9348a01d4194ac70d0bb297a916f90ff6419efa5c328"
+          },
+          "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/references/dlrover-persistence-replicas.md": {
+            "bytes": 8398,
+            "sha256": "055f4357f79026fcbcc766bbe77db610376ec353623f8898f5d14a30ee69e7bc"
+          },
+          "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/references/sources.md": {
+            "bytes": 7984,
+            "sha256": "9425b1aab963a2c8d35f3db897bd15ee69699bf5b0d0c8425c3c8bba3e9a0233"
+          },
+          "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/SKILL.md": {
+            "bytes": 3068,
+            "sha256": "25d04aed9782b2a8ff1186828becdc7fce93c4418039ceaef9b8dcc5ec9bcd24"
+          },
+          "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/metadata.json": {
+            "bytes": 478,
+            "sha256": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7"
+          },
+          "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/references/source-notes.md": {
+            "bytes": 8521,
+            "sha256": "9f9184d46b19312138a6e805e439e43dc0594a0cc6b46675255a582dc946ca5b"
+          },
+          "src/inferrsi/data/skill_library/methods/dense-performance-evidence/SKILL.md": {
+            "bytes": 8305,
+            "sha256": "53cc1867a13f54d1006d50517dea8848869ca50df3f0cd610b2424cc8d18c933"
+          },
+          "src/inferrsi/data/skill_library/methods/dense-performance-evidence/metadata.json": {
+            "bytes": 707,
+            "sha256": "a0d1e3b17140295510a764045d1b222e19be3a3406e6d1e7dddaad49ff474490"
+          },
+          "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/SKILL.md": {
+            "bytes": 6667,
+            "sha256": "82d0099d3c33706f46c657d4fa7f76d3d486e019b36760d1de001921328a60cc"
+          },
+          "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/metadata.json": {
+            "bytes": 547,
+            "sha256": "37379a4d4a058ca46da2b20a920dc01be9ec11e557c32b2332702ec76027f4b5"
+          },
+          "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/references/source-notes.md": {
+            "bytes": 44796,
+            "sha256": "ea416831fd7d0874e06d0a76fb0c7196df53072b717673a0cdf635db6bfd917b"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-benchmark-design/SKILL.md": {
+            "bytes": 8974,
+            "sha256": "19ebc1f567aadf49f7a534a18fdb629c72c0e4ce043d360983136491cb72bfab"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-benchmark-design/metadata.json": {
+            "bytes": 5923,
+            "sha256": "18310402a8ade03fd48ac687f79ba9d447871e90337de84c0e1bc15c0b827306"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-benchmark-design/references/benchmark-commands.md": {
+            "bytes": 8466,
+            "sha256": "0898424e0bff3a6246e1eeeda6abd6c5bd974c047fe32f3bbd5ab9d31ce17cd3"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-benchmark-design/references/task-benchmark-selection.md": {
+            "bytes": 25086,
+            "sha256": "0d4bb080a754e26846e81680023db0dff5e6a4c96be4327c411c3b72ac4ef8e2"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/SKILL.md": {
+            "bytes": 6644,
+            "sha256": "89a6a82c692f7f5192f4d7f4c3a5ff08a4df3b1c4bdaa0aafb7e04292d8aa5b4"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/admission-drain-occupancy.json": {
+            "bytes": 1354,
+            "sha256": "3319e23bbe2bd40e2d45fb89dc87bdc92d4df2b5ddd8086fcf84813b43885b7c"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/admission-drain-occupancy.md": {
+            "bytes": 3134,
+            "sha256": "e88bbac5d253a8e7e3a3d5e21047f23f2970c11ac23887d4303ab94fca52dd27"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/drama3-rejected-directions.json": {
+            "bytes": 1459,
+            "sha256": "2cc0a526fe0509551607a03e7626dd3d7844e643c4afb194837f75cddead8b12"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/drama3-rejected-directions.md": {
+            "bytes": 3557,
+            "sha256": "6f143bc363a3290fc008c08e57cb075f1776d1c8cca8a6a59297d5b1fbbf61bf"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-ascend-rejected-directions.json": {
+            "bytes": 1293,
+            "sha256": "b3d538f79c56724ccb2895cbf50cf2e6b0fec3979be62a684a8230cd932dff55"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-ascend-rejected-directions.md": {
+            "bytes": 2564,
+            "sha256": "be396983812262515899a3e2ba4f01b23ff92e1b836881882c7580372b8ae552"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-b300-rejected-directions.json": {
+            "bytes": 1196,
+            "sha256": "146ea1564a46ed03092c7dc5796b318fc05b04b1821105ee00eb7fa7fa1333e2"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-b300-rejected-directions.md": {
+            "bytes": 2656,
+            "sha256": "cce2930f544a0f78f21797fba4d8ded298e4a81f76c34d7977b0e31d361bc738"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/metadata.json": {
+            "bytes": 535,
+            "sha256": "fb55067fb45970319787068a3206110641cceaf0ccbabd706be6d0b3d2afc2ed"
+          },
+          "src/inferrsi/data/skill_library/methods/kernel-writing-guide/SKILL.md": {
+            "bytes": 10029,
+            "sha256": "01fb1712d7eb988c5e412f02340d2e48b4f8564fcf4ab232dc5897869d0d50a0"
+          },
+          "src/inferrsi/data/skill_library/methods/kernel-writing-guide/metadata.json": {
+            "bytes": 653,
+            "sha256": "0189d184337662056fc6952c7beffc1bf6154c4fa173ae3e3e463bd22d70f36c"
+          },
+          "src/inferrsi/data/skill_library/methods/kernel-writing-guide/references/akg-affine-tail-guards.md": {
+            "bytes": 11941,
+            "sha256": "113deb2db10b171db3425bb3f1792bee763706b794c8cc2e222b4cc0d0114118"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/SKILL.md": {
+            "bytes": 12914,
+            "sha256": "882e7850b9b49ad692a2d11618701afe0f4e83727ea78df114be2029923c725a"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/metadata.json": {
+            "bytes": 1403,
+            "sha256": "9f1fd8a2af52bc964021e21f16ecbdb891e08f2057370e700d67f86605d333db"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 3597,
+            "sha256": "157e74ba180426b279b548a729769f152bf18f9a78ffca4ebfcaf26b7d0c5292"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/MECHANISMS.md": {
+            "bytes": 5875,
+            "sha256": "18bb2f5d7e52773c516420a142edfb531d0b9c822a2934d60f3234ac1cb9d40e"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/NEXT_GATE_LEADS.md": {
+            "bytes": 941,
+            "sha256": "16a264c7159f9c2c700ce8ba46e63f3e636b277c987da4970f6d2fa085fcb053"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/knowledge/adaptation-interfaces.md": {
+            "bytes": 3893,
+            "sha256": "5e761957cb5ecd93926e6ffc13deee7c830c6e8f0ee3c3846e9dd3b50dc74446"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/references/source-notes.md": {
+            "bytes": 13370,
+            "sha256": "54d83c3713045d59a48821bcf93598374ba4a630e3891ad39498f919317c4fb8"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/updates/optimization-search-policy.md": {
+            "bytes": 4238,
+            "sha256": "d3db44fa9a6831b86a02c638b323ed943e69fe256866213f10d68c72d762879e"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/SKILL.md": {
+            "bytes": 15324,
+            "sha256": "c5f78961c23e3c8d659f5675b522518ab06e1f2a8ea8bc74c5d428ffb41693d5"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/metadata.json": {
+            "bytes": 3669,
+            "sha256": "60b5b24ead590b0fb222398fba47822bf6723690da683e256a450bb5c0eeb806"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 7349,
+            "sha256": "083e462e0a204dce599362d46712fcfb392d053c5cda56216e22e28d4b8fa0fa"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/MECHANISMS.md": {
+            "bytes": 4003,
+            "sha256": "742c249082ceca230326f4407a40130fe624553b796ae5a157a8a4c15a7e1398"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/README.md": {
+            "bytes": 1204,
+            "sha256": "17b2d1c7c14124d59ac8a0d5ad4200ad2803d991e784a1a98cd97188cc364999"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/READSET.md": {
+            "bytes": 3937,
+            "sha256": "66459ce393858b2b2160f1e2973373bfc4b9902a12644631934b4ebe3140296a"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/STATIC_CHECKS.md": {
+            "bytes": 1111,
+            "sha256": "cb0aa7d148352ebb7b1678fc42a557a099575054716a341dfa25b6a998029064"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/cases/decision-cases.md": {
+            "bytes": 5491,
+            "sha256": "781a675a6533e0a18c044aaaee0d010da7013b6504654b8b8292a35176db801e"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/govern-optimization-experiments/procedure.md": {
+            "bytes": 3823,
+            "sha256": "ce036a4d8a0908da204515c47e2ff9623548d09de264fa65bd58fedce52a78a0"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/govern-optimization-experiments/references/source-notes.md": {
+            "bytes": 21659,
+            "sha256": "f9a8254e45473ddb2b73e576957b6d3db83c62b99dff1922f4d1e606e569559e"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/knowledge/bounds-predictions-and-decisions.md": {
+            "bytes": 5297,
+            "sha256": "5780d6eb5257612ec45deb94f85f334a6ecabf7e466d1678706041f6549bda79"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/redirect-cross-layer-optimization/procedure.md": {
+            "bytes": 3648,
+            "sha256": "6593a63fce9c474b515663a560c94b7df048dc5f6e77a95b8e54e4a64cd10a09"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/redirect-cross-layer-optimization/references/source-notes.md": {
+            "bytes": 26469,
+            "sha256": "6ff3a7bc31f1172db92aedaf7df1f1aac688bf769227b583a4cd82bfad52da67"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/agentic-kernel.md": {
+            "bytes": 41353,
+            "sha256": "7190573b85f94a4391bb0d887ccbd37a31b3f08bb0531114ad7afd72fe1567a7"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/flashvector.md": {
+            "bytes": 9830,
+            "sha256": "3926369acd88ff3c370bee5718e74dd15f3219c3e4f6311095e30e917bffa4cf"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/inferencebench.md": {
+            "bytes": 9294,
+            "sha256": "21f47111197a0b7f8d6384ea918d139dc8318a62ca3313ce060048271f50b639"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/prior-coverage.md": {
+            "bytes": 3254,
+            "sha256": "35d47ac078591221fcb666219bcd824f9400fc17aafecdc0499d8336f7178238"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/vidur.md": {
+            "bytes": 7736,
+            "sha256": "d38602af5219f5feeb49e6cd6bf53d66098c77fad141dae676f1d8a6743cf25d"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/roofline-contract.md": {
+            "bytes": 7091,
+            "sha256": "419ec4bcaec45ac579a133bce083d4f48779f26e664d7a86f720fa65f343b1d1"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/roofline-example.json": {
+            "bytes": 2015,
+            "sha256": "597cd3205df043cff3ae9d15924c77934e8d8eb300c00c4840f6bb2921f90faf"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/scripts/roofline.py": {
+            "bytes": 9004,
+            "sha256": "524de48a6764401695fe1444f1fa3d38f363069b0cd2fba4715a86ca4b28a5d4"
+          },
+          "src/inferrsi/data/skill_library/methods/plan-activation-residency/SKILL.md": {
+            "bytes": 3507,
+            "sha256": "542d8a1e6963f5019986e1fbd79d3a247a045a51487c09313026087ac457c15e"
+          },
+          "src/inferrsi/data/skill_library/methods/plan-activation-residency/metadata.json": {
+            "bytes": 478,
+            "sha256": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7"
+          },
+          "src/inferrsi/data/skill_library/methods/plan-activation-residency/references/source-notes.md": {
+            "bytes": 13907,
+            "sha256": "5ff52655cfd06fcf970ed2a7d4d4fe1c31115b102af2d41497de0c74848ed2b3"
+          },
+          "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/SKILL.md": {
+            "bytes": 13800,
+            "sha256": "7e20d3390f2501263e5e298e026c47cac16c234f2397833131e15ae5d34c0b97"
+          },
+          "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/metadata.json": {
+            "bytes": 1903,
+            "sha256": "d0e4d6950250e4cec43d82e23324c9301a925d4c22be30b4d2713d790cb6504b"
+          },
+          "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/references/capability-presence-audit.md": {
+            "bytes": 11461,
+            "sha256": "c21eb4a6aabe811a49a6540a34c54dbcfdcf7efae3052a79b54034c7838026a5"
+          },
+          "src/inferrsi/data/skill_library/methods/skill-evolution-playbook/SKILL.md": {
+            "bytes": 9448,
+            "sha256": "9cd37f21fe0d862069e4798edca8a04d59f8b840e55d521901594a63dddd9ba3"
+          },
+          "src/inferrsi/data/skill_library/methods/skill-evolution-playbook/metadata.json": {
+            "bytes": 527,
+            "sha256": "92b36338b9257e2e41b95e4370694d7de52ed84bc7971056b3bfced07b0993d8"
+          },
+          "src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md": {
+            "bytes": 6865,
+            "sha256": "64bd2c6b795d966045d1d3cca55976d3d49b08b06df7c7e06a70c2824cf55b7e"
+          },
+          "src/inferrsi/data/skill_library/methods/training-input-pipeline/metadata.json": {
+            "bytes": 587,
+            "sha256": "80659099fce6a6ec05e9084b75452c877a5812e30f60a78f31ef8c7efcdc1bac"
+          },
+          "src/inferrsi/data/skill_library/methods/training-input-pipeline/references/sources.md": {
+            "bytes": 2637,
+            "sha256": "60e0f3e88655d8e1d5eb4797f39ca4370deda01614f265b8d2ea4559d8ac020f"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/SKILL.md": {
+            "bytes": 7200,
+            "sha256": "bdfa5a4ef37d9f303c3a23cd6f98c21137b97ac5ed58e0ed5fac6d71d33e52ef"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/metadata.json": {
+            "bytes": 976,
+            "sha256": "d42b614af9cee6ee0d4a25dc5d57c7b98438feed1e807a7c9f1755109a86e9cf"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/mindspeed-historical.md": {
+            "bytes": 3384,
+            "sha256": "eb97f935ed30517a8fc9edec5e9cddc04f5d679d156c16d75bd29c534593360a"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 7368,
+            "sha256": "c55c6ad2131ef788866411dcc7056f221420ebff30c95c6bc919b09a4818cb73"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/cases/paper-grounded-cases.md": {
+            "bytes": 5649,
+            "sha256": "19bd25b1a944adb2351ebc6b1d795150ea033041a44f6cc261e881677812a101"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/knowledge/offload-delta.md": {
+            "bytes": 2026,
+            "sha256": "f8740fc5c8bbbe5ac13ff1cfec7ca499569d0b46e8cd56cf1f5a44c3a5e72242"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/notes/training-memory-scheduling.md": {
+            "bytes": 18660,
+            "sha256": "9aa6d3e01452dff95c8a8940ff62ead8c146ce215c9175b78c9f2ef508a7b2de"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/references/source-readset.md": {
+            "bytes": 22946,
+            "sha256": "2a5c166767f56fa47e7eba7386f1f36da8cb73702e5b5bbe14d6a456ca9ebfae"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/sources.md": {
+            "bytes": 2752,
+            "sha256": "6d248f58af8b350fc66c22898988d6f451ac60d03bcd805cd6e84df5e55174ac"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/stas-decision-notes.md": {
+            "bytes": 4086,
+            "sha256": "f1a993e6c4ca9a035a2effcaf5500081f227b0cc56cc885a6b22fda0041fc744"
+          },
+          "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md": {
+            "bytes": 25969,
+            "sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213"
+          },
+          "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json": {
+            "bytes": 2418,
+            "sha256": "798cbff6f10b7c24b188080ebf233b188b0e98ef2c6fdb5e00e1fe3eff727c0f"
+          },
+          "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md": {
+            "bytes": 8656,
+            "sha256": "5a39f14b102b0cee304541e011aa31e44e6b13527d1bee9987c3f3b66c064d3d"
+          },
+          "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md": {
+            "bytes": 6392,
+            "sha256": "6c640f28ee3f0253a17b81461c865cb9f947367002cb75bf72550c4572d02e97"
+          },
+          "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/SKILL.md": {
+            "bytes": 7181,
+            "sha256": "a46110891eca860915e314d6921a22856d631453b22b17837fe30262d6780535"
+          },
+          "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/metadata.json": {
+            "bytes": 2175,
+            "sha256": "4e6100dfb9fc7bc00b19915f4fd38ad509704f2a609c658adf4d632d4e05b6c1"
+          },
+          "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/references/afd-plugin-LICENSE.md": {
+            "bytes": 12747,
+            "sha256": "0e8566ea362276878e1f38f97da49964dbea301c15d4aa9d727f97a4c39fd943"
+          },
+          "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/references/source-notes.md": {
+            "bytes": 17240,
+            "sha256": "5073d87450cb60d1b245308dff8807e73749c07de976d50f2fa26694fad81a72"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/LICENSE": {
+            "bytes": 35149,
+            "sha256": "3972dc9744f6499f0f9b2dbf76696f2ae7ad8af9b23dde66d6af86c9dfb36986"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/SKILL.md": {
+            "bytes": 4507,
+            "sha256": "c993bd6ff8e9b5f82f39f175ec8242341136e4ee0808af414e1d94ad3e196c5e"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/metadata.json": {
+            "bytes": 2032,
+            "sha256": "74742851a002a51c208362cdcd2a59fdab5d1734d8492b1f93fbec3cf21e0841"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/references/graph-cache-invalidation.md": {
+            "bytes": 7679,
+            "sha256": "614c9a4e6943d24d7d7be7d632682ce8b1dcc239124a37d9dc1d487113772eac"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/references/model-residency.md": {
+            "bytes": 7191,
+            "sha256": "014c68386e7cdfdb7b4f566836e8b2ff61bd5a5a506e7b4723f83fe80c5d4de7"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/references/repo-provenance.md": {
+            "bytes": 5648,
+            "sha256": "ca996ef515af42a3ce304747fa3f12a31c00533deaace5ba2ce21aefa0a62e2e"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/references/validation-boundaries.md": {
+            "bytes": 3523,
+            "sha256": "1913dedbeb558d872d6a0e015cc11d822910c3cb259a236db29fb60ad8a64263"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md": {
+            "bytes": 12294,
+            "sha256": "4703beb6e5194ef2ff2c0b5ec69f64e35417a4bc0dfb50a10c80a920b77ef3b5"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/metadata.json": {
+            "bytes": 2333,
+            "sha256": "014be24d0e669f4acbc0717dc323eec784026cf185186619d5a19d6d84496131"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/source-notes.md": {
+            "bytes": 12036,
+            "sha256": "fd917bd5ac80bee9c9f4525f8e0300f6d67dc144e2a16632af42a67f3d53d098"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/validation.md": {
+            "bytes": 5402,
+            "sha256": "60e71e70c5519bcdf6c171eec2be2f342e8801bedd746d6005f11adc646dfaff"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/SKILL.md": {
+            "bytes": 11202,
+            "sha256": "f157562d1edc905cb4fc07246567e3c5b844efb9fbf5fe39f0d40553a1b1a307"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/metadata.json": {
+            "bytes": 2156,
+            "sha256": "4d68bf4c1767eaee12ade963bbd551d0e61d1bcd81a984bb02fd3aae46965919"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/references/source-notes.md": {
+            "bytes": 11633,
+            "sha256": "6fa955022f16cd598f44fd6521901d1cb868b06cec545cabe2b82769a641b5d6"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/references/validation.md": {
+            "bytes": 7196,
+            "sha256": "673d93f6700052974b340fab25b3438e0e6344960569ac9abec7d7720f662288"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/LICENSE": {
+            "bytes": 1077,
+            "sha256": "663e51edcd9a3a8f858018ce6924ad2895a24c269c26c39ad75325bcd0f7e5bc"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/SKILL.md": {
+            "bytes": 5432,
+            "sha256": "fdb4d9483c6efa836f9e59617da605fa9c2186e11c2041b05d5c81850ba0475d"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/metadata.json": {
+            "bytes": 4923,
+            "sha256": "fe626a77af3f084f36fd73ffdb6f67976e7dceeafb94ba22078ef03c3a96cfae"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/THIRD_PARTY_NOTICES.md": {
+            "bytes": 1774,
+            "sha256": "a0d77db7d2aca0eda811ab3c386cc32b3143f2c2609a042cf4642954f0f6fb09"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/minimax-h3-apple.md": {
+            "bytes": 12059,
+            "sha256": "57ce01754ba96c378f13a91305ef3d418b926f99b2753f0af99315572bce063a"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/minimax-h3-weight-inventory.json": {
+            "bytes": 10664,
+            "sha256": "fbaacaeec340d3ce893556ee864405a0746af4d13530a38cb3d5aa9d60929271"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/phase-residency.md": {
+            "bytes": 9408,
+            "sha256": "2836a550eb2be7f5a862f51ec7b135ebd9715bf3c4faa84a7ecbc9d9c11f5d8d"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/precision-and-streaming.md": {
+            "bytes": 9440,
+            "sha256": "7e166c89193f4c7aed852ec95346a2de2b5008dac002e2e1862be1b3e57f6093"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/source-identity.json": {
+            "bytes": 7747,
+            "sha256": "730f8dd23648a3f8f47b9726cc613d893a79d1817d515eec609a9b06d8a9ff90"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/source-notes.md": {
+            "bytes": 9448,
+            "sha256": "7304bf9369efc157dc7aa40686dd828cc14db45c49442e908b3d3cc4b329eb86"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/validation.md": {
+            "bytes": 7880,
+            "sha256": "0754873a758e048480938314894e25b504ae0c269549312e7f8e6020d83fe294"
+          },
+          "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/SKILL.md": {
+            "bytes": 9338,
+            "sha256": "95e81b6bc096a954fa73139ea5bc001d1efa6c99747161da62f012b494e5c013"
+          },
+          "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/metadata.json": {
+            "bytes": 1718,
+            "sha256": "1d30b31dc8352fc6a8d7b8abbef8118c840e47e5b90795dae0c758bf9717278c"
+          },
+          "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/references/source-notes.md": {
+            "bytes": 15234,
+            "sha256": "3d12a401e9decb55a2f4f3b2bfa0d360ef5110f63a0469dfe285654e78054190"
+          },
+          "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md": {
+            "bytes": 10812,
+            "sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27"
+          },
+          "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/metadata.json": {
+            "bytes": 572,
+            "sha256": "c221189b8d85fac3bd41ef1c7cd560f1aebc8cfa0381c182d1d571a7169be829"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/SKILL.md": {
+            "bytes": 3788,
+            "sha256": "bf1177cb9985c12a9bf11edb64771018e1e314c4713920de8fd4a989ff70115f"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/metadata.json": {
+            "bytes": 1299,
+            "sha256": "d86cecb615e93ba0f6be09d1ead102cd5a17068c07e1be2549ac5cd8f7519279"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/cache-stop.md": {
+            "bytes": 2026,
+            "sha256": "bba6b5c73261065309cbcf11b1e591eef48916f0c79ceaaf74d98a9fa96b0992"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/generation-batch.md": {
+            "bytes": 1733,
+            "sha256": "f16d3c57d45e63bcaffd3871aa41c76972ee11f7dbf1e12b45d9b170303cad9b"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/source-map.md": {
+            "bytes": 6358,
+            "sha256": "91d26181f246307f2113f94b12ac9587d689b654dcd216e9f14b11cf5ebb650c"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/validation.md": {
+            "bytes": 1665,
+            "sha256": "73b5dd347017baba519ecc85415aead9993c1751d116b9ba4f07689f58c5f783"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/LICENSE": {
+            "bytes": 11352,
+            "sha256": "a54c8553df5d4313d293d1aef3bf2f13952bda8b7c7a1fdd729b2bbe27ee6015"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/SKILL.md": {
+            "bytes": 2304,
+            "sha256": "d75f5b2881b240ec2d9d4f5cc8e3f7d53e4378cb01732a2fb9699da2901ba38c"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/references/repo-provenance.md": {
+            "bytes": 1778,
+            "sha256": "5e8dbe168a8885e209d4b4532944ffa7fead658b23fd6588054d57b22476f63b"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/references/repo-routing-metadata.json": {
+            "bytes": 511,
+            "sha256": "6820707cb9bd877d1635928cf5a0efe38e4fe37dc04036dc70d53e479ed9bf7c"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/references/snapshot.md": {
+            "bytes": 3488,
+            "sha256": "0fb5d776349de845dc314f16ce73a06c9caf5db84dc1c293327d3b3ca1337c80"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-architecture/SKILL.md": {
+            "bytes": 2355,
+            "sha256": "766b2d66d2056dd779abd9e492cb3af6e25899298f44964f565509f130a7cb24"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-architecture/references/runtime.md": {
+            "bytes": 7734,
+            "sha256": "36bba9874fe054acc5217f9c265f5ae309c83943e0d46228370ae1bcdc69421e"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-extension/SKILL.md": {
+            "bytes": 2193,
+            "sha256": "9af38bca840d0004fc2ed049141476fc40fd40dcfa843f38a404a7a314c19425"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-extension/references/integration.md": {
+            "bytes": 5775,
+            "sha256": "6ce9383ca17bacff60a180f8fb8c43ea47eaa37117fb8da6051ad86135d228ef"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-serving/SKILL.md": {
+            "bytes": 2335,
+            "sha256": "990b6bdebfe22612de49dbebcb16810e8abef94d70691aa883b7937c85210eb6"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-serving/references/recipes-and-tuning.md": {
+            "bytes": 8918,
+            "sha256": "94afb762bbe185f58606154b3f0a5af56e7089cb84ea94e26cfcb9f93c95439f"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/LICENSE": {
+            "bytes": 11346,
+            "sha256": "1495e1e757ef4d0925a2350563cf5754bb23c51701a8ec4fb3c5cdcbedae6747"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/SKILL.md": {
+            "bytes": 5466,
+            "sha256": "ae72d1f0cc750699bf79d7a138a97219209bca7b7e5f19cd427d306245eaf841"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/LICENSE": {
+            "bytes": 11346,
+            "sha256": "1495e1e757ef4d0925a2350563cf5754bb23c51701a8ec4fb3c5cdcbedae6747"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/SKILL.md": {
+            "bytes": 5738,
+            "sha256": "b0a4ba4751087126143ef86983798deb0ed00e24b6ff06bf6ef76a9df6e274f2"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/metadata.json": {
+            "bytes": 2844,
+            "sha256": "6ed57dfb4f25ecb34b81f2188b055e05d172ba2581e6a3b605b814b2f733a4d6"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/references/source-notes.md": {
+            "bytes": 19512,
+            "sha256": "d0fcfd1177ae2f351041c5b2abba7f5121cec133006d9e4dca8f74ddb567332d"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/architecture-and-memory.md": {
+            "bytes": 7735,
+            "sha256": "798e747022b7ec135cc4861c2c9ac44fb14b503bb77a0068ee7fbea494b764a7"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/native-candidates.md": {
+            "bytes": 3961,
+            "sha256": "abafcc9a34df196a6365145554887b8d37afb36c6e236c92e50a4a06c8a4ec93"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/parallelism-and-extension.md": {
+            "bytes": 6324,
+            "sha256": "e3be7f4b51264ac68682a64d264e4d96857d5bb9f9d99cae5f327c56984a9e6b"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/repo-provenance.md": {
+            "bytes": 2274,
+            "sha256": "534d95ae8d0b29bcbc8f6707b3d38fb964e31f5b66ba13517a9a6b8f309bbb38"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/repo-routing-metadata.json": {
+            "bytes": 625,
+            "sha256": "b2b879c1cc42ddc2a5b95ae24115eb891562fa9a6e96d151f4ddf6f7719fe73c"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/source-map.md": {
+            "bytes": 8945,
+            "sha256": "c2adad64ceb0c305cdc4859c03012b3e0a4ab26cd0cb289bbd76e192b0cc3794"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/tuning-and-validation.md": {
+            "bytes": 10832,
+            "sha256": "49b3343b3292cc098a787de9d24d968b444d0ef4f73fedb5d8183390e8f39c9c"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-parallel-extension/SKILL.md": {
+            "bytes": 2316,
+            "sha256": "18bd54ebdd629e1ba6dfa589f391c7cc3c5355785c2ae82f741b64ee8022464f"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-runtime-memory/SKILL.md": {
+            "bytes": 2154,
+            "sha256": "33b1a65aba00fca819d2b6cc42202e7708e9155e4d8f4373a2387f4ee2f2fa3e"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-tuning-validation/SKILL.md": {
+            "bytes": 2018,
+            "sha256": "744b2ef9122101d3f50cbbb40fe020e2b3d3888adbe2d86041041e7c7ce424fc"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE": {
+            "bytes": 1062,
+            "sha256": "b53fa08f515cb5a6fff7b9fd8fcd0961a4b80df29d74372d25e1f9171aa042ee"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md": {
+            "bytes": 5504,
+            "sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json": {
+            "bytes": 2202,
+            "sha256": "f7b7581a3f44ba9774480c559408fc66d6a4d6d1dbbeabe4529f1eb7d4dd5ad6"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md": {
+            "bytes": 8136,
+            "sha256": "7b537cb972347dd68e303960f142892346d8b19b884e4d08016ea36a23bdaa7a"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json": {
+            "bytes": 3617,
+            "sha256": "af4830370b6eddc2d176a9ae14da4ab56a90b0fe048a01a89ee051df4365d8e9"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md": {
+            "bytes": 6589,
+            "sha256": "b67affbd7d237d82b7f51be49f1537e9820fe56ec70a71ee76809f9e599ca693"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md": {
+            "bytes": 3394,
+            "sha256": "5dcf5fa7bf734fbe13214f57d2dfb40ff8215d9cd34e072a9e615bf103b82fc2"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/LICENSE": {
+            "bytes": 1093,
+            "sha256": "949b9b0aada7476006469e58ee4ffb0e0f68c2d9137e90a356757b61a09d37f0"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/SKILL.md": {
+            "bytes": 2813,
+            "sha256": "0e5daeb6409c916be45ecd7ce2531ea5f1c60a4a53e2b0ed25dc1e2cdc5c17ed"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/design-space-analysis.md": {
+            "bytes": 27736,
+            "sha256": "11ecafe71530ddd4839268ea1b24b3361a3895012a705859ca06dae29fa72630"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/repo-provenance.md": {
+            "bytes": 2913,
+            "sha256": "05f17d0ae996ad6c12314bdd29f3427574f6aa611ea60effd18598a0f386638b"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/repo-routing-metadata.json": {
+            "bytes": 332,
+            "sha256": "a10ba49b7e6565109b566c105e17551d042075a1ff649c6f371134e3b7402c74"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/trial-format.md": {
+            "bytes": 2291,
+            "sha256": "92d8cf38549268ae4eef382890662d03e18a7dff858ad45ee5eb6f47eb39c8c0"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/troubleshooting.md": {
+            "bytes": 2079,
+            "sha256": "7301cfa5a5f6cfbf9576c3845a96c6b4de8f8a6509f2330970a8e2e4669b3d5a"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/scripts/summarize_trials.py": {
+            "bytes": 5170,
+            "sha256": "6a243a08da15659ced6519f49f52a49a675bd9e952740c3ab03dd21964eefbe2"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/design-space/SKILL.md": {
+            "bytes": 1826,
+            "sha256": "857bea3c64db45dc035728a41523142ac65778b41f63092148b8ab936f971d76"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/design-space/references/design-space.md": {
+            "bytes": 24231,
+            "sha256": "0a428dcd12d5290b554e9da6795a110f94d98f128771602e1ea5f608048b135e"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/tune-execution/SKILL.md": {
+            "bytes": 1145,
+            "sha256": "cdf51c537e7237296d52ef6cda9f2e4c9d9ce3fc87265a369cd9eb7497c53512"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/tune-execution/references/tuning.md": {
+            "bytes": 11310,
+            "sha256": "7b69aa7872d60dd3fb91e41b77a632572070dcf80dd5b5f6dc1722d8540177b1"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/validate-engine/SKILL.md": {
+            "bytes": 4613,
+            "sha256": "90bfd975e25f18895ce0e85f6f35d24149ce026305a4d479aea4924510053460"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/validate-engine/references/validation.md": {
+            "bytes": 16895,
+            "sha256": "dae846f4c47c68731b96ec1d50cb008c03871c72078f1c5b0f4a9423be9ba397"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/LICENSE": {
+            "bytes": 11357,
+            "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/SKILL.md": {
+            "bytes": 4945,
+            "sha256": "df3e63dc28de2dac2aa92824fd88719159138650d96487332e0effca3f0e2a80"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/references/evidence.md": {
+            "bytes": 4689,
+            "sha256": "ae0d51ba56b34ed81fba34de8fd3685bebf2f160b188fa53719936002881d220"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/references/repo-provenance.md": {
+            "bytes": 2221,
+            "sha256": "43bbfe431434ca4ad6157ba35cbb9ce2d89e4065f427abfdd5f372f51ba0c9c9"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/references/repo-routing-metadata.json": {
+            "bytes": 342,
+            "sha256": "7784b4d98d096718c4f4cc42ce86d5ec113c469844e12de4690b0fccca59d329"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-architecture/SKILL.md": {
+            "bytes": 1336,
+            "sha256": "fb3ff50a7b9122a2545a97ef5179011237c4c4ca214dde7236ad7bf79a7c8623"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-architecture/references/architecture.md": {
+            "bytes": 5600,
+            "sha256": "db845eb4207ddeedb8ae381f1b91704a42cb53ba0ee6c7ee581510ba1a412190"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-serving/SKILL.md": {
+            "bytes": 1343,
+            "sha256": "8ab3209b838635395b22deca6aa879172016531d45c2374eb8e498e7f71bb40f"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-serving/references/serving.md": {
+            "bytes": 5049,
+            "sha256": "9bd5eff300d9755137a7600c8f7a7753bb22d09c8853e08062d7e535662ba8a3"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/SKILL.md": {
+            "bytes": 2227,
+            "sha256": "93c83bc4b799b32f50ff54b870cbbde050a790af16d9b7b062039c4987c80148"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/references/abort-acknowledgment.md": {
+            "bytes": 8547,
+            "sha256": "e95ae5b2cc410917ddf5e9066890d0bef9264104529bdca5ac12bdf56fe80ae3"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/references/validation.md": {
+            "bytes": 4632,
+            "sha256": "1542061f9c76fff2acb6529a58238301484b0e23ec0c697d2fdf129e200ea973"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/LICENSE": {
+            "bytes": 11324,
+            "sha256": "50e6751797c50dedd75ef1b8a0d9e42f5f8472e9fbce91f34718e9f97b0c780a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/SKILL.md": {
+            "bytes": 2863,
+            "sha256": "38e522b8abeb0dc140a2194340432ed41dc8b569d33b63d5d4c59aa1092cd2d1"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/SKILL.md": {
+            "bytes": 6664,
+            "sha256": "313733a65cc48442b18dda1c4665139a689963a90e1032a2235bc6f40e3a8e06"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/metadata.json": {
+            "bytes": 781,
+            "sha256": "b43fc3a7ccae59c664fcdf9ff77962edcc07c2a3df8b59324db551642bd23e23"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/references/source-notes.md": {
+            "bytes": 13853,
+            "sha256": "3665b034594fc8cc767c408002f4e16f8803369242405027c7f37ce6c9ecebe9"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/SKILL.md": {
+            "bytes": 6565,
+            "sha256": "5fb9ea1d1373e110d9b506f69963c4857338b7dbf5e38ec9a49b59a0ff683ed0"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/metadata.json": {
+            "bytes": 1372,
+            "sha256": "78459416734f61a8faa819d9a1a3ad7b9a985e8bf8e149813c604d15be6bb92f"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 9006,
+            "sha256": "82b24be82451ee121892f61575d0cf4ecb9920813a25ccb26cce2b45dda6d8b3"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/migration-is-not-free-memory.md": {
+            "bytes": 2145,
+            "sha256": "2ac168f2725cba7fc92dfcc67c01bf2258dff6a2dbe1ddb063bc60c175d82d0a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/parallelism-change-with-state-cost.md": {
+            "bytes": 2069,
+            "sha256": "5accc81086e4500d66f894ef6af9951e1cbb76941f6c631aab5c5561a786d8d8"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/preemption-under-kv-pressure.md": {
+            "bytes": 2630,
+            "sha256": "462200a6d0b193ef01f32374a463e0f2838ca8edf33d9ab898def65ffd0b490a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/knowledge/scheduling-state-boundaries.md": {
+            "bytes": 7488,
+            "sha256": "d4ef29869959b835a5dcb06ddb2fd8e6cff49bb176994dc1d46eb9a852f6228d"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/fastserve.md": {
+            "bytes": 9289,
+            "sha256": "9b4e275843000ef453d508d9c3c9c3f8ea1de5545fbd01d7abe74baa8a4931a7"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/llumnix.md": {
+            "bytes": 10181,
+            "sha256": "f252ce33ac2db017701ca1d49813c5ec1996c4c92900cf35bb6bda7fc59c0f18"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/loongserve.md": {
+            "bytes": 12422,
+            "sha256": "7bc6d1e256d461b04acc8ed856cf99108faceda9cde729ab95bcca596fe019ca"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/orca.md": {
+            "bytes": 8826,
+            "sha256": "f5a3b5870d64226408a8fcda5d69483769d823f8a5efd7f11698c8da483ab195"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/fastserve-readset.md": {
+            "bytes": 7173,
+            "sha256": "9866bd0c78b2b638998b3b632f6bae26aef032462c81abee336d309af5b07831"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/llumnix-readset.md": {
+            "bytes": 9386,
+            "sha256": "3a2819d65ad24087c08530492ef39d8ea046ed1f780d53512f1d23e2b3578487"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/loongserve-readset.md": {
+            "bytes": 12490,
+            "sha256": "f1a383e7bdbbd5a5a472d7f6e1292c8a17d4910ff736ca5a6aa422e1bee74f8c"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/orca-readset.md": {
+            "bytes": 4796,
+            "sha256": "a497ef23b7e60c0fbc15d95b7e07cd20547cfe56f37c24dfe6ed70d79ef7a3cf"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/source-notes.md": {
+            "bytes": 9575,
+            "sha256": "84033fd514773f10c0bddd9914fabb980497ba2b308f226a922c3e1888413827"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/references/evidence-map.json": {
+            "bytes": 14384,
+            "sha256": "dd8bbab5d285ed3a1a19d2f06ad84e8c444251690d8f673def3e2681765ef6c1"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/references/repo-provenance.json": {
+            "bytes": 1313,
+            "sha256": "4e705a964697712e950e02bedc41320806a41311b7f4dcf32f923d60b1d2e919"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/references/repo-provenance.md": {
+            "bytes": 3261,
+            "sha256": "6e841c38c35ff2cd6ff8139cb1a6c8f75701647897f1089a94a5c62928609941"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/references/repo-routing-metadata.json": {
+            "bytes": 432,
+            "sha256": "a20b54f682eb52311e445ac62b95d1935c001296f5cf23b7764cb04477d535c9"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/SKILL.md": {
+            "bytes": 1835,
+            "sha256": "2a945b5fddacfb90d1cd7cb2b2f767067d641061423a65e1c19afecec7f3c06a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/references/capability-boundaries.md": {
+            "bytes": 6560,
+            "sha256": "c8081d03f04f7481080c46c7dcf728962729bc122f49f9502d50cd1e3d04bcf3"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/references/runtime-and-invariants.md": {
+            "bytes": 8265,
+            "sha256": "afa1f967545021f36bc1b59dbf6aa2c367f4d5ece4f08dce755e2017cf50dbe5"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/extend-validate/SKILL.md": {
+            "bytes": 1731,
+            "sha256": "495679e5b99341bf80daeaacd2d4c2699aba16bc16c447cbd8d2fe17f7559798"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/extend-validate/references/change-boundaries-and-validation.md": {
+            "bytes": 7624,
+            "sha256": "f7fdd2d699f09d96fbf5ca9382f9b56113c870db2688d8e1118eb8163519fc4a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/SKILL.md": {
+            "bytes": 1812,
+            "sha256": "0ef4d2e3a96cbb277557e169a3fc8edf0f0504aad03259771b432a33ee008f78"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/references/commands-and-measurements.md": {
+            "bytes": 5667,
+            "sha256": "76ab51083b14595c49fa1c255cb7e5ebb996f14930a51ab3cbe864cc334b5f80"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/references/experiment-decisions.md": {
+            "bytes": 7644,
+            "sha256": "242593c0d386e57104b57a09bc7db97c2bbbcdb905bb68a397cfe612e07fb083"
+          },
+          "src/inferrsi/data/skill_library/skill-index.json": {
+            "bytes": 7514,
+            "sha256": "8ce35e18f7629260d67a95617b9918eecff73ed8d9370dd3d4595c725ada9aa1"
+          },
+          "src/inferrsi/data/skill_library/workflows/cite-check-downstream-pr/SKILL.md": {
+            "bytes": 1093,
+            "sha256": "485932cdf3b1fba5dc69be73e75d38f86a7b79442d26e82a296c963a7598d226"
+          },
+          "src/inferrsi/data/skill_library/workflows/claim-contract/SKILL.md": {
+            "bytes": 1075,
+            "sha256": "3af1dbbe81d91cb05fc2ef15a2ae25d7592a08558e1a5535f49c52f33a488c55"
+          },
+          "src/inferrsi/data/skill_library/workflows/competitor-scan/SKILL.md": {
+            "bytes": 1088,
+            "sha256": "0ede58ace65a9324a538b4816add1684f230ba536d4364fe5a5dd2c0bbed3776"
+          },
+          "src/inferrsi/data/skill_library/workflows/draft-acceptance-issue/SKILL.md": {
+            "bytes": 1230,
+            "sha256": "ce983acfb7d7d3272775e7cf00c92e4ddadd97511b602575a76e5ab7ade01b56"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/SKILL.md": {
+            "bytes": 5681,
+            "sha256": "364beed2ca617e71f4f955bbd024740d411b94eeb9557d7e055d4463cc22caba"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/cross-repository-mechanisms.md": {
+            "bytes": 17050,
+            "sha256": "28efa1b25f548763180b70d8d0ddebb7d4e7227df189f871fd233fe1d4881303"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/import.md": {
+            "bytes": 3463,
+            "sha256": "c8a57db94768cc6cd4193932def75ba67ec453e56ee0e54f176fb82f1af53091"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/model-aware-preparation.md": {
+            "bytes": 9056,
+            "sha256": "787c017aade68ee1133b979d636fa24062302cef6174c001171c963947ee8da4"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/overview.md": {
+            "bytes": 8099,
+            "sha256": "7ed0a966239b9481b82dd349245cfbfe966c030fc916771c91fd0692464ffc56"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/repository-source-scope.md": {
+            "bytes": 11967,
+            "sha256": "85202ca209171c50573c2fc209acb1e937d8211e1652e4ca9c6cc4bce04d03eb"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/source-manifest.json": {
+            "bytes": 1613,
+            "sha256": "0ce1678803d1ede6b485fb6b4b1c88aba909056ee6e3b5af691899ceb2056240"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/SKILL.md": {
+            "bytes": 6502,
+            "sha256": "74cd863df2e80dc2bda1713d24207809173a3116f30325e3487e1e3517cfec1e"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/metadata.json": {
+            "bytes": 3236,
+            "sha256": "12fc9272bd5fbf02dd9b897d8ca40984b72ae54f20492ace5304711bbc4d71c3"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/COMPARISON.md": {
+            "bytes": 4267,
+            "sha256": "38587678a96581f5c2dc697d982127343565504bffa191352811dffcc165aa23"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 5557,
+            "sha256": "fbc4a96b6f6c477706eed83eaa4fcc29a14f881637629d6c004601ed049f2523"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/QA.md": {
+            "bytes": 1853,
+            "sha256": "3f3785ec39a6938a5d1f2e874e99153fe92c0e666dd3e78a499c8b47d891346a"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/README.md": {
+            "bytes": 2151,
+            "sha256": "221785e91c45588f82b3caab6d52f1ea1d102b16139c20a0802921e2789e6d28"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/UNREAD_LEADS.md": {
+            "bytes": 1038,
+            "sha256": "6c1d87ebe491d27666703149eb05ad582c7e5aff77f0ac5c11a7b89c946f96ee"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/cases/methodology-counterexamples.md": {
+            "bytes": 5213,
+            "sha256": "f9fc21753ea41eef77ca020d78d031e945e5dd835b5d09c5ff0e53749ceb2ca4"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/knowledge/evidence-routing.md": {
+            "bytes": 5282,
+            "sha256": "406d5d74c3fef4c1a23d495492e21a084c3c46495a1a84b385703ed104358874"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/learning-to-commit.md": {
+            "bytes": 2847,
+            "sha256": "9c2d1d738e896d05c00f53e6adf14c9df2501ecc3f0020e71b5a1c2192547086"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/protocol-details.md": {
+            "bytes": 6520,
+            "sha256": "41298c6bde3a28f0712fa06a18bbc4ac59aea6d17cc9eb1318744f4121e3bab6"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/readset.md": {
+            "bytes": 7086,
+            "sha256": "6f953d20d8958c68339c9ceeae348604d0fea3e9c6ccb9b64a63d3b210426a22"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/repo-to-skill.md": {
+            "bytes": 3151,
+            "sha256": "5ebdac2a2bbbe6614f9d4096bc9895bd685acdf76414e2c3c1fb7b50d57c738d"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/skillopt.md": {
+            "bytes": 3040,
+            "sha256": "148075daa67e19f4b4351710498438d17c279a9bce56b2d01cab26ff3be33e99"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/source-manifest.json": {
+            "bytes": 2431,
+            "sha256": "716cd49ed01824ee9f6b6fc55ebf2b24fdd5d5a3e677f6e28e34a472616bbe25"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/trace2skill.md": {
+            "bytes": 2917,
+            "sha256": "230d13b77ba819b5e3edee147523e8cad6bfce95afccc140fb4946d6c8747a2e"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/updates/bounded-updates.md": {
+            "bytes": 7630,
+            "sha256": "d02c5bd0109846d1e1987de074810321cb4af277cd3f66ab2ac268aadd5ee142"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/skill-factory-design-20261014.md": {
+            "bytes": 9465,
+            "sha256": "fb8ce1ad83b7a6ddc657e89ee2eebb710873b213cfd85bdd6d823c8691b3cb9f"
+          },
+          "src/inferrsi/data/skill_library/workflows/rsi-evidence/SKILL.md": {
+            "bytes": 3680,
+            "sha256": "a55610874b658b86a39a1a049efb4e4eab0140f6579ed74dd7718a4ca8cedaca"
+          }
+        },
+        "sha256": "6e19365a0ad8e2951af0f3106f93aa545b4ad00b6e0f0317a20f32065b6a05fb"
+      },
+      "environment": {
+        "filesystem_cache": "not controlled; process cold is not OS-cache cold",
+        "implementation": "CPython",
+        "logical_cpu_count": 8,
+        "machine": "AMD64",
+        "perf_counter_resolution_seconds": 1e-07,
+        "python_executable_sha256": "737a7e3b71e3578f8432acc7dd88c452e593622c544bc13da4789d69c63da5ae",
+        "python_version": "3.12.6",
+        "release": "10",
+        "site_initializers_disabled": true,
+        "system": "Windows"
+      },
+      "git_head": "4817f92121df73fd9cd5fb4a319e3c56a68b4a5b",
+      "knowledge_record_count": 108,
+      "library": {
+        "bytes": 2294958,
+        "file_count": 342,
+        "sha256": "be8726296e6186601692dc2d95d99bfbba76f0b4c758de3a88a9b2a523eaa37e"
+      },
+      "modules": {
+        "inferrsi": {
+          "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+          "source": "src/inferrsi/__init__.py"
+        },
+        "inferrsi.catalog": {
+          "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+          "source": "src/inferrsi/catalog.py"
+        },
+        "inferrsi.cli": {
+          "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+          "source": "src/inferrsi/cli.py"
+        },
+        "inferrsi.knowledge_operations": {
+          "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+          "source": "src/inferrsi/knowledge_operations.py"
+        }
+      },
+      "packaged_skill_count": 60,
+      "python_sources": {
+        "bytes": 274444,
+        "file_count": 53,
+        "files": {
+          "src/inferrsi/__init__.py": {
+            "bytes": 102,
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47"
+          },
+          "src/inferrsi/__main__.py": {
+            "bytes": 48,
+            "sha256": "935a1c1166b0c1ea35a82256345000bf2c73ded718d77773bc27a71ecce28f7d"
+          },
+          "src/inferrsi/_skill_selection.py": {
+            "bytes": 5790,
+            "sha256": "4640399f4013425589f6702e2db844820619ba35967d1c3519978eb307168440"
+          },
+          "src/inferrsi/arguments.py": {
+            "bytes": 12200,
+            "sha256": "4456600cf4a362cf069594fe36db46932cf5cb14ddd37683aaf3538f5cf76b4d"
+          },
+          "src/inferrsi/catalog.py": {
+            "bytes": 6855,
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5"
+          },
+          "src/inferrsi/claim.py": {
+            "bytes": 805,
+            "sha256": "5f24c4fa91b48f0bb4d4554531607763fc8fc108e90f332816f06cfb520b5cc1"
+          },
+          "src/inferrsi/cli.py": {
+            "bytes": 14205,
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88"
+          },
+          "src/inferrsi/contracts.py": {
+            "bytes": 119,
+            "sha256": "03000cb9246ce8c0967147533f7ad25df918ba2674839801453ffb870f496490"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/scripts/roofline.py": {
+            "bytes": 9004,
+            "sha256": "524de48a6764401695fe1444f1fa3d38f363069b0cd2fba4715a86ca4b28a5d4"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/scripts/summarize_trials.py": {
+            "bytes": 5170,
+            "sha256": "6a243a08da15659ced6519f49f52a49a675bd9e952740c3ab03dd21964eefbe2"
+          },
+          "src/inferrsi/evaluation.py": {
+            "bytes": 183,
+            "sha256": "482c40710d428206c59fe3696703248cd86c4b3c7d19ad41017c1369bd6d780e"
+          },
+          "src/inferrsi/evidence_paths.py": {
+            "bytes": 859,
+            "sha256": "cc46231e0b8334ef5742defdda20e0093e22b30c86af59df1fb3ae4b5caa70fa"
+          },
+          "src/inferrsi/execution.py": {
+            "bytes": 755,
+            "sha256": "f43d8cbb4d43a72a7ff8a10b29a5fb9cf4cb22af54dd35943283c843fba0da92"
+          },
+          "src/inferrsi/harness/__init__.py": {
+            "bytes": 177,
+            "sha256": "1006e4713003f44bfb32b2e385a918e1652b9e9833801e4555674a71e57fb0d5"
+          },
+          "src/inferrsi/harness/evaluation.py": {
+            "bytes": 4160,
+            "sha256": "9b00b2e93959f16b7b76bb4b257df7dcf48520a06fd6adae96e9a165a05e6e80"
+          },
+          "src/inferrsi/harness/improvement.py": {
+            "bytes": 5777,
+            "sha256": "3faa0f5928a707a98f35f45d1ca18e226c898941e6a07b3829eebeb53e70aeb8"
+          },
+          "src/inferrsi/harness/reporting.py": {
+            "bytes": 3566,
+            "sha256": "38c58f02c44e8c8aeacf46527e142be0e3142a83e4b2693246d0ca76c0ddd7c3"
+          },
+          "src/inferrsi/harness/runner.py": {
+            "bytes": 6567,
+            "sha256": "a566307b9b586655941b627f428b5121ed38d7ac80bccd1e1799e443fb0de911"
+          },
+          "src/inferrsi/harness/scenarios/__init__.py": {
+            "bytes": 79,
+            "sha256": "4e97913c806077f0cadfeb29d986d2fa75b55ab8766cb86e6e3e9a195a9a1f5a"
+          },
+          "src/inferrsi/harness/scenarios/external.py": {
+            "bytes": 1016,
+            "sha256": "84d5f6920501a9d882cc4fffbc721d2051fdbf7c90694bde9cbfba3c2b866304"
+          },
+          "src/inferrsi/harness/scenarios/media.py": {
+            "bytes": 2914,
+            "sha256": "67d979a3ed6f13a8e3e1e40a304f78e5a3e6c2556b3aa8a11c41156745d7050a"
+          },
+          "src/inferrsi/harness/skill_evaluation.py": {
+            "bytes": 11093,
+            "sha256": "b710b8e53758757efd68c703449eb4bab7eb7ec50ae3dfc52a010cdebd8c3bf5"
+          },
+          "src/inferrsi/infra/__init__.py": {
+            "bytes": 85,
+            "sha256": "8779897733779266841b0f79b84b1dc43f0fa57c013855c906c809829ecfc77c"
+          },
+          "src/inferrsi/infra/artifacts.py": {
+            "bytes": 1040,
+            "sha256": "e72e1bebdbb88e42de430238489252386851f9e6645ed7780f2aad2e90338120"
+          },
+          "src/inferrsi/infra/campaign.py": {
+            "bytes": 1028,
+            "sha256": "be505fba1592611bd84ceaca55afbac027205413a2639baefa0c3bb343230209"
+          },
+          "src/inferrsi/infra/commands.py": {
+            "bytes": 811,
+            "sha256": "1890f115b3a2c5c6832173bf60432ad595bd6bf1c69783ecd460c6f8961194f1"
+          },
+          "src/inferrsi/infra/environment.py": {
+            "bytes": 2067,
+            "sha256": "60e46c2aefc216db202b41f44542e7b893e08e0721c667b9ccc50411d1ac84ee"
+          },
+          "src/inferrsi/infra/executors/__init__.py": {
+            "bytes": 64,
+            "sha256": "327c3cd5c821c832ff260a41dba0ee1e439cdabefc9edfb9cad1cd38d14bc87a"
+          },
+          "src/inferrsi/infra/executors/local.py": {
+            "bytes": 3654,
+            "sha256": "37bcbca9f3261c8c58a8bf806c9da40720b18d0676c82e8ed24c3901ac9bbe77"
+          },
+          "src/inferrsi/infra/http.py": {
+            "bytes": 634,
+            "sha256": "f65d6c2b55b6a89694caaf6dcdf7d74f417e9ced103eca1b3f218e305c6b6cb2"
+          },
+          "src/inferrsi/infra/inspection.py": {
+            "bytes": 2415,
+            "sha256": "d1c66deb0cced529bca996d2f88b60f51fa3417ff1cae365c650fec808946ecd"
+          },
+          "src/inferrsi/infra/platforms/__init__.py": {
+            "bytes": 1494,
+            "sha256": "9e88e8b9a396ed90c221e1585135c0a1b58da41c1bb73ed9237c53b6684abfed"
+          },
+          "src/inferrsi/infra/platforms/cuda.py": {
+            "bytes": 3309,
+            "sha256": "9e0e4276e1b2984605493b14b82fba3e86f6635204719c318302e624f66f8b6f"
+          },
+          "src/inferrsi/infra/platforms/mlx.py": {
+            "bytes": 1971,
+            "sha256": "f3b81470db28bd6183ccca2e763c5e079686bba6cdfe6c08ebe2cdf64216f2e3"
+          },
+          "src/inferrsi/infra/processes.py": {
+            "bytes": 8151,
+            "sha256": "3b96ebc7a50913e78e86ff3dcba74c8ee8d03d547d9591ca6488f390ac1e6189"
+          },
+          "src/inferrsi/infra/targets/__init__.py": {
+            "bytes": 186,
+            "sha256": "1f3ea75dded7c6cc967754c7b6d4f30ef00da8191a749938aabc3c6bb8f86eb3"
+          },
+          "src/inferrsi/infra/targets/git.py": {
+            "bytes": 2910,
+            "sha256": "6327c4f126086f310b6de309fc7659c9759da28ed76efaeeb94a3d2fba29a0c2"
+          },
+          "src/inferrsi/inspection.py": {
+            "bytes": 692,
+            "sha256": "afa44bf4cdfbb58a318c21f9b292ca142d1bbb22750e35a4cc8244da5350ebb3"
+          },
+          "src/inferrsi/knowledge_operations.py": {
+            "bytes": 27444,
+            "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b"
+          },
+          "src/inferrsi/knowledge_store.py": {
+            "bytes": 13700,
+            "sha256": "5e794e28906ce4687aff7f14a507e267184640fc8a4fac19e0ced279ce6a6cb6"
+          },
+          "src/inferrsi/model/__init__.py": {
+            "bytes": 77,
+            "sha256": "3c1dba4f5159fd9fbc8d4d96e2a83c8441491c1b1dc7b42031a3fc1782ed9d9f"
+          },
+          "src/inferrsi/model/anthropic_agent.py": {
+            "bytes": 24696,
+            "sha256": "b0ef9cd542544292eeb90ab05ef410be9ed6e5d1a40d729890f49562c28c12f0"
+          },
+          "src/inferrsi/model/protocols/__init__.py": {
+            "bytes": 85,
+            "sha256": "5a2f54483b164de04e309dbebb87a700b213682b0f44a50d89a752dd71a4ce47"
+          },
+          "src/inferrsi/model/protocols/image.py": {
+            "bytes": 2047,
+            "sha256": "251a1f8f885da2cd8e9d3baba766c525afe293091606b4e8668c43ba2cb7cb5f"
+          },
+          "src/inferrsi/model/protocols/metrics.py": {
+            "bytes": 610,
+            "sha256": "17db16d6ef15b41fb60d8f09507893accc43f9cb3ad686b8dc4e9839713ebc56"
+          },
+          "src/inferrsi/model/validators/__init__.py": {
+            "bytes": 65,
+            "sha256": "1d4b6ffa968e283219ddcc8a09de9280238a8d65ae2dc8cd2b60889877a8eb0b"
+          },
+          "src/inferrsi/model/validators/flare.py": {
+            "bytes": 19975,
+            "sha256": "3696b0609c772f2222b448833608565a033123d7e92b11fbc545dd356fec590f"
+          },
+          "src/inferrsi/model/validators/image.py": {
+            "bytes": 3204,
+            "sha256": "79470b6b06a22d53ec6c40636a0b3d821a2e6ee2f7bb7aed3dac7241504add1a"
+          },
+          "src/inferrsi/reporting.py": {
+            "bytes": 127,
+            "sha256": "1f771890ab3e69b1aa9246b548a7825596bdf904f6c395dd218208084266853d"
+          },
+          "src/inferrsi/research.py": {
+            "bytes": 5213,
+            "sha256": "9bfe59dd17b5476c18ce50d671711053c62800521c50d6b710e8140a2b4dca04"
+          },
+          "src/inferrsi/runspec.py": {
+            "bytes": 4398,
+            "sha256": "dca0952425caf42095ef0c7e2feb8ef4740cc422ecb61e777f203962d9b87709"
+          },
+          "src/inferrsi/skill_catalog.py": {
+            "bytes": 9938,
+            "sha256": "65f5a1f3f6894009abb98ec2fec30a63cc1bb2f6ce6aeb01f9f7a3dce0439f2d"
+          },
+          "src/inferrsi/summary_chart.py": {
+            "bytes": 40910,
+            "sha256": "d29023d60f6be804c1b01c9eb73765d4824d6c3bb870fe22c99a24544bc1d8bc"
+          }
+        },
+        "sha256": "4faa529dc47f23c24f8c5b8e0110d8e413aab20df5ad61db591fb751564a6448"
+      },
+      "tracked_source_dirty": false
+    }
+  },
+  "pairs": [
+    {
+      "pair": 0,
+      "order": "AB",
+      "A": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 1598122600,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 2368962700,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 57596100,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 1930111100,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 1840695000,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 55828800,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 1721433000,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 2317566300,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 69933100,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 1842333500,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 2279583500,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 82093200,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 2470582500,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 2657212800,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 2345841000,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 2474300900,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 643242400,
+            "decode_ns": 29000,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 3111096000,
+            "decode_ns": 40000,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 2849083500,
+            "decode_ns": 3763000,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 2785271200,
+            "decode_ns": 7065600,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      },
+      "B": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 1923862500,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 2126043000,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 38347200,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 1586684000,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 1725672900,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 48281500,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 1726587100,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 1530965600,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 45815100,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 1272102400,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 1407347700,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 42448300,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 1217508600,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 1356572000,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 1412448200,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 1410985400,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 401732400,
+            "decode_ns": 12900,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 2727919900,
+            "decode_ns": 30700,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 2173445200,
+            "decode_ns": 2006900,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 2131496100,
+            "decode_ns": 4209800,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      }
+    },
+    {
+      "pair": 1,
+      "order": "BA",
+      "B": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 2568397400,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 1850186000,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 56018800,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 1563240700,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 2104588700,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 63694600,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 1897882700,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 3220638000,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 54528400,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 1656563000,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 1992348500,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 58071400,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 1703707700,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 1870658500,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 1546350400,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 1755327600,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 536016000,
+            "decode_ns": 16700,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 2644827800,
+            "decode_ns": 23400,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 2201420800,
+            "decode_ns": 1965300,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 2577305400,
+            "decode_ns": 4404100,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      },
+      "A": {
+        "api": [
+          {
+            "case": "dev01",
+            "mode": "hot_relevance",
+            "ns": 2685380000,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "cold_relevance",
+            "ns": 2259572100,
+            "output": {
+              "serialized_bytes": 434223,
+              "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+            }
+          },
+          {
+            "case": "dev01",
+            "mode": "search",
+            "ns": 49892600,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_relevance",
+            "ns": 1628240700,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_relevance",
+            "ns": 1925371200,
+            "output": {
+              "serialized_bytes": 430369,
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "search",
+            "ns": 54436400,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_relevance",
+            "ns": 1815182800,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_relevance",
+            "ns": 1954253900,
+            "output": {
+              "serialized_bytes": 435629,
+              "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "search",
+            "ns": 65386000,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "hot_relevance",
+            "ns": 1840794000,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "cold_relevance",
+            "ns": 1798481500,
+            "output": {
+              "serialized_bytes": 435641,
+              "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+            }
+          },
+          {
+            "case": "dev13",
+            "mode": "search",
+            "ns": 62451800,
+            "output": {
+              "serialized_bytes": 2,
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "hot_default_diagnostic",
+            "ns": 1639306600,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev05",
+            "mode": "cold_default_diagnostic",
+            "ns": 1817855800,
+            "output": {
+              "serialized_bytes": 312,
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "hot_context_material",
+            "ns": 1670411400,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          },
+          {
+            "case": "dev09",
+            "mode": "cold_context_material",
+            "ns": 2067584500,
+            "output": {
+              "serialized_bytes": 541917,
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+            }
+          }
+        ],
+        "modules": {
+          "inferrsi": {
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+            "source": "src/inferrsi/__init__.py"
+          },
+          "inferrsi.catalog": {
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+            "source": "src/inferrsi/catalog.py"
+          },
+          "inferrsi.cli": {
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+            "source": "src/inferrsi/cli.py"
+          },
+          "inferrsi.knowledge_operations": {
+            "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+            "source": "src/inferrsi/knowledge_operations.py"
+          }
+        },
+        "warmup_cache_state": {
+          "library_populated": true,
+          "records_populated": true
+        },
+        "cli": [
+          {
+            "mode": "search",
+            "case": "dev05",
+            "process_ns": 567731900,
+            "decode_ns": 12900,
+            "scope": "repo=inferrsi",
+            "stdout_sha256": "a5338d955b09046ec0b16f3a9625b7955c763aae07dc722e474e6078745f932f",
+            "stdout_bytes": 4,
+            "output": {
+              "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945",
+              "serialized_bytes": 2
+            }
+          },
+          {
+            "mode": "default_diagnostic",
+            "case": "dev05",
+            "process_ns": 2796663100,
+            "decode_ns": 23200,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "ed878194d9e2d2359c59a7100724705ad6643fc4620ee09907df5f22b6244159",
+            "stdout_bytes": 368,
+            "output": {
+              "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859",
+              "serialized_bytes": 312
+            }
+          },
+          {
+            "mode": "relevance",
+            "case": "dev05",
+            "process_ns": 2330269000,
+            "decode_ns": 2204000,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "fc665edca70efc1a745f4cf8d151ae2fde5c9369ecc4fc166fd638591957beeb",
+            "stdout_bytes": 470481,
+            "output": {
+              "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc",
+              "serialized_bytes": 430369
+            }
+          },
+          {
+            "mode": "context_material",
+            "case": "dev09",
+            "process_ns": 2420095100,
+            "decode_ns": 4290300,
+            "scope": "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": "d12cc142402f841f3d695cedc32edef46d4f5c1060dcb583749dd3cd9608cbf7",
+            "stdout_bytes": 586103,
+            "output": {
+              "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b",
+              "serialized_bytes": 541917
+            }
+          }
+        ]
+      }
+    }
+  ],
+  "elapsed_ns": 175635254600,
+  "summary": {
+    "primary": {
+      "paired_median_reduction": 0.04903038727940762,
+      "confidence_95": [
+        0.027771264850477206,
+        0.07028950970833803
+      ],
+      "paired_reductions": [
+        0.07028950970833803,
+        0.027771264850477206
+      ],
+      "bootstrap": {
+        "method": "percentile bootstrap of paired reductions; each draw samples 20 reductions with replacement, then computes their median",
+        "iterations": 10000,
+        "seed": 19020261027,
+        "confidence": 0.95,
+        "quantiles": [
+          0.025,
+          0.975
+        ],
+        "quantile_method": "linear interpolation at (n-1)*p"
+      }
+    },
+    "performance_gate_passed": false,
+    "output_parity_failures": [],
+    "latency_ns": {
+      "api/cold_context_material/dev09/A": {
+        "n": 2,
+        "p50": 2270942700.0,
+        "p95": 2453965080.0
+      },
+      "api/cold_context_material/dev09/B": {
+        "n": 2,
+        "p50": 1583156500.0,
+        "p95": 1738110490.0
+      },
+      "api/cold_default_diagnostic/dev05/A": {
+        "n": 2,
+        "p50": 2237534300.0,
+        "p95": 2615244950.0
+      },
+      "api/cold_default_diagnostic/dev05/B": {
+        "n": 2,
+        "p50": 1613615250.0,
+        "p95": 1844954175.0
+      },
+      "api/cold_relevance/dev01/A": {
+        "n": 2,
+        "p50": 2314267400.0,
+        "p95": 2363493170.0
+      },
+      "api/cold_relevance/dev01/B": {
+        "n": 2,
+        "p50": 1988114500.0,
+        "p95": 2112250150.0
+      },
+      "api/cold_relevance/dev05/A": {
+        "n": 2,
+        "p50": 1883033100.0,
+        "p95": 1921137390.0
+      },
+      "api/cold_relevance/dev05/B": {
+        "n": 2,
+        "p50": 1915130800.0,
+        "p95": 2085642910.0
+      },
+      "api/cold_relevance/dev09/A": {
+        "n": 2,
+        "p50": 2135910100.0,
+        "p95": 2299400680.0
+      },
+      "api/cold_relevance/dev09/B": {
+        "n": 2,
+        "p50": 2375801800.0,
+        "p95": 3136154380.0
+      },
+      "api/cold_relevance/dev13/A": {
+        "n": 2,
+        "p50": 2039032500.0,
+        "p95": 2255528400.0
+      },
+      "api/cold_relevance/dev13/B": {
+        "n": 2,
+        "p50": 1699848100.0,
+        "p95": 1963098460.0
+      },
+      "api/hot_context_material/dev09/A": {
+        "n": 2,
+        "p50": 2008126200.0,
+        "p95": 2312069520.0
+      },
+      "api/hot_context_material/dev09/B": {
+        "n": 2,
+        "p50": 1479399300.0,
+        "p95": 1539655290.0
+      },
+      "api/hot_default_diagnostic/dev05/A": {
+        "n": 2,
+        "p50": 2054944550.0,
+        "p95": 2429018705.0
+      },
+      "api/hot_default_diagnostic/dev05/B": {
+        "n": 2,
+        "p50": 1460608150.0,
+        "p95": 1679397745.0
+      },
+      "api/hot_relevance/dev01/A": {
+        "n": 2,
+        "p50": 2141751300.0,
+        "p95": 2631017130.0
+      },
+      "api/hot_relevance/dev01/B": {
+        "n": 2,
+        "p50": 2246129950.0,
+        "p95": 2536170655.0
+      },
+      "api/hot_relevance/dev05/A": {
+        "n": 2,
+        "p50": 1779175900.0,
+        "p95": 1915017580.0
+      },
+      "api/hot_relevance/dev05/B": {
+        "n": 2,
+        "p50": 1574962350.0,
+        "p95": 1585511835.0
+      },
+      "api/hot_relevance/dev09/A": {
+        "n": 2,
+        "p50": 1768307900.0,
+        "p95": 1810495310.0
+      },
+      "api/hot_relevance/dev09/B": {
+        "n": 2,
+        "p50": 1812234900.0,
+        "p95": 1889317920.0
+      },
+      "api/hot_relevance/dev13/A": {
+        "n": 2,
+        "p50": 1841563750.0,
+        "p95": 1842256525.0
+      },
+      "api/hot_relevance/dev13/B": {
+        "n": 2,
+        "p50": 1464332700.0,
+        "p95": 1637339970.0
+      },
+      "api/search/dev01/A": {
+        "n": 2,
+        "p50": 53744350.0,
+        "p95": 57210925.0
+      },
+      "api/search/dev01/B": {
+        "n": 2,
+        "p50": 47183000.0,
+        "p95": 55135220.0
+      },
+      "api/search/dev05/A": {
+        "n": 2,
+        "p50": 55132600.0,
+        "p95": 55759180.0
+      },
+      "api/search/dev05/B": {
+        "n": 2,
+        "p50": 55988050.0,
+        "p95": 62923945.0
+      },
+      "api/search/dev09/A": {
+        "n": 2,
+        "p50": 67659550.0,
+        "p95": 69705745.0
+      },
+      "api/search/dev09/B": {
+        "n": 2,
+        "p50": 50171750.0,
+        "p95": 54092735.0
+      },
+      "api/search/dev13/A": {
+        "n": 2,
+        "p50": 72272500.0,
+        "p95": 81111130.0
+      },
+      "api/search/dev13/B": {
+        "n": 2,
+        "p50": 50259850.0,
+        "p95": 57290245.0
+      },
+      "cli/context_material/dev09/decode_ns/A": {
+        "n": 2,
+        "p50": 5677950.0,
+        "p95": 6926835.0
+      },
+      "cli/context_material/dev09/decode_ns/B": {
+        "n": 2,
+        "p50": 4306950.0,
+        "p95": 4394385.0
+      },
+      "cli/context_material/dev09/process_ns/A": {
+        "n": 2,
+        "p50": 2602683150.0,
+        "p95": 2767012395.0
+      },
+      "cli/context_material/dev09/process_ns/B": {
+        "n": 2,
+        "p50": 2354400750.0,
+        "p95": 2555014935.0
+      },
+      "cli/default_diagnostic/dev05/decode_ns/A": {
+        "n": 2,
+        "p50": 31600.0,
+        "p95": 39160.0
+      },
+      "cli/default_diagnostic/dev05/decode_ns/B": {
+        "n": 2,
+        "p50": 27050.0,
+        "p95": 30335.0
+      },
+      "cli/default_diagnostic/dev05/process_ns/A": {
+        "n": 2,
+        "p50": 2953879550.0,
+        "p95": 3095374355.0
+      },
+      "cli/default_diagnostic/dev05/process_ns/B": {
+        "n": 2,
+        "p50": 2686373850.0,
+        "p95": 2723765295.0
+      },
+      "cli/relevance/dev05/decode_ns/A": {
+        "n": 2,
+        "p50": 2983500.0,
+        "p95": 3685050.0
+      },
+      "cli/relevance/dev05/decode_ns/B": {
+        "n": 2,
+        "p50": 1986100.0,
+        "p95": 2004820.0
+      },
+      "cli/relevance/dev05/process_ns/A": {
+        "n": 2,
+        "p50": 2589676250.0,
+        "p95": 2823142775.0
+      },
+      "cli/relevance/dev05/process_ns/B": {
+        "n": 2,
+        "p50": 2187433000.0,
+        "p95": 2200022020.0
+      },
+      "cli/search/dev05/decode_ns/A": {
+        "n": 2,
+        "p50": 20950.0,
+        "p95": 28195.0
+      },
+      "cli/search/dev05/decode_ns/B": {
+        "n": 2,
+        "p50": 14800.0,
+        "p95": 16510.0
+      },
+      "cli/search/dev05/process_ns/A": {
+        "n": 2,
+        "p50": 605487150.0,
+        "p95": 639466875.0
+      },
+      "cli/search/dev05/process_ns/B": {
+        "n": 2,
+        "p50": 468874200.0,
+        "p95": 529301820.0
+      }
+    }
+  },
+  "source_and_corpus_unchanged_at_end": {
+    "A": true,
+    "B": true
+  },
+  "completed_at_utc": "2026-10-27T11:16:18.113257+00:00"
+}
diff --git a/evidence/knowledge-retrieval-20261027/results/dry-run-identity.json b/evidence/knowledge-retrieval-20261027/results/dry-run-identity.json
new file mode 100644
index 00000000..1d4272dd
--- /dev/null
+++ b/evidence/knowledge-retrieval-20261027/results/dry-run-identity.json
@@ -0,0 +1,2099 @@
+{
+  "corpus": {
+    "bytes": 2883003,
+    "file_count": 458,
+    "files": {
+      "src/inferrsi/data/knowledge_packs/README.md": {
+        "bytes": 14172,
+        "sha256": "b877e67d9cccf0dcbe57419eeb3ff9374e5bd03ff8eec63adef1025c446140e9"
+      },
+      "src/inferrsi/data/knowledge_packs/dependency/diffusers/cuda/diffusion/group-offload-transfer-overlap.json": {
+        "bytes": 3277,
+        "sha256": "9bec3d912192efa2f2c853d3fb70bdae21a75875630a02294fcd084b18c30414"
+      },
+      "src/inferrsi/data/knowledge_packs/dependency/flashattention/cuda/attention/generation-specific-dispatch.json": {
+        "bytes": 3085,
+        "sha256": "90290d7bc80b1626863e0520c91bb196d93ebfe08190a450e9de59c6505ae8ad"
+      },
+      "src/inferrsi/data/knowledge_packs/dependency/flashinfer/cuda/llm/paged-decode-dispatch.json": {
+        "bytes": 3745,
+        "sha256": "0f2986494fcc9b751df71e598fd61caa68e28bc4b0ac401e4cc763bc3f75283b"
+      },
+      "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/cross-stream-dependency.json": {
+        "bytes": 3809,
+        "sha256": "b829315e55adf8e11f0bd581d5f45dd130d7eb3f81ce7de532f63c7b9c27cbb5"
+      },
+      "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/evaluation-completion.json": {
+        "bytes": 3841,
+        "sha256": "19d2e506ac4f50895776a83a062059aa7a052175c8e4b21dd0d46e1a5a6716a1"
+      },
+      "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/post-detachment-buffer-lifetime.json": {
+        "bytes": 4663,
+        "sha256": "18d78c0e542dacb42ca33e5264d92aef80fa06d163fd16d4feafd5ed0827ab99"
+      },
+      "src/inferrsi/data/knowledge_packs/dependency/pytorch/cuda/compiler/cudagraph-shapes-and-lifetimes.json": {
+        "bytes": 3623,
+        "sha256": "7fef4122ff794967e61f55d02a4a6b558d7e498d9513607f5ca86db6386928a8"
+      },
+      "src/inferrsi/data/knowledge_packs/dependency/transformers/common/llm/kv-cache-strategy-contracts.json": {
+        "bytes": 3072,
+        "sha256": "a2df26052112a9f3fc6242b2bd636ab13991442db57e56646168d6c52032d38d"
+      },
+      "src/inferrsi/data/knowledge_packs/dependency/triton/common/kernel/bandwidth-fusion-resource-limits.json": {
+        "bytes": 2941,
+        "sha256": "6e9673d1a600a6ad34820e4325f2a6417f44f0f96d8ed0bcf21466d998f19ee4"
+      },
+      "src/inferrsi/data/knowledge_packs/peer/cache_dit/common/diffusion/residual-cache-calibration.json": {
+        "bytes": 3841,
+        "sha256": "bb9509ad0225a91b52bc6787477581d05f0ecd5e44fa59f08f718eea36a2ac45"
+      },
+      "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/bounded-windowed-replay.json": {
+        "bytes": 10959,
+        "sha256": "43613225e9ccdbd34c0f7e3e0767887b9bbfd0f38b2c21684677d99e31a432f0"
+      },
+      "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/speculative-verification-budget.json": {
+        "bytes": 10177,
+        "sha256": "a729e1fc00be44e1480a5dc65bbcdd58df8599681df74dbc651a5a49f27e9028"
+      },
+      "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/zero-overhead-deferred-scheduling.json": {
+        "bytes": 8191,
+        "sha256": "ed5426af145f066f6183bd316c4073c46f324f5ee83aa9bd616387c0d0e7149c"
+      },
+      "src/inferrsi/data/knowledge_packs/peer/lmdeploy/cuda/llm/kv-quantization-capacity-cost.json": {
+        "bytes": 3366,
+        "sha256": "873997b38981ccc03ff73e6fb84e96022f293a019843efd8c4aa50f27878ac9a"
+      },
+      "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/codec-constants-and-state.json": {
+        "bytes": 4250,
+        "sha256": "1cecd7905fef5af3ef8952a4d803900a7df91d662ebae34db47cc0dbf352d4e7"
+      },
+      "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/fixed-customvoice-prompt.json": {
+        "bytes": 3660,
+        "sha256": "ea5e8fdf54cb731b255b73aa0cf47ef53fac103baa2bd5246be45f95d54cb593"
+      },
+      "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/graph-batching-state-ownership.json": {
+        "bytes": 3950,
+        "sha256": "130741a9f7ac9d366784d6c4dd42860262235caff0c277de1e8a8f8ec933b37f"
+      },
+      "src/inferrsi/data/knowledge_packs/peer/sglang/cuda/llm/hierarchical-prefix-cache.json": {
+        "bytes": 3641,
+        "sha256": "b50ca58df28d4343545d07adf0da23f8f181297b30aba7870e5690630e7d7315"
+      },
+      "src/inferrsi/data/knowledge_packs/peer/tensorrt_llm/cuda/llm/prefill-decode-disaggregation.json": {
+        "bytes": 2967,
+        "sha256": "349856dd98628f94987e014889cd1b15d0e3ae6ab6c576cdde917f28fbafb102"
+      },
+      "src/inferrsi/data/knowledge_packs/peer/xdit/cuda/diffusion/parallelism-freshness-and-topology.json": {
+        "bytes": 3236,
+        "sha256": "3674b5916cb7b384951bf3e449b6796b161878daabcc1d7e7f6d6f8ff861cc17"
+      },
+      "src/inferrsi/data/repo_packs/afd_plugin/pack.json": {
+        "bytes": 502,
+        "sha256": "00042abeabdc9c0e24f1a4edaba47e233df398aef0ad4e181fc70191d383da5f"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-aclgraph-capture-replay.json": {
+        "bytes": 4311,
+        "sha256": "327414443ea9537ae320aa405912c338351bc06cefdc0b99352e1eea36f61dc3"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-device-health.json": {
+        "bytes": 4115,
+        "sha256": "633fa5fdcea1e1e9edda8a218d6a3fde0d12637848348e1e2fe7d1b0bd744a16"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-distributed-parallel.json": {
+        "bytes": 5841,
+        "sha256": "709ca0d061e4d6917fe8fa049fe93165e71dfecf96c7bf264ced25346381f603"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-graph-mode.json": {
+        "bytes": 4135,
+        "sha256": "bb82ab3d628b91be957e557545810ef5c20790be7d43296096f20a21f45ec391"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-model-adaptation.json": {
+        "bytes": 4137,
+        "sha256": "7ea1ab0f0044afff935a91071b36ac09bc46b5ef044fe2a68ab0608256ddd853"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-profiling.json": {
+        "bytes": 4186,
+        "sha256": "0d27058fd8d0ef8b524b7a60d1cb76950622746c94b5a03915e64e6edb4ab3d4"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-tuning.json": {
+        "bytes": 4439,
+        "sha256": "fcb49d71f59d8b39c1eaa03d24e95e2cbfe5f9d7f6e0f22df03ee0d60f6a568a"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-optimization-casebook.json": {
+        "bytes": 4371,
+        "sha256": "05783630bfaf3c7ef99b848865f4bbab50d0b52e80ad04fcd028c624d7638e9b"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-performance-analysis.json": {
+        "bytes": 7200,
+        "sha256": "b6eea897c12d0c6ac9d172089535c82c25eb71c09d7a4972e64812cc4682b5c9"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascendc-tiling-design.json": {
+        "bytes": 4505,
+        "sha256": "e69b6e89dd342ba55ae54b5fac14aa1b91ee2a361da40cb669c287ce519f4517"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/cann-attribution-overlap.json": {
+        "bytes": 4915,
+        "sha256": "8ac01b60f5cd37fd594d6cb5119e4e6525199d017296fc8779dc68a576f5525c"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/torchnpu-device-memory-lifecycle.json": {
+        "bytes": 4349,
+        "sha256": "7a79d0eb2c50381d40086e6db72d4bdd2cec5d2261bfb00cf73a60701a94da2a"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/torchnpu-pinned-host-memory-lifecycle.json": {
+        "bytes": 4431,
+        "sha256": "8bb34dcfbebb48667a94434b036a1a408b970af9890b047a3de325f364ed5f7c"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/assess-serving-preemption.json": {
+        "bytes": 4198,
+        "sha256": "e3d2a1a72d0d980f31fc85deee27db26bc0daa0c709e8ae99e42deb814a4a7f4"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/assess-split-backward-schedule.json": {
+        "bytes": 3915,
+        "sha256": "a8561794786eb0a897d831f00c92b4fdd3b0492316c1b1d8ad13c1f8f75ac815"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-h3-runtime-residency.json": {
+        "bytes": 12169,
+        "sha256": "602dd625874bd4a89861d5bb25881cc63e4b5a7fdedc2867089f0952943f6eff"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-llama-runtime-state.json": {
+        "bytes": 5816,
+        "sha256": "64f6ab8a95172378350323cead08bc6c201c624b711e6242938a6a21a8e83ad7"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-sd-cpp-residency.json": {
+        "bytes": 8349,
+        "sha256": "95ed97b717255be3b817e5c69debd6d41197559dbf21771507d5da79134a72e1"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-sglang-diffusion-lifetimes.json": {
+        "bytes": 7276,
+        "sha256": "5e5889223878c6cf42a49c6f09c60e9d9add88cb7e7d0df050c4e75d88a05aa6"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-vllm-kv-lifetimes.json": {
+        "bytes": 4275,
+        "sha256": "00424005a5eeb8b15f239edd2cb2c71047787e3ecf55924860cb7a78d137b586"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/bound-windowed-attention-replay.json": {
+        "bytes": 7877,
+        "sha256": "98b2291147a796ef59cf20a075ff662998ee6a529f85b49e9c6fb62fe41fad9f"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/budget-speculative-verification.json": {
+        "bytes": 7490,
+        "sha256": "66448f59460c9dffac46e74160388fab650cf8bbbd5bdc95c1465b70daf0c2dc"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/calibrate-serving-simulation.json": {
+        "bytes": 3992,
+        "sha256": "055b70e0b86f65156775b519e4b8cbf06a956a8993b06a66175167cedb78f2cb"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/cgroup-resource-controls.json": {
+        "bytes": 4801,
+        "sha256": "ad406b6dd5d9415bd28d448e64269a388b50e11f9eb7f5a9adb6b0a6246fd303"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/checkpoint-recovery-consistency.json": {
+        "bytes": 4740,
+        "sha256": "ccb85314a831bde58f8cc515580d001b174c11dca0bbcbc6221303acf629046d"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/choose-parallel-training-plan.json": {
+        "bytes": 3901,
+        "sha256": "40afcb41c59988fc1af8fdd00d2340800c54e96847dd0de16ecddf7ac221d1b4"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/cite-check-downstream-pr.json": {
+        "bytes": 2533,
+        "sha256": "dabff14684d5ae9a2b05a9743ff5b76d00a90fdb29319588aba67f5087aad706"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/claim-contract.json": {
+        "bytes": 2469,
+        "sha256": "78f15f6533434fc29adf1df651f7a15eb8ba81733f8b02c710d90102541d707b"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/comfy-ui.json": {
+        "bytes": 6919,
+        "sha256": "1496807d958df055f720ca93e6dcf7f7e8f506f3f12c94c9a2166f789b00058d"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/competitor-scan.json": {
+        "bytes": 2449,
+        "sha256": "c2e97e49e5185710cdfaa08ac1c2b1c19b958e68ebfca885d5aee0ed8c9f8c8b"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dcp-async-save-performance.json": {
+        "bytes": 4538,
+        "sha256": "359a5acf8304a78fc074688676f9a0be62f56f463d0960a544e409f49843bdaf"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dense-performance-evidence.json": {
+        "bytes": 4227,
+        "sha256": "d87f93a9dc9ddc583e3929baaac85f7a425666dd6dd71665933f18344ccec834"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/diagnose-vllm-admission-pressure.json": {
+        "bytes": 9789,
+        "sha256": "a1de40def775b2e840c27330b67e6f37c3c244db1731fab2bd3e8a787143210e"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/diffusers-invocation-state.json": {
+        "bytes": 6109,
+        "sha256": "914d2ffe9776ba931bed780a2a270c1525e3c6cbaf0735b8531064d3180f3233"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/draft-acceptance-issue.json": {
+        "bytes": 2549,
+        "sha256": "6aaa63dbfe6cda982ccd3a4e368527b80fd2873672c000c66d06d28e40a2bf14"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/evaluate-elastic-serving-placement.json": {
+        "bytes": 4334,
+        "sha256": "fab1fa814e789ec2d86587fcf7de7c38174464b9e9ba0cf53ef3fb470053e7c4"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/generation-cache-lifecycle.json": {
+        "bytes": 5972,
+        "sha256": "e5679cdecb722bbd17280913913da8473052ad09e9e7a13b3a9b95463a139ef6"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/host-memory-management.json": {
+        "bytes": 6530,
+        "sha256": "a36aea21169281866b2468860d3ffeb912e61edd571308b02f82939f17f694a8"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/host-overhead-optimization.json": {
+        "bytes": 4243,
+        "sha256": "05fd7f054966b96892980d585352bc1c3fcd47f439cb34268fdf07feee58a998"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/inference-benchmark-design.json": {
+        "bytes": 10735,
+        "sha256": "091bb086c6f0123dbb5db78968c13f846cad135e8969982efe557644b7c7613b"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/inference-optimization-casebook.json": {
+        "bytes": 6531,
+        "sha256": "c512025d7a59934c91665e24752d9393ca9b6a9dfb624c72789294defae93c0e"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/kernel-writing-guide.json": {
+        "bytes": 4545,
+        "sha256": "f1f0ca1d11d3550c9b8c5fdbed8f6cd09b877ac14341f973d58dca02b5d8995d"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/kv-tier-management.json": {
+        "bytes": 4609,
+        "sha256": "6141f61fb6b0d147b704ab7a2627e9babdb53a90e15b659152a3d565967afddf"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/lazy-evaluation-stream-lifetimes.json": {
+        "bytes": 5934,
+        "sha256": "34f47009a8a33f8907bd78f110b95fcd62409bef1b38e1055acb46b2eab285ff"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/npu-skill-factory.json": {
+        "bytes": 11868,
+        "sha256": "24c1d3bff288e86b99ea674e3252ec6f54d9ba944274aef6a7666197afd6788a"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/optimization-search-policy.json": {
+        "bytes": 7101,
+        "sha256": "d7bf71d5491765f62eec02fe5013dc1a5ef56e1516892da48593c6165a86366d"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/os-process-placement.json": {
+        "bytes": 6304,
+        "sha256": "e6e8fe5d7d66f3e98f05762d94441726dfa757ce7fbffd2b35402b3f327f8fb3"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/performance-ceiling-analysis.json": {
+        "bytes": 13695,
+        "sha256": "47c034d96026e1c2a02848d9c9c3caf4024a5033664bfd089ad30d9f23a80deb"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/plan-activation-residency.json": {
+        "bytes": 3871,
+        "sha256": "436e364020cc854fd52a10ee76e4593c7519bb0546c8aa5238fcfd89158e8ab0"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/repository-evolution-analysis.json": {
+        "bytes": 5865,
+        "sha256": "e0af402c3320ebd9bd36c774cb52c0708721a86ca446c05f6e94306888838e30"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/rsi-evidence.json": {
+        "bytes": 2475,
+        "sha256": "c8b4bb5a8c88009ac5b1526e82639715381c940d3c3eb295f0333cae6c2d3ca9"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/skill-evolution-playbook.json": {
+        "bytes": 4091,
+        "sha256": "0e063c2e7686de0c27c061e1aac0eb2d8c4fa83246e32ddeeb0e28126c176425"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/storage-io-measurement.json": {
+        "bytes": 4783,
+        "sha256": "80e589e3a0b67d3a045e03feeadc45e720ceb277e6124c917eec04863eb67419"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/training-input-pipeline.json": {
+        "bytes": 4296,
+        "sha256": "ae696bea8d2b7bfdaa29b6e5bc0bd861abcc0e42e7d9bec6ae20c6e7d6f27669"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/training-resource-budget.json": {
+        "bytes": 6944,
+        "sha256": "ea5e5e864e3aa34bd00be67c8e741877ae01480f039652073b65b71dfcd54c97"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/weight-loading-optimization.json": {
+        "bytes": 8139,
+        "sha256": "7f2d1fbc6b1cfd80c3d6ed8df107d4ca9b97655626ee078c345445a303b1b367"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/audit-afd-p2p-exchanges.json": {
+        "bytes": 6640,
+        "sha256": "9ac45aa784b430ad1ca9ffddecce8dbae9740cfe9c98c10b330e01d350015a38"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/diagnose-attention-bottlenecks.json": {
+        "bytes": 7602,
+        "sha256": "e525d75578de1cc164e6553d01efbd0a014e74646200580f59b1161c030413ad"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/host-overhead-profiling.json": {
+        "bytes": 3965,
+        "sha256": "e9e6ae8c5b9986e83cade3bd029baae112859e01aa6c96f741a264f3bee44adc"
+      },
+      "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/reason-hopper-kernels.json": {
+        "bytes": 15086,
+        "sha256": "758a649e651d8f9635ac2df45d18043e156c8d54b8f8dbc95f8e68ebcb866a28"
+      },
+      "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/parallelism-and-extension.json": {
+        "bytes": 3399,
+        "sha256": "1253f5f01bcf1498703c7f34960f0c50580df5ba8f140af557a903e296ad81d3"
+      },
+      "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/runtime-radix-hicache.json": {
+        "bytes": 3317,
+        "sha256": "5a0b2aed51649af214cdc1034f57650b7ce195c934a17452ee67a35131f80cdd"
+      },
+      "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/tuning-and-validation.json": {
+        "bytes": 3385,
+        "sha256": "ef4d4f7182da4433d59d72dda0d21b4aeafadd73260be11ab8d4cb36c20fafc6"
+      },
+      "src/inferrsi/data/repo_packs/sglang/pack.json": {
+        "bytes": 620,
+        "sha256": "bce2e0c42796a9ef499d459c37b56d1d56f4b19344841359c534fca29c0359ce"
+      },
+      "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/audio/serving-and-streaming-quality.json": {
+        "bytes": 3440,
+        "sha256": "97fc1d5a11a39a0634dae9e9602a4ec47a4d2348e121076c2d5c836cf0692114"
+      },
+      "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/omni/extension-lifecycle-validation.json": {
+        "bytes": 3482,
+        "sha256": "7da8dc5d2655099de2bcd5f0fa922d9e1bcb0bfd2f571f026a004a7f402a75b6"
+      },
+      "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/omni/pipeline-state-and-transport.json": {
+        "bytes": 3404,
+        "sha256": "0c57692772019504264e39bf0444a69e20cff86f6e112b072c53cf7a4299cc82"
+      },
+      "src/inferrsi/data/repo_packs/sglang_omni/pack.json": {
+        "bytes": 630,
+        "sha256": "8e70336bc2aac0c74732a026337813652b5d5c6e1fc60f5571e9442345077b96"
+      },
+      "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/architecture-search-space.json": {
+        "bytes": 3294,
+        "sha256": "2c5ad72d5bab4f7082fc3be1532c347444cde38f5a84639fe16b22134b68685e"
+      },
+      "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/controlled-execution-tuning.json": {
+        "bytes": 3255,
+        "sha256": "6a2bf611a432213919e1c237687507ceb0ab753cccad0e934dc77f86604805e9"
+      },
+      "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/engine-validation-contract.json": {
+        "bytes": 3295,
+        "sha256": "ef89775bc82965fe87a832c7ec8a5cf55cb9eba4ecfc6279caea9ce3b0410c41"
+      },
+      "src/inferrsi/data/repo_packs/strata/pack.json": {
+        "bytes": 617,
+        "sha256": "643b32defb8db9aa230f624856f4fb4b595b5c23db58d9453091f5420cff71c3"
+      },
+      "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/bounded-serving-measurement.json": {
+        "bytes": 3239,
+        "sha256": "fb00ddac14ef6b06f48da9ba335529bd55d5d83c2a2fe331f0aa1ca16afceb37"
+      },
+      "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/extension-correctness-gates.json": {
+        "bytes": 3343,
+        "sha256": "4ea7083d277df205ea4c361bee87a8207951596e9ef6cfa136cc7d2aeab02f46"
+      },
+      "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/runtime-and-kv-ownership.json": {
+        "bytes": 3280,
+        "sha256": "e6142c1065754716a5f26f42119722908dc8fff63e667c986d57f1227b8610bd"
+      },
+      "src/inferrsi/data/repo_packs/vllm/pack.json": {
+        "bytes": 508,
+        "sha256": "1705d96710249c4514c15d8d011b43c1a1ae7c15fbc9ddf11ef2d0e6b2ab7d07"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/minimax-h3-reference-kv-reuse.json": {
+        "bytes": 4047,
+        "sha256": "80b853fc594d387ab48bed04dca245e0a22b8cdaf7c9b853551ab3bafa302e84"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/minimax-h3-sp-head-buckets.json": {
+        "bytes": 3742,
+        "sha256": "c4d29d1bceb5e2cf86151bf53c9dafb3d83428fb1e9d2a848d852bb8824bf857"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/qwen-image-21-eager-t2i.json": {
+        "bytes": 5268,
+        "sha256": "ef107d43cc448219eda4d64ab961b61cda6350e20d30d94de1ff7c4ffff7e775"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/audio/streaming-serving-measurement.json": {
+        "bytes": 3389,
+        "sha256": "2680f32f28907f6abc7a7e1e3ecab4f46ee412f40f4898ed2f371ffadf47ae4d"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/measurement-and-profiling.json": {
+        "bytes": 4472,
+        "sha256": "c8b25583cd160024c7472a20debea34cb6d4bc79714b4bad2953555adc9f794e"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/production-integration-gates.json": {
+        "bytes": 6112,
+        "sha256": "649b49ff64d3e9e88bd620027dabf9d126ca69bbe91b448f28e845ecbe1f8e2f"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/step-state-and-output-validation.json": {
+        "bytes": 5090,
+        "sha256": "e1ffe28c4fa54af739cdef2fedf4755bba18e078f8380d4fde8138ea712e5778"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/upstream-review-gates.json": {
+        "bytes": 7614,
+        "sha256": "5127709bb85ce31008d19468e1b4fcb5454a71776f69f991841cd9ef7558f9c4"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/omni/stage-ownership-and-search.json": {
+        "bytes": 3427,
+        "sha256": "e75d6c2829ae22881893f5da2ad7edba099fd64ccfb0fc93d30473e06e9d4d80"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/audio/concurrency-scaling-state-and-overlap.json": {
+        "bytes": 25324,
+        "sha256": "cb47fabb232db81ed525cbb54371becc7e4a28b2bae9d8492f60dfb6f31563e8"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-adaln-schedule-cache.json": {
+        "bytes": 4253,
+        "sha256": "ddbba3bb72b5789938595a82d3280e861f75f6ed54f05a1d2b7dcb1089d30f63"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-cuda-mxfp8.json": {
+        "bytes": 4468,
+        "sha256": "e4c538f03f4ab90a4acc8e3d8721bbd226aec1ee9bde45757a7c27d3d970821d"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-fp32-modulation.json": {
+        "bytes": 3625,
+        "sha256": "abb8d94a1b025f0e2c8e085275e96ee85d68aeb27cb12d0918e4647f8255b693"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-qk-norm-rope.json": {
+        "bytes": 3706,
+        "sha256": "4c1c4acfb9cc0baa90bc28fed1903020ea1ea38755b84315711795db45e48bb6"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-ref2va-teacache.json": {
+        "bytes": 3835,
+        "sha256": "8aeb3f99ac4227cf8b3a5cece13d938b19082e265a557eddae27b7e495bcc22e"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-vae-stacked-tiling.json": {
+        "bytes": 3508,
+        "sha256": "8cc18eacbbfd8fc4d3a6b4152c4b584d559fc9b15377da56ef072e542d807c93"
+      },
+      "src/inferrsi/data/repo_packs/vllm_omni/pack.json": {
+        "bytes": 566,
+        "sha256": "58c971e9dcf6ac15752629ed4d9426aeb5b5d857fc37b5bcf921aa2e81c0cdd5"
+      },
+      "src/inferrsi/data/repo_packs/vllm_rlt/pack.json": {
+        "bytes": 649,
+        "sha256": "6097041b8d026740060e527f49b1b4e288315b1c3227f7583805b86d8fa415f7"
+      },
+      "src/inferrsi/data/skill_library/README.md": {
+        "bytes": 18819,
+        "sha256": "ae583f3578425df1dc91ba32b8b6543c0469c9a817ebea873a77051d6b13e1ce"
+      },
+      "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/SKILL.md": {
+        "bytes": 5494,
+        "sha256": "7581b56336d07213a87d01477793ae1a8acf0b7cf876fce36c9a761ee32c1ff4"
+      },
+      "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/metadata.json": {
+        "bytes": 1429,
+        "sha256": "68ac7f64b7ab947fc2db664bc3990f102c1cc5539b42350aa5818d0bbf17d86c"
+      },
+      "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/ownership.md": {
+        "bytes": 2081,
+        "sha256": "40688cc44183027efbf2fca8d3548c299cdf2e33fef54f76ab3549353eb9bd1b"
+      },
+      "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/source-map.md": {
+        "bytes": 4635,
+        "sha256": "10562a75683bd9bd07ec2e37a1cbd0f93f2534d6f85b55a566adb2d97485215f"
+      },
+      "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/validation.md": {
+        "bytes": 1477,
+        "sha256": "5a0b854b87d091d7d505b0bdd5c03330052507a851ab6edf35b93a6675f2d4f7"
+      },
+      "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/SKILL.md": {
+        "bytes": 4010,
+        "sha256": "a95569bcc0da6e715fd74783a6a1e0489e92b680c582f15dafbf98df589d505a"
+      },
+      "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/metadata.json": {
+        "bytes": 1519,
+        "sha256": "9c2293c8956508de987c886c7224d673d22912e7e915b026b05abdd287e97401"
+      },
+      "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/execution-lifetime.md": {
+        "bytes": 2147,
+        "sha256": "c4bc41e08aca68f661325f61e6f0db1760a079dc362f295c847bdc374fee3bf6"
+      },
+      "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/source-map.md": {
+        "bytes": 5918,
+        "sha256": "798b9cb89ed6b885eb47464ca9f137d2802a54571b9327e631d7742d3e5abd5d"
+      },
+      "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/validation.md": {
+        "bytes": 1608,
+        "sha256": "ebfc00bf7d6ab0a1dc415953ae84d56ccdc713d5614aff85e433c4a843793bff"
+      },
+      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md": {
+        "bytes": 7322,
+        "sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845"
+      },
+      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json": {
+        "bytes": 587,
+        "sha256": "db7ad368394ccbfdeff6a16e3423c300d51305b1f95ff5c3936df91738a8cc35"
+      },
+      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md": {
+        "bytes": 5314,
+        "sha256": "22afe4f4d77f441096cc657d55d763473bfcbea8fb531c8e493fe49ab84955f7"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/SKILL.md": {
+        "bytes": 6745,
+        "sha256": "d494fc7e539cbb7388189bd7b3f04bb5c4983fdd7a82484348aad1cee7f741f2"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/metadata.json": {
+        "bytes": 558,
+        "sha256": "71c68431663ed2336cccc985a354adab6b13b4cfd6123a18a5d8d53c28268509"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/references/source-notes.md": {
+        "bytes": 13004,
+        "sha256": "6ffbfb68cc7e04923b4fbca2a73a7555b7c3b45a6f40819bb5f3213ef05cf88d"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/SKILL.md": {
+        "bytes": 5977,
+        "sha256": "03829039860ece35519f2e4c339a48b5014ec25e07a2c150a3efb6a3a4e35b4a"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/metadata.json": {
+        "bytes": 559,
+        "sha256": "80d80c15514919826fa72d6f6cf26b76db8accd85d21187fc239232ca454e7bf"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/SKILL.md": {
+        "bytes": 14703,
+        "sha256": "4ce9dcadd1ae3c7a8cc633336760d0a3c382f354f1b7028e3a10d28fa0d0c9fa"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/metadata.json": {
+        "bytes": 1403,
+        "sha256": "b6cb25879b1bca5f5af08b48776f2e664cf5dcd11a0efb1c4f56879e2afa9006"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/references/awesome-ascend-communication-tests.md": {
+        "bytes": 9388,
+        "sha256": "0cb27a3d24409171bb414d7ebf0c49ad62be5f46482f29624695496ba9eb3e63"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/SKILL.md": {
+        "bytes": 5839,
+        "sha256": "9203cc5a14a1d65df5fbf275734b93be023e06527459b2f79864f28a338cb44a"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/metadata.json": {
+        "bytes": 536,
+        "sha256": "72b41e7845915d34b09219624e82976386f3125816293540c2c38143fe6b477f"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/SKILL.md": {
+        "bytes": 6440,
+        "sha256": "8fdf46fb2743b67a66bc069d22a45305a222debfb606777bc451b4b8f31d9cc6"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/metadata.json": {
+        "bytes": 536,
+        "sha256": "be681a3206e5e183b4c9670c9d7784bcffce18c891449e5b15f94ca5f1324b9a"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/SKILL.md": {
+        "bytes": 12988,
+        "sha256": "401981cf7828edcab0c9ea882f7ecd83badfa3c2a95006e912cf99d67b3d6a12"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/metadata.json": {
+        "bytes": 760,
+        "sha256": "efd37bb23acb500dad4ec8ebe7ddfe5c3f196374052825f923ac85da8445f06b"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/SKILL.md": {
+        "bytes": 12236,
+        "sha256": "38b151bc6b4e9734c6fa41efb7a30343559400c0c75bbdd26bd8a968b8ea4592"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/metadata.json": {
+        "bytes": 557,
+        "sha256": "04d2002fc8a0a7becf86b0b73daaf2da8716ab466d7f988da43d41a10f26680e"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/SKILL.md": {
+        "bytes": 9172,
+        "sha256": "647f5e9a288f2c4e1617ba53df59d9d827df749eb4fe86182d21a9c41175caf8"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/metadata.json": {
+        "bytes": 536,
+        "sha256": "f425f7c1abbd616387273e8932c6bd02f4b1da012f099286ad4171e0f6037b22"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/SKILL.md": {
+        "bytes": 9880,
+        "sha256": "a057b7f86a24f34a8f84abbbb4910d58d4e50ff47060e1c404600017e3febe89"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/metadata.json": {
+        "bytes": 2297,
+        "sha256": "9e705625c8308013c287576b0247da6f6116d817a4a535cbecff80abd6ea35ee"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-CANN-2.0-LICENSE.md": {
+        "bytes": 5262,
+        "sha256": "89b1c114df62cfd183004f38d18d07f186436c3b37203dd69c3e84becf9d2282"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-stage-evidence.md": {
+        "bytes": 11789,
+        "sha256": "f56e4125325c9c1b9235a1d64fd732861c7686444492422daff44ba8dd648be4"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md": {
+        "bytes": 11860,
+        "sha256": "93aa1b775ced304cebfe6c4dec0503ddb2260ecb3cb01edc17ce329446c5ce2b"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/metadata.json": {
+        "bytes": 552,
+        "sha256": "f6d94f104cb70969024b2f4542078a240e200712c66279eca8b94f60bee1dd71"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md": {
+        "bytes": 8972,
+        "sha256": "0e6f094dc554abea12e07def50f20b233b47127779a8aefad595b4ab1efd24c2"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/metadata.json": {
+        "bytes": 569,
+        "sha256": "acccc8eb86de0df6caf78a9b3c6db76a2a686adc4a5ba0029a4d62d50ff9a8f8"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md": {
+        "bytes": 5989,
+        "sha256": "8428df7157af024fa17cfe0b3818ed54bd9f468306dda46eb8d4f0edf4f3f230"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/metadata.json": {
+        "bytes": 579,
+        "sha256": "c650dce88c07b636d68e063dbda4d319f6f7f3ecbcdd5e47f393cbfc3af7c71d"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/references/source-notes.md": {
+        "bytes": 11470,
+        "sha256": "80cba4d3a884e7ee2618326cf74b9f9c3a34e4d9cf443707f6bbe673f2d5ba65"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/SKILL.md": {
+        "bytes": 6159,
+        "sha256": "23e93ea7c98f99743509fdc04139d78ed2e7123c63e80b366c6beaba7eb70d0d"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/metadata.json": {
+        "bytes": 581,
+        "sha256": "97836a38465106d816a9ace894a7b277d6d330f9accade8f8493ce6f0fb57ef9"
+      },
+      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/references/source-notes.md": {
+        "bytes": 13493,
+        "sha256": "435193590b4a2ba32e2e4c6fa9d7e8f889528d412cbbb48e35ec5433f7dab384"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/SKILL.md": {
+        "bytes": 7137,
+        "sha256": "a25738dbb3b9daa15c5827cb536a20fe13f7c8b9dd14aa3fb7cfa7e132879431"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/metadata.json": {
+        "bytes": 1007,
+        "sha256": "3abe626a6e4c73c943429f609ce2c62aa77aae7e1726475ebbbdfbe1794f6f7d"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/mechanisms.md": {
+        "bytes": 17376,
+        "sha256": "fe08b2292fa587be3874774ddf41063b84c579a0085df422ec6710d8aa90b8b7"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/INTEGRATION_NOTES.md": {
+        "bytes": 4024,
+        "sha256": "371b01ee201ef51fecbfd6da892d517aaba295930b9395025f03f34e21c030a1"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/NEXT_GATE_LEADS.md": {
+        "bytes": 1623,
+        "sha256": "35b8b58ae5db02b2df51ccbe5cdf870a7383d17f345f9ee5eb7b0a4146dfc74b"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/README.md": {
+        "bytes": 2888,
+        "sha256": "e0b1ec1b39c10067f5ddce58320851ad089568c4b2296a24777389097cb5b388"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/io-reduction-can-plateau.md": {
+        "bytes": 2671,
+        "sha256": "5cc35f514224ac6e6efd2da6db221bba468b2f4db66807295e67b240bcdfdba5"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/overlap-and-fp8-counter-signals.md": {
+        "bytes": 3692,
+        "sha256": "80c8a60a6df55f3d84fab23e03567d8b38e23bcdbd63cc09881e46db1e2888ab"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/partitioning-after-fusion.md": {
+        "bytes": 3132,
+        "sha256": "4d4cabe0991eab554dcfbd4fa9c8199e61fa4767709b0c4cd646e2588bb398a2"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/knowledge/attention-cost-model-boundaries.md": {
+        "bytes": 5223,
+        "sha256": "5ed8f888443e38ccb72545b6025447aab7b92936f78c151fed680e582126239b"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/sources.md": {
+        "bytes": 9069,
+        "sha256": "53436e0de4e6dda2d29426591bf61f8526378cdc6f9ec7967b88b1bcd4c61e6f"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/host-overhead-profiling/SKILL.md": {
+        "bytes": 11199,
+        "sha256": "326e566fbe30d3ba04e4b50c3e87aad9df037f0ac4ad6fe129165ae97dc5f91d"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/host-overhead-profiling/metadata.json": {
+        "bytes": 527,
+        "sha256": "76e34c1c64aba473a9e126b032a776417caa0f913f6ac3cf4dba3a5d123b1ee0"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/SKILL.md": {
+        "bytes": 4905,
+        "sha256": "cec4e3aa1eb64c31c54b5f5ca40af2b61b7919d94be4174968581d5f752c0538"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/metadata.json": {
+        "bytes": 7453,
+        "sha256": "7520da1cc97412631a600957ecb18aa5a7a4461ea21e76c237330180193ed898"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/fp8.md": {
+        "bytes": 6287,
+        "sha256": "31d9c211c11c2224007abbdcd4059913c4a54833fe2d2f926411c5238fcd3501"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/cutlass-bsd.md": {
+        "bytes": 1547,
+        "sha256": "42fec630f410aa308f70a51a89fadcd19586fa620f9831a32bee528a9a10000e"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/deepgemm-mit.md": {
+        "bytes": 1065,
+        "sha256": "6c73a30c14a2938abaa6afe98ebb19e6d81dac5bd4ace0cb05c61c20afa37b85"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/kernelwiki-kda-mit.md": {
+        "bytes": 1063,
+        "sha256": "84aab57158a914f448b38d8be29dd842b634a08c192d844f416a256beed4ae94"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/notices.md": {
+        "bytes": 3240,
+        "sha256": "876ec016644bf5331b3f257985e126c0e61980e3bfd82569b7b0c3f6b2c6c64c"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/resources-and-scheduling.md": {
+        "bytes": 8919,
+        "sha256": "1ce25b426736072435a1ead9a2a8e1f56df5d0abb8198e8c4fdcb34321db4964"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/source-inventory.json": {
+        "bytes": 120506,
+        "sha256": "1dc09c895c970a0c9708158bcc6dd65cdb736274fd2c35d44542c70407d3849d"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/sources.md": {
+        "bytes": 9004,
+        "sha256": "6b74b9e774f2a80db0860510d9e6940d36284430f29c1e124f115a9eaf25833d"
+      },
+      "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/synchronization.md": {
+        "bytes": 6843,
+        "sha256": "4322f6b3c0e4368bf8c192b4ec752529767f50d19c0c38dad8fdf9cd781e41d9"
+      },
+      "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/SKILL.md": {
+        "bytes": 7348,
+        "sha256": "fd1f5a7e6a55714730044155f5b38c1d700383d4e16c3e91963bbfbe1c773af5"
+      },
+      "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/metadata.json": {
+        "bytes": 572,
+        "sha256": "19e40763fedc20990651bb989c32a7693306c5c54af46c3d4be77fd4f87b6c03"
+      },
+      "src/inferrsi/data/skill_library/hardware/host/host-memory-management/SKILL.md": {
+        "bytes": 23213,
+        "sha256": "2c807f3df2da2f533d25183de5771f3d412058d16410994292ef8f48c49f9b3a"
+      },
+      "src/inferrsi/data/skill_library/hardware/host/host-memory-management/metadata.json": {
+        "bytes": 1684,
+        "sha256": "9ac6af135f0b6d76a76bc201b6bbd8ac9afe5201dd1e2ea82c0ac4826de401da"
+      },
+      "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-LICENSE.md": {
+        "bytes": 11357,
+        "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4"
+      },
+      "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-native-offload-readset.json": {
+        "bytes": 6267,
+        "sha256": "cc6c2011be746002ba3ff96ccefab1809863748ddd17eb102b75e7b0e134fe55"
+      },
+      "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-native-offload.md": {
+        "bytes": 8966,
+        "sha256": "6c2a72b5ba3708a01d229e0576570837a617e6a42dbb0b5312a26a65a6c31542"
+      },
+      "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/SKILL.md": {
+        "bytes": 16072,
+        "sha256": "36ef88826a32ac3147b869001004ded17a8d5a491e25bfe105005ba39ea9d1eb"
+      },
+      "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/metadata.json": {
+        "bytes": 539,
+        "sha256": "ae71c389dfdb1a19d267feecb4c6f1fbb4c9afad501ff6d8fab7e38b29bcc37a"
+      },
+      "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/references/deepinfer-zero-overhead-scheduling.md": {
+        "bytes": 10894,
+        "sha256": "f452bf40132c9195d739f1c86b3aa18857138c7ed0f1861af5853d93b9955924"
+      },
+      "src/inferrsi/data/skill_library/hardware/host/os-process-placement/SKILL.md": {
+        "bytes": 10327,
+        "sha256": "2003492e4d89fde0443b180d74933e764bd1979435e109c8d26e20eb8f6e5053"
+      },
+      "src/inferrsi/data/skill_library/hardware/host/os-process-placement/metadata.json": {
+        "bytes": 1399,
+        "sha256": "ec02995c075f0a540921f57b4846f2cafe334eb90f44389f17c3c50457c3e9c5"
+      },
+      "src/inferrsi/data/skill_library/hardware/host/os-process-placement/references/llama-numa-thread-readset.md": {
+        "bytes": 6617,
+        "sha256": "748b254b3b043a5ddb0f480eb4a566854d27eedd7b111499863e9bf0a5a6951e"
+      },
+      "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md": {
+        "bytes": 8813,
+        "sha256": "63b883c10a8d3e7d918abfc7175b72e2b401ff1e02697f94b8382ec1c8f369b7"
+      },
+      "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/metadata.json": {
+        "bytes": 558,
+        "sha256": "0a41c7b00978332b54ee1e4a7c1ade88ecde3e961128f24d1b4f7708caff5ff5"
+      },
+      "src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md": {
+        "bytes": 6402,
+        "sha256": "0a5f48d9352c5bf460775ef47f9136829978811f4a16e84983e299782cae305d"
+      },
+      "src/inferrsi/data/skill_library/methods/assess-serving-preemption/metadata.json": {
+        "bytes": 537,
+        "sha256": "77ce5ba16245cbd80d46d0abdf80aaf26fc0924fa2d2a5624ff430de68a1daf0"
+      },
+      "src/inferrsi/data/skill_library/methods/assess-serving-preemption/references/source-notes.md": {
+        "bytes": 30405,
+        "sha256": "f29fcecfba0490008a354c11b42403983fe165e56ff907ad61cc6040d2b6407f"
+      },
+      "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/SKILL.md": {
+        "bytes": 3181,
+        "sha256": "1e64f28ea71672fb53c56f6108b87864738c6d03d6dcd32eedaef9e0625e96e3"
+      },
+      "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/metadata.json": {
+        "bytes": 478,
+        "sha256": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7"
+      },
+      "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/references/source-notes.md": {
+        "bytes": 10548,
+        "sha256": "61b1609f71c37b61d136b8c4d39f793de87cffb1cf31c8ec737a1c3fa6fb1b4e"
+      },
+      "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/SKILL.md": {
+        "bytes": 3884,
+        "sha256": "ac773c6f1c110a664dfca28a3ae9129b10c07614e60ca3640dea776220e7b352"
+      },
+      "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/metadata.json": {
+        "bytes": 648,
+        "sha256": "261ce9f644ea6eebb90c830f41fb8b320d2b401456e5bd8d494095a44b8620d0"
+      },
+      "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/references/source-notes.md": {
+        "bytes": 8415,
+        "sha256": "9696d3c0ef96b6611787446bd6f96d4d6255ca8d107105de8c1d9af83edb9fa5"
+      },
+      "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/SKILL.md": {
+        "bytes": 6416,
+        "sha256": "6b7db1eb2780dec27858cb7aa0a3c093c5b5dfa4c30261aebfeaa255213ff688"
+      },
+      "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/metadata.json": {
+        "bytes": 582,
+        "sha256": "4dc8659806539b4784ea9348a01d4194ac70d0bb297a916f90ff6419efa5c328"
+      },
+      "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/references/dlrover-persistence-replicas.md": {
+        "bytes": 8398,
+        "sha256": "055f4357f79026fcbcc766bbe77db610376ec353623f8898f5d14a30ee69e7bc"
+      },
+      "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/references/sources.md": {
+        "bytes": 7984,
+        "sha256": "9425b1aab963a2c8d35f3db897bd15ee69699bf5b0d0c8425c3c8bba3e9a0233"
+      },
+      "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/SKILL.md": {
+        "bytes": 3068,
+        "sha256": "25d04aed9782b2a8ff1186828becdc7fce93c4418039ceaef9b8dcc5ec9bcd24"
+      },
+      "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/metadata.json": {
+        "bytes": 478,
+        "sha256": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7"
+      },
+      "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/references/source-notes.md": {
+        "bytes": 8521,
+        "sha256": "9f9184d46b19312138a6e805e439e43dc0594a0cc6b46675255a582dc946ca5b"
+      },
+      "src/inferrsi/data/skill_library/methods/dense-performance-evidence/SKILL.md": {
+        "bytes": 8305,
+        "sha256": "53cc1867a13f54d1006d50517dea8848869ca50df3f0cd610b2424cc8d18c933"
+      },
+      "src/inferrsi/data/skill_library/methods/dense-performance-evidence/metadata.json": {
+        "bytes": 707,
+        "sha256": "a0d1e3b17140295510a764045d1b222e19be3a3406e6d1e7dddaad49ff474490"
+      },
+      "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/SKILL.md": {
+        "bytes": 6667,
+        "sha256": "82d0099d3c33706f46c657d4fa7f76d3d486e019b36760d1de001921328a60cc"
+      },
+      "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/metadata.json": {
+        "bytes": 547,
+        "sha256": "37379a4d4a058ca46da2b20a920dc01be9ec11e557c32b2332702ec76027f4b5"
+      },
+      "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/references/source-notes.md": {
+        "bytes": 44796,
+        "sha256": "ea416831fd7d0874e06d0a76fb0c7196df53072b717673a0cdf635db6bfd917b"
+      },
+      "src/inferrsi/data/skill_library/methods/inference-benchmark-design/SKILL.md": {
+        "bytes": 8974,
+        "sha256": "19ebc1f567aadf49f7a534a18fdb629c72c0e4ce043d360983136491cb72bfab"
+      },
+      "src/inferrsi/data/skill_library/methods/inference-benchmark-design/metadata.json": {
+        "bytes": 5923,
+        "sha256": "18310402a8ade03fd48ac687f79ba9d447871e90337de84c0e1bc15c0b827306"
+      },
+      "src/inferrsi/data/skill_library/methods/inference-benchmark-design/references/benchmark-commands.md": {
+        "bytes": 8466,
+        "sha256": "0898424e0bff3a6246e1eeeda6abd6c5bd974c047fe32f3bbd5ab9d31ce17cd3"
+      },
+      "src/inferrsi/data/skill_library/methods/inference-benchmark-design/references/task-benchmark-selection.md": {
+        "bytes": 25086,
+        "sha256": "0d4bb080a754e26846e81680023db0dff5e6a4c96be4327c411c3b72ac4ef8e2"
+      },
+      "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/SKILL.md": {
+        "bytes": 6644,
+        "sha256": "89a6a82c692f7f5192f4d7f4c3a5ff08a4df3b1c4bdaa0aafb7e04292d8aa5b4"
+      },
+      "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/admission-drain-occupancy.json": {
+        "bytes": 1354,
+        "sha256": "3319e23bbe2bd40e2d45fb89dc87bdc92d4df2b5ddd8086fcf84813b43885b7c"
+      },
+      "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/admission-drain-occupancy.md": {
+        "bytes": 3134,
+        "sha256": "e88bbac5d253a8e7e3a3d5e21047f23f2970c11ac23887d4303ab94fca52dd27"
+      },
+      "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/drama3-rejected-directions.json": {
+        "bytes": 1459,
+        "sha256": "2cc0a526fe0509551607a03e7626dd3d7844e643c4afb194837f75cddead8b12"
+      },
+      "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/drama3-rejected-directions.md": {
+        "bytes": 3557,
+        "sha256": "6f143bc363a3290fc008c08e57cb075f1776d1c8cca8a6a59297d5b1fbbf61bf"
+      },
+      "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-ascend-rejected-directions.json": {
+        "bytes": 1293,
+        "sha256": "b3d538f79c56724ccb2895cbf50cf2e6b0fec3979be62a684a8230cd932dff55"
+      },
+      "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-ascend-rejected-directions.md": {
+        "bytes": 2564,
+        "sha256": "be396983812262515899a3e2ba4f01b23ff92e1b836881882c7580372b8ae552"
+      },
+      "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-b300-rejected-directions.json": {
+        "bytes": 1196,
+        "sha256": "146ea1564a46ed03092c7dc5796b318fc05b04b1821105ee00eb7fa7fa1333e2"
+      },
+      "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-b300-rejected-directions.md": {
+        "bytes": 2656,
+        "sha256": "cce2930f544a0f78f21797fba4d8ded298e4a81f76c34d7977b0e31d361bc738"
+      },
+      "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/metadata.json": {
+        "bytes": 535,
+        "sha256": "fb55067fb45970319787068a3206110641cceaf0ccbabd706be6d0b3d2afc2ed"
+      },
+      "src/inferrsi/data/skill_library/methods/kernel-writing-guide/SKILL.md": {
+        "bytes": 10029,
+        "sha256": "01fb1712d7eb988c5e412f02340d2e48b4f8564fcf4ab232dc5897869d0d50a0"
+      },
+      "src/inferrsi/data/skill_library/methods/kernel-writing-guide/metadata.json": {
+        "bytes": 653,
+        "sha256": "0189d184337662056fc6952c7beffc1bf6154c4fa173ae3e3e463bd22d70f36c"
+      },
+      "src/inferrsi/data/skill_library/methods/kernel-writing-guide/references/akg-affine-tail-guards.md": {
+        "bytes": 11941,
+        "sha256": "113deb2db10b171db3425bb3f1792bee763706b794c8cc2e222b4cc0d0114118"
+      },
+      "src/inferrsi/data/skill_library/methods/optimization-search-policy/SKILL.md": {
+        "bytes": 12914,
+        "sha256": "882e7850b9b49ad692a2d11618701afe0f4e83727ea78df114be2029923c725a"
+      },
+      "src/inferrsi/data/skill_library/methods/optimization-search-policy/metadata.json": {
+        "bytes": 1403,
+        "sha256": "9f1fd8a2af52bc964021e21f16ecbdb891e08f2057370e700d67f86605d333db"
+      },
+      "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/INTEGRATION_NOTES.md": {
+        "bytes": 3597,
+        "sha256": "157e74ba180426b279b548a729769f152bf18f9a78ffca4ebfcaf26b7d0c5292"
+      },
+      "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/MECHANISMS.md": {
+        "bytes": 5875,
+        "sha256": "18bb2f5d7e52773c516420a142edfb531d0b9c822a2934d60f3234ac1cb9d40e"
+      },
+      "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/NEXT_GATE_LEADS.md": {
+        "bytes": 941,
+        "sha256": "16a264c7159f9c2c700ce8ba46e63f3e636b277c987da4970f6d2fa085fcb053"
+      },
+      "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/knowledge/adaptation-interfaces.md": {
+        "bytes": 3893,
+        "sha256": "5e761957cb5ecd93926e6ffc13deee7c830c6e8f0ee3c3846e9dd3b50dc74446"
+      },
+      "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/references/source-notes.md": {
+        "bytes": 13370,
+        "sha256": "54d83c3713045d59a48821bcf93598374ba4a630e3891ad39498f919317c4fb8"
+      },
+      "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/updates/optimization-search-policy.md": {
+        "bytes": 4238,
+        "sha256": "d3db44fa9a6831b86a02c638b323ed943e69fe256866213f10d68c72d762879e"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/SKILL.md": {
+        "bytes": 15324,
+        "sha256": "c5f78961c23e3c8d659f5675b522518ab06e1f2a8ea8bc74c5d428ffb41693d5"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/metadata.json": {
+        "bytes": 3669,
+        "sha256": "60b5b24ead590b0fb222398fba47822bf6723690da683e256a450bb5c0eeb806"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/INTEGRATION_NOTES.md": {
+        "bytes": 7349,
+        "sha256": "083e462e0a204dce599362d46712fcfb392d053c5cda56216e22e28d4b8fa0fa"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/MECHANISMS.md": {
+        "bytes": 4003,
+        "sha256": "742c249082ceca230326f4407a40130fe624553b796ae5a157a8a4c15a7e1398"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/README.md": {
+        "bytes": 1204,
+        "sha256": "17b2d1c7c14124d59ac8a0d5ad4200ad2803d991e784a1a98cd97188cc364999"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/READSET.md": {
+        "bytes": 3937,
+        "sha256": "66459ce393858b2b2160f1e2973373bfc4b9902a12644631934b4ebe3140296a"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/STATIC_CHECKS.md": {
+        "bytes": 1111,
+        "sha256": "cb0aa7d148352ebb7b1678fc42a557a099575054716a341dfa25b6a998029064"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/cases/decision-cases.md": {
+        "bytes": 5491,
+        "sha256": "781a675a6533e0a18c044aaaee0d010da7013b6504654b8b8292a35176db801e"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/govern-optimization-experiments/procedure.md": {
+        "bytes": 3823,
+        "sha256": "ce036a4d8a0908da204515c47e2ff9623548d09de264fa65bd58fedce52a78a0"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/govern-optimization-experiments/references/source-notes.md": {
+        "bytes": 21659,
+        "sha256": "f9a8254e45473ddb2b73e576957b6d3db83c62b99dff1922f4d1e606e569559e"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/knowledge/bounds-predictions-and-decisions.md": {
+        "bytes": 5297,
+        "sha256": "5780d6eb5257612ec45deb94f85f334a6ecabf7e466d1678706041f6549bda79"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/redirect-cross-layer-optimization/procedure.md": {
+        "bytes": 3648,
+        "sha256": "6593a63fce9c474b515663a560c94b7df048dc5f6e77a95b8e54e4a64cd10a09"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/redirect-cross-layer-optimization/references/source-notes.md": {
+        "bytes": 26469,
+        "sha256": "6ff3a7bc31f1172db92aedaf7df1f1aac688bf769227b583a4cd82bfad52da67"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/agentic-kernel.md": {
+        "bytes": 41353,
+        "sha256": "7190573b85f94a4391bb0d887ccbd37a31b3f08bb0531114ad7afd72fe1567a7"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/flashvector.md": {
+        "bytes": 9830,
+        "sha256": "3926369acd88ff3c370bee5718e74dd15f3219c3e4f6311095e30e917bffa4cf"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/inferencebench.md": {
+        "bytes": 9294,
+        "sha256": "21f47111197a0b7f8d6384ea918d139dc8318a62ca3313ce060048271f50b639"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/prior-coverage.md": {
+        "bytes": 3254,
+        "sha256": "35d47ac078591221fcb666219bcd824f9400fc17aafecdc0499d8336f7178238"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/vidur.md": {
+        "bytes": 7736,
+        "sha256": "d38602af5219f5feeb49e6cd6bf53d66098c77fad141dae676f1d8a6743cf25d"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/roofline-contract.md": {
+        "bytes": 7091,
+        "sha256": "419ec4bcaec45ac579a133bce083d4f48779f26e664d7a86f720fa65f343b1d1"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/roofline-example.json": {
+        "bytes": 2015,
+        "sha256": "597cd3205df043cff3ae9d15924c77934e8d8eb300c00c4840f6bb2921f90faf"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/scripts/roofline.py": {
+        "bytes": 9004,
+        "sha256": "524de48a6764401695fe1444f1fa3d38f363069b0cd2fba4715a86ca4b28a5d4"
+      },
+      "src/inferrsi/data/skill_library/methods/plan-activation-residency/SKILL.md": {
+        "bytes": 3507,
+        "sha256": "542d8a1e6963f5019986e1fbd79d3a247a045a51487c09313026087ac457c15e"
+      },
+      "src/inferrsi/data/skill_library/methods/plan-activation-residency/metadata.json": {
+        "bytes": 478,
+        "sha256": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7"
+      },
+      "src/inferrsi/data/skill_library/methods/plan-activation-residency/references/source-notes.md": {
+        "bytes": 13907,
+        "sha256": "5ff52655cfd06fcf970ed2a7d4d4fe1c31115b102af2d41497de0c74848ed2b3"
+      },
+      "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/SKILL.md": {
+        "bytes": 13800,
+        "sha256": "7e20d3390f2501263e5e298e026c47cac16c234f2397833131e15ae5d34c0b97"
+      },
+      "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/metadata.json": {
+        "bytes": 1903,
+        "sha256": "d0e4d6950250e4cec43d82e23324c9301a925d4c22be30b4d2713d790cb6504b"
+      },
+      "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/references/capability-presence-audit.md": {
+        "bytes": 11461,
+        "sha256": "c21eb4a6aabe811a49a6540a34c54dbcfdcf7efae3052a79b54034c7838026a5"
+      },
+      "src/inferrsi/data/skill_library/methods/skill-evolution-playbook/SKILL.md": {
+        "bytes": 9448,
+        "sha256": "9cd37f21fe0d862069e4798edca8a04d59f8b840e55d521901594a63dddd9ba3"
+      },
+      "src/inferrsi/data/skill_library/methods/skill-evolution-playbook/metadata.json": {
+        "bytes": 527,
+        "sha256": "92b36338b9257e2e41b95e4370694d7de52ed84bc7971056b3bfced07b0993d8"
+      },
+      "src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md": {
+        "bytes": 6865,
+        "sha256": "64bd2c6b795d966045d1d3cca55976d3d49b08b06df7c7e06a70c2824cf55b7e"
+      },
+      "src/inferrsi/data/skill_library/methods/training-input-pipeline/metadata.json": {
+        "bytes": 587,
+        "sha256": "80659099fce6a6ec05e9084b75452c877a5812e30f60a78f31ef8c7efcdc1bac"
+      },
+      "src/inferrsi/data/skill_library/methods/training-input-pipeline/references/sources.md": {
+        "bytes": 2637,
+        "sha256": "60e0f3e88655d8e1d5eb4797f39ca4370deda01614f265b8d2ea4559d8ac020f"
+      },
+      "src/inferrsi/data/skill_library/methods/training-resource-budget/SKILL.md": {
+        "bytes": 7200,
+        "sha256": "bdfa5a4ef37d9f303c3a23cd6f98c21137b97ac5ed58e0ed5fac6d71d33e52ef"
+      },
+      "src/inferrsi/data/skill_library/methods/training-resource-budget/metadata.json": {
+        "bytes": 976,
+        "sha256": "d42b614af9cee6ee0d4a25dc5d57c7b98438feed1e807a7c9f1755109a86e9cf"
+      },
+      "src/inferrsi/data/skill_library/methods/training-resource-budget/references/mindspeed-historical.md": {
+        "bytes": 3384,
+        "sha256": "eb97f935ed30517a8fc9edec5e9cddc04f5d679d156c16d75bd29c534593360a"
+      },
+      "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/INTEGRATION_NOTES.md": {
+        "bytes": 7368,
+        "sha256": "c55c6ad2131ef788866411dcc7056f221420ebff30c95c6bc919b09a4818cb73"
+      },
+      "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/cases/paper-grounded-cases.md": {
+        "bytes": 5649,
+        "sha256": "19bd25b1a944adb2351ebc6b1d795150ea033041a44f6cc261e881677812a101"
+      },
+      "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/knowledge/offload-delta.md": {
+        "bytes": 2026,
+        "sha256": "f8740fc5c8bbbe5ac13ff1cfec7ca499569d0b46e8cd56cf1f5a44c3a5e72242"
+      },
+      "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/notes/training-memory-scheduling.md": {
+        "bytes": 18660,
+        "sha256": "9aa6d3e01452dff95c8a8940ff62ead8c146ce215c9175b78c9f2ef508a7b2de"
+      },
+      "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/references/source-readset.md": {
+        "bytes": 22946,
+        "sha256": "2a5c166767f56fa47e7eba7386f1f36da8cb73702e5b5bbe14d6a456ca9ebfae"
+      },
+      "src/inferrsi/data/skill_library/methods/training-resource-budget/references/sources.md": {
+        "bytes": 2752,
+        "sha256": "6d248f58af8b350fc66c22898988d6f451ac60d03bcd805cd6e84df5e55174ac"
+      },
+      "src/inferrsi/data/skill_library/methods/training-resource-budget/references/stas-decision-notes.md": {
+        "bytes": 4086,
+        "sha256": "f1a993e6c4ca9a035a2effcaf5500081f227b0cc56cc885a6b22fda0041fc744"
+      },
+      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md": {
+        "bytes": 25969,
+        "sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213"
+      },
+      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json": {
+        "bytes": 2418,
+        "sha256": "798cbff6f10b7c24b188080ebf233b188b0e98ef2c6fdb5e00e1fe3eff727c0f"
+      },
+      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md": {
+        "bytes": 8656,
+        "sha256": "5a39f14b102b0cee304541e011aa31e44e6b13527d1bee9987c3f3b66c064d3d"
+      },
+      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md": {
+        "bytes": 6392,
+        "sha256": "6c640f28ee3f0253a17b81461c865cb9f947367002cb75bf72550c4572d02e97"
+      },
+      "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/SKILL.md": {
+        "bytes": 7181,
+        "sha256": "a46110891eca860915e314d6921a22856d631453b22b17837fe30262d6780535"
+      },
+      "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/metadata.json": {
+        "bytes": 2175,
+        "sha256": "4e6100dfb9fc7bc00b19915f4fd38ad509704f2a609c658adf4d632d4e05b6c1"
+      },
+      "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/references/afd-plugin-LICENSE.md": {
+        "bytes": 12747,
+        "sha256": "0e8566ea362276878e1f38f97da49964dbea301c15d4aa9d727f97a4c39fd943"
+      },
+      "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/references/source-notes.md": {
+        "bytes": 17240,
+        "sha256": "5073d87450cb60d1b245308dff8807e73749c07de976d50f2fa26694fad81a72"
+      },
+      "src/inferrsi/data/skill_library/repositories/comfy-ui/LICENSE": {
+        "bytes": 35149,
+        "sha256": "3972dc9744f6499f0f9b2dbf76696f2ae7ad8af9b23dde66d6af86c9dfb36986"
+      },
+      "src/inferrsi/data/skill_library/repositories/comfy-ui/SKILL.md": {
+        "bytes": 4507,
+        "sha256": "c993bd6ff8e9b5f82f39f175ec8242341136e4ee0808af414e1d94ad3e196c5e"
+      },
+      "src/inferrsi/data/skill_library/repositories/comfy-ui/metadata.json": {
+        "bytes": 2032,
+        "sha256": "74742851a002a51c208362cdcd2a59fdab5d1734d8492b1f93fbec3cf21e0841"
+      },
+      "src/inferrsi/data/skill_library/repositories/comfy-ui/references/graph-cache-invalidation.md": {
+        "bytes": 7679,
+        "sha256": "614c9a4e6943d24d7d7be7d632682ce8b1dcc239124a37d9dc1d487113772eac"
+      },
+      "src/inferrsi/data/skill_library/repositories/comfy-ui/references/model-residency.md": {
+        "bytes": 7191,
+        "sha256": "014c68386e7cdfdb7b4f566836e8b2ff61bd5a5a506e7b4723f83fe80c5d4de7"
+      },
+      "src/inferrsi/data/skill_library/repositories/comfy-ui/references/repo-provenance.md": {
+        "bytes": 5648,
+        "sha256": "ca996ef515af42a3ce304747fa3f12a31c00533deaace5ba2ce21aefa0a62e2e"
+      },
+      "src/inferrsi/data/skill_library/repositories/comfy-ui/references/validation-boundaries.md": {
+        "bytes": 3523,
+        "sha256": "1913dedbeb558d872d6a0e015cc11d822910c3cb259a236db29fb60ad8a64263"
+      },
+      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md": {
+        "bytes": 12294,
+        "sha256": "4703beb6e5194ef2ff2c0b5ec69f64e35417a4bc0dfb50a10c80a920b77ef3b5"
+      },
+      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/metadata.json": {
+        "bytes": 2333,
+        "sha256": "014be24d0e669f4acbc0717dc323eec784026cf185186619d5a19d6d84496131"
+      },
+      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/source-notes.md": {
+        "bytes": 12036,
+        "sha256": "fd917bd5ac80bee9c9f4525f8e0300f6d67dc144e2a16632af42a67f3d53d098"
+      },
+      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/validation.md": {
+        "bytes": 5402,
+        "sha256": "60e71e70c5519bcdf6c171eec2be2f342e8801bedd746d6005f11adc646dfaff"
+      },
+      "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/SKILL.md": {
+        "bytes": 11202,
+        "sha256": "f157562d1edc905cb4fc07246567e3c5b844efb9fbf5fe39f0d40553a1b1a307"
+      },
+      "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/metadata.json": {
+        "bytes": 2156,
+        "sha256": "4d68bf4c1767eaee12ade963bbd551d0e61d1bcd81a984bb02fd3aae46965919"
+      },
+      "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/references/source-notes.md": {
+        "bytes": 11633,
+        "sha256": "6fa955022f16cd598f44fd6521901d1cb868b06cec545cabe2b82769a641b5d6"
+      },
+      "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/references/validation.md": {
+        "bytes": 7196,
+        "sha256": "673d93f6700052974b340fab25b3438e0e6344960569ac9abec7d7720f662288"
+      },
+      "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/LICENSE": {
+        "bytes": 1077,
+        "sha256": "663e51edcd9a3a8f858018ce6924ad2895a24c269c26c39ad75325bcd0f7e5bc"
+      },
+      "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/SKILL.md": {
+        "bytes": 5432,
+        "sha256": "fdb4d9483c6efa836f9e59617da605fa9c2186e11c2041b05d5c81850ba0475d"
+      },
+      "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/metadata.json": {
+        "bytes": 4923,
+        "sha256": "fe626a77af3f084f36fd73ffdb6f67976e7dceeafb94ba22078ef03c3a96cfae"
+      },
+      "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/THIRD_PARTY_NOTICES.md": {
+        "bytes": 1774,
+        "sha256": "a0d77db7d2aca0eda811ab3c386cc32b3143f2c2609a042cf4642954f0f6fb09"
+      },
+      "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/minimax-h3-apple.md": {
+        "bytes": 12059,
+        "sha256": "57ce01754ba96c378f13a91305ef3d418b926f99b2753f0af99315572bce063a"
+      },
+      "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/minimax-h3-weight-inventory.json": {
+        "bytes": 10664,
+        "sha256": "fbaacaeec340d3ce893556ee864405a0746af4d13530a38cb3d5aa9d60929271"
+      },
+      "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/phase-residency.md": {
+        "bytes": 9408,
+        "sha256": "2836a550eb2be7f5a862f51ec7b135ebd9715bf3c4faa84a7ecbc9d9c11f5d8d"
+      },
+      "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/precision-and-streaming.md": {
+        "bytes": 9440,
+        "sha256": "7e166c89193f4c7aed852ec95346a2de2b5008dac002e2e1862be1b3e57f6093"
+      },
+      "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/source-identity.json": {
+        "bytes": 7747,
+        "sha256": "730f8dd23648a3f8f47b9726cc613d893a79d1817d515eec609a9b06d8a9ff90"
+      },
+      "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/source-notes.md": {
+        "bytes": 9448,
+        "sha256": "7304bf9369efc157dc7aa40686dd828cc14db45c49442e908b3d3cc4b329eb86"
+      },
+      "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/validation.md": {
+        "bytes": 7880,
+        "sha256": "0754873a758e048480938314894e25b504ae0c269549312e7f8e6020d83fe294"
+      },
+      "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/SKILL.md": {
+        "bytes": 9338,
+        "sha256": "95e81b6bc096a954fa73139ea5bc001d1efa6c99747161da62f012b494e5c013"
+      },
+      "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/metadata.json": {
+        "bytes": 1718,
+        "sha256": "1d30b31dc8352fc6a8d7b8abbef8118c840e47e5b90795dae0c758bf9717278c"
+      },
+      "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/references/source-notes.md": {
+        "bytes": 15234,
+        "sha256": "3d12a401e9decb55a2f4f3b2bfa0d360ef5110f63a0469dfe285654e78054190"
+      },
+      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md": {
+        "bytes": 10812,
+        "sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27"
+      },
+      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/metadata.json": {
+        "bytes": 572,
+        "sha256": "c221189b8d85fac3bd41ef1c7cd560f1aebc8cfa0381c182d1d571a7169be829"
+      },
+      "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/SKILL.md": {
+        "bytes": 3788,
+        "sha256": "bf1177cb9985c12a9bf11edb64771018e1e314c4713920de8fd4a989ff70115f"
+      },
+      "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/metadata.json": {
+        "bytes": 1299,
+        "sha256": "d86cecb615e93ba0f6be09d1ead102cd5a17068c07e1be2549ac5cd8f7519279"
+      },
+      "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/cache-stop.md": {
+        "bytes": 2026,
+        "sha256": "bba6b5c73261065309cbcf11b1e591eef48916f0c79ceaaf74d98a9fa96b0992"
+      },
+      "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/generation-batch.md": {
+        "bytes": 1733,
+        "sha256": "f16d3c57d45e63bcaffd3871aa41c76972ee11f7dbf1e12b45d9b170303cad9b"
+      },
+      "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/source-map.md": {
+        "bytes": 6358,
+        "sha256": "91d26181f246307f2113f94b12ac9587d689b654dcd216e9f14b11cf5ebb650c"
+      },
+      "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/validation.md": {
+        "bytes": 1665,
+        "sha256": "73b5dd347017baba519ecc85415aead9993c1751d116b9ba4f07689f58c5f783"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang-omni/LICENSE": {
+        "bytes": 11352,
+        "sha256": "a54c8553df5d4313d293d1aef3bf2f13952bda8b7c7a1fdd729b2bbe27ee6015"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang-omni/SKILL.md": {
+        "bytes": 2304,
+        "sha256": "d75f5b2881b240ec2d9d4f5cc8e3f7d53e4378cb01732a2fb9699da2901ba38c"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang-omni/references/repo-provenance.md": {
+        "bytes": 1778,
+        "sha256": "5e8dbe168a8885e209d4b4532944ffa7fead658b23fd6588054d57b22476f63b"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang-omni/references/repo-routing-metadata.json": {
+        "bytes": 511,
+        "sha256": "6820707cb9bd877d1635928cf5a0efe38e4fe37dc04036dc70d53e479ed9bf7c"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang-omni/references/snapshot.md": {
+        "bytes": 3488,
+        "sha256": "0fb5d776349de845dc314f16ce73a06c9caf5db84dc1c293327d3b3ca1337c80"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-architecture/SKILL.md": {
+        "bytes": 2355,
+        "sha256": "766b2d66d2056dd779abd9e492cb3af6e25899298f44964f565509f130a7cb24"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-architecture/references/runtime.md": {
+        "bytes": 7734,
+        "sha256": "36bba9874fe054acc5217f9c265f5ae309c83943e0d46228370ae1bcdc69421e"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-extension/SKILL.md": {
+        "bytes": 2193,
+        "sha256": "9af38bca840d0004fc2ed049141476fc40fd40dcfa843f38a404a7a314c19425"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-extension/references/integration.md": {
+        "bytes": 5775,
+        "sha256": "6ce9383ca17bacff60a180f8fb8c43ea47eaa37117fb8da6051ad86135d228ef"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-serving/SKILL.md": {
+        "bytes": 2335,
+        "sha256": "990b6bdebfe22612de49dbebcb16810e8abef94d70691aa883b7937c85210eb6"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-serving/references/recipes-and-tuning.md": {
+        "bytes": 8918,
+        "sha256": "94afb762bbe185f58606154b3f0a5af56e7089cb84ea94e26cfcb9f93c95439f"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang/LICENSE": {
+        "bytes": 11346,
+        "sha256": "1495e1e757ef4d0925a2350563cf5754bb23c51701a8ec4fb3c5cdcbedae6747"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang/SKILL.md": {
+        "bytes": 5466,
+        "sha256": "ae72d1f0cc750699bf79d7a138a97219209bca7b7e5f19cd427d306245eaf841"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/LICENSE": {
+        "bytes": 11346,
+        "sha256": "1495e1e757ef4d0925a2350563cf5754bb23c51701a8ec4fb3c5cdcbedae6747"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/SKILL.md": {
+        "bytes": 5738,
+        "sha256": "b0a4ba4751087126143ef86983798deb0ed00e24b6ff06bf6ef76a9df6e274f2"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/metadata.json": {
+        "bytes": 2844,
+        "sha256": "6ed57dfb4f25ecb34b81f2188b055e05d172ba2581e6a3b605b814b2f733a4d6"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/references/source-notes.md": {
+        "bytes": 19512,
+        "sha256": "d0fcfd1177ae2f351041c5b2abba7f5121cec133006d9e4dca8f74ddb567332d"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang/references/architecture-and-memory.md": {
+        "bytes": 7735,
+        "sha256": "798e747022b7ec135cc4861c2c9ac44fb14b503bb77a0068ee7fbea494b764a7"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang/references/native-candidates.md": {
+        "bytes": 3961,
+        "sha256": "abafcc9a34df196a6365145554887b8d37afb36c6e236c92e50a4a06c8a4ec93"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang/references/parallelism-and-extension.md": {
+        "bytes": 6324,
+        "sha256": "e3be7f4b51264ac68682a64d264e4d96857d5bb9f9d99cae5f327c56984a9e6b"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang/references/repo-provenance.md": {
+        "bytes": 2274,
+        "sha256": "534d95ae8d0b29bcbc8f6707b3d38fb964e31f5b66ba13517a9a6b8f309bbb38"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang/references/repo-routing-metadata.json": {
+        "bytes": 625,
+        "sha256": "b2b879c1cc42ddc2a5b95ae24115eb891562fa9a6e96d151f4ddf6f7719fe73c"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang/references/source-map.md": {
+        "bytes": 8945,
+        "sha256": "c2adad64ceb0c305cdc4859c03012b3e0a4ab26cd0cb289bbd76e192b0cc3794"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang/references/tuning-and-validation.md": {
+        "bytes": 10832,
+        "sha256": "49b3343b3292cc098a787de9d24d968b444d0ef4f73fedb5d8183390e8f39c9c"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-parallel-extension/SKILL.md": {
+        "bytes": 2316,
+        "sha256": "18bd54ebdd629e1ba6dfa589f391c7cc3c5355785c2ae82f741b64ee8022464f"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-runtime-memory/SKILL.md": {
+        "bytes": 2154,
+        "sha256": "33b1a65aba00fca819d2b6cc42202e7708e9155e4d8f4373a2387f4ee2f2fa3e"
+      },
+      "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-tuning-validation/SKILL.md": {
+        "bytes": 2018,
+        "sha256": "744b2ef9122101d3f50cbbb40fe020e2b3d3888adbe2d86041041e7c7ce424fc"
+      },
+      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE": {
+        "bytes": 1062,
+        "sha256": "b53fa08f515cb5a6fff7b9fd8fcd0961a4b80df29d74372d25e1f9171aa042ee"
+      },
+      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md": {
+        "bytes": 5504,
+        "sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6"
+      },
+      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json": {
+        "bytes": 2202,
+        "sha256": "f7b7581a3f44ba9774480c559408fc66d6a4d6d1dbbeabe4529f1eb7d4dd5ad6"
+      },
+      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md": {
+        "bytes": 8136,
+        "sha256": "7b537cb972347dd68e303960f142892346d8b19b884e4d08016ea36a23bdaa7a"
+      },
+      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json": {
+        "bytes": 3617,
+        "sha256": "af4830370b6eddc2d176a9ae14da4ab56a90b0fe048a01a89ee051df4365d8e9"
+      },
+      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md": {
+        "bytes": 6589,
+        "sha256": "b67affbd7d237d82b7f51be49f1537e9820fe56ec70a71ee76809f9e599ca693"
+      },
+      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md": {
+        "bytes": 3394,
+        "sha256": "5dcf5fa7bf734fbe13214f57d2dfb40ff8215d9cd34e072a9e615bf103b82fc2"
+      },
+      "src/inferrsi/data/skill_library/repositories/strata/LICENSE": {
+        "bytes": 1093,
+        "sha256": "949b9b0aada7476006469e58ee4ffb0e0f68c2d9137e90a356757b61a09d37f0"
+      },
+      "src/inferrsi/data/skill_library/repositories/strata/SKILL.md": {
+        "bytes": 2813,
+        "sha256": "0e5daeb6409c916be45ecd7ce2531ea5f1c60a4a53e2b0ed25dc1e2cdc5c17ed"
+      },
+      "src/inferrsi/data/skill_library/repositories/strata/references/design-space-analysis.md": {
+        "bytes": 27736,
+        "sha256": "11ecafe71530ddd4839268ea1b24b3361a3895012a705859ca06dae29fa72630"
+      },
+      "src/inferrsi/data/skill_library/repositories/strata/references/repo-provenance.md": {
+        "bytes": 2913,
+        "sha256": "05f17d0ae996ad6c12314bdd29f3427574f6aa611ea60effd18598a0f386638b"
+      },
+      "src/inferrsi/data/skill_library/repositories/strata/references/repo-routing-metadata.json": {
+        "bytes": 332,
+        "sha256": "a10ba49b7e6565109b566c105e17551d042075a1ff649c6f371134e3b7402c74"
+      },
+      "src/inferrsi/data/skill_library/repositories/strata/references/trial-format.md": {
+        "bytes": 2291,
+        "sha256": "92d8cf38549268ae4eef382890662d03e18a7dff858ad45ee5eb6f47eb39c8c0"
+      },
+      "src/inferrsi/data/skill_library/repositories/strata/references/troubleshooting.md": {
+        "bytes": 2079,
+        "sha256": "7301cfa5a5f6cfbf9576c3845a96c6b4de8f8a6509f2330970a8e2e4669b3d5a"
+      },
+      "src/inferrsi/data/skill_library/repositories/strata/scripts/summarize_trials.py": {
+        "bytes": 5170,
+        "sha256": "6a243a08da15659ced6519f49f52a49a675bd9e952740c3ab03dd21964eefbe2"
+      },
+      "src/inferrsi/data/skill_library/repositories/strata/sub-skills/design-space/SKILL.md": {
+        "bytes": 1826,
+        "sha256": "857bea3c64db45dc035728a41523142ac65778b41f63092148b8ab936f971d76"
+      },
+      "src/inferrsi/data/skill_library/repositories/strata/sub-skills/design-space/references/design-space.md": {
+        "bytes": 24231,
+        "sha256": "0a428dcd12d5290b554e9da6795a110f94d98f128771602e1ea5f608048b135e"
+      },
+      "src/inferrsi/data/skill_library/repositories/strata/sub-skills/tune-execution/SKILL.md": {
+        "bytes": 1145,
+        "sha256": "cdf51c537e7237296d52ef6cda9f2e4c9d9ce3fc87265a369cd9eb7497c53512"
+      },
+      "src/inferrsi/data/skill_library/repositories/strata/sub-skills/tune-execution/references/tuning.md": {
+        "bytes": 11310,
+        "sha256": "7b69aa7872d60dd3fb91e41b77a632572070dcf80dd5b5f6dc1722d8540177b1"
+      },
+      "src/inferrsi/data/skill_library/repositories/strata/sub-skills/validate-engine/SKILL.md": {
+        "bytes": 4613,
+        "sha256": "90bfd975e25f18895ce0e85f6f35d24149ce026305a4d479aea4924510053460"
+      },
+      "src/inferrsi/data/skill_library/repositories/strata/sub-skills/validate-engine/references/validation.md": {
+        "bytes": 16895,
+        "sha256": "dae846f4c47c68731b96ec1d50cb008c03871c72078f1c5b0f4a9423be9ba397"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm-omni/LICENSE": {
+        "bytes": 11357,
+        "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm-omni/SKILL.md": {
+        "bytes": 4945,
+        "sha256": "df3e63dc28de2dac2aa92824fd88719159138650d96487332e0effca3f0e2a80"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm-omni/references/evidence.md": {
+        "bytes": 4689,
+        "sha256": "ae0d51ba56b34ed81fba34de8fd3685bebf2f160b188fa53719936002881d220"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm-omni/references/repo-provenance.md": {
+        "bytes": 2221,
+        "sha256": "43bbfe431434ca4ad6157ba35cbb9ce2d89e4065f427abfdd5f372f51ba0c9c9"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm-omni/references/repo-routing-metadata.json": {
+        "bytes": 342,
+        "sha256": "7784b4d98d096718c4f4cc42ce86d5ec113c469844e12de4690b0fccca59d329"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-architecture/SKILL.md": {
+        "bytes": 1336,
+        "sha256": "fb3ff50a7b9122a2545a97ef5179011237c4c4ca214dde7236ad7bf79a7c8623"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-architecture/references/architecture.md": {
+        "bytes": 5600,
+        "sha256": "db845eb4207ddeedb8ae381f1b91704a42cb53ba0ee6c7ee581510ba1a412190"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-serving/SKILL.md": {
+        "bytes": 1343,
+        "sha256": "8ab3209b838635395b22deca6aa879172016531d45c2374eb8e498e7f71bb40f"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-serving/references/serving.md": {
+        "bytes": 5049,
+        "sha256": "9bd5eff300d9755137a7600c8f7a7753bb22d09c8853e08062d7e535662ba8a3"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/SKILL.md": {
+        "bytes": 2227,
+        "sha256": "93c83bc4b799b32f50ff54b870cbbde050a790af16d9b7b062039c4987c80148"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/references/abort-acknowledgment.md": {
+        "bytes": 8547,
+        "sha256": "e95ae5b2cc410917ddf5e9066890d0bef9264104529bdca5ac12bdf56fe80ae3"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/references/validation.md": {
+        "bytes": 4632,
+        "sha256": "1542061f9c76fff2acb6529a58238301484b0e23ec0c697d2fdf129e200ea973"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/LICENSE": {
+        "bytes": 11324,
+        "sha256": "50e6751797c50dedd75ef1b8a0d9e42f5f8472e9fbce91f34718e9f97b0c780a"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/SKILL.md": {
+        "bytes": 2863,
+        "sha256": "38e522b8abeb0dc140a2194340432ed41dc8b569d33b63d5d4c59aa1092cd2d1"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/SKILL.md": {
+        "bytes": 6664,
+        "sha256": "313733a65cc48442b18dda1c4665139a689963a90e1032a2235bc6f40e3a8e06"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/metadata.json": {
+        "bytes": 781,
+        "sha256": "b43fc3a7ccae59c664fcdf9ff77962edcc07c2a3df8b59324db551642bd23e23"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/references/source-notes.md": {
+        "bytes": 13853,
+        "sha256": "3665b034594fc8cc767c408002f4e16f8803369242405027c7f37ce6c9ecebe9"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/SKILL.md": {
+        "bytes": 6565,
+        "sha256": "5fb9ea1d1373e110d9b506f69963c4857338b7dbf5e38ec9a49b59a0ff683ed0"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/metadata.json": {
+        "bytes": 1372,
+        "sha256": "78459416734f61a8faa819d9a1a3ad7b9a985e8bf8e149813c604d15be6bb92f"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/INTEGRATION_NOTES.md": {
+        "bytes": 9006,
+        "sha256": "82b24be82451ee121892f61575d0cf4ecb9920813a25ccb26cce2b45dda6d8b3"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/migration-is-not-free-memory.md": {
+        "bytes": 2145,
+        "sha256": "2ac168f2725cba7fc92dfcc67c01bf2258dff6a2dbe1ddb063bc60c175d82d0a"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/parallelism-change-with-state-cost.md": {
+        "bytes": 2069,
+        "sha256": "5accc81086e4500d66f894ef6af9951e1cbb76941f6c631aab5c5561a786d8d8"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/preemption-under-kv-pressure.md": {
+        "bytes": 2630,
+        "sha256": "462200a6d0b193ef01f32374a463e0f2838ca8edf33d9ab898def65ffd0b490a"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/knowledge/scheduling-state-boundaries.md": {
+        "bytes": 7488,
+        "sha256": "d4ef29869959b835a5dcb06ddb2fd8e6cff49bb176994dc1d46eb9a852f6228d"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/fastserve.md": {
+        "bytes": 9289,
+        "sha256": "9b4e275843000ef453d508d9c3c9c3f8ea1de5545fbd01d7abe74baa8a4931a7"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/llumnix.md": {
+        "bytes": 10181,
+        "sha256": "f252ce33ac2db017701ca1d49813c5ec1996c4c92900cf35bb6bda7fc59c0f18"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/loongserve.md": {
+        "bytes": 12422,
+        "sha256": "7bc6d1e256d461b04acc8ed856cf99108faceda9cde729ab95bcca596fe019ca"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/orca.md": {
+        "bytes": 8826,
+        "sha256": "f5a3b5870d64226408a8fcda5d69483769d823f8a5efd7f11698c8da483ab195"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/fastserve-readset.md": {
+        "bytes": 7173,
+        "sha256": "9866bd0c78b2b638998b3b632f6bae26aef032462c81abee336d309af5b07831"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/llumnix-readset.md": {
+        "bytes": 9386,
+        "sha256": "3a2819d65ad24087c08530492ef39d8ea046ed1f780d53512f1d23e2b3578487"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/loongserve-readset.md": {
+        "bytes": 12490,
+        "sha256": "f1a383e7bdbbd5a5a472d7f6e1292c8a17d4910ff736ca5a6aa422e1bee74f8c"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/orca-readset.md": {
+        "bytes": 4796,
+        "sha256": "a497ef23b7e60c0fbc15d95b7e07cd20547cfe56f37c24dfe6ed70d79ef7a3cf"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/source-notes.md": {
+        "bytes": 9575,
+        "sha256": "84033fd514773f10c0bddd9914fabb980497ba2b308f226a922c3e1888413827"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/references/evidence-map.json": {
+        "bytes": 14384,
+        "sha256": "dd8bbab5d285ed3a1a19d2f06ad84e8c444251690d8f673def3e2681765ef6c1"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/references/repo-provenance.json": {
+        "bytes": 1313,
+        "sha256": "4e705a964697712e950e02bedc41320806a41311b7f4dcf32f923d60b1d2e919"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/references/repo-provenance.md": {
+        "bytes": 3261,
+        "sha256": "6e841c38c35ff2cd6ff8139cb1a6c8f75701647897f1089a94a5c62928609941"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/references/repo-routing-metadata.json": {
+        "bytes": 432,
+        "sha256": "a20b54f682eb52311e445ac62b95d1935c001296f5cf23b7764cb04477d535c9"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/SKILL.md": {
+        "bytes": 1835,
+        "sha256": "2a945b5fddacfb90d1cd7cb2b2f767067d641061423a65e1c19afecec7f3c06a"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/references/capability-boundaries.md": {
+        "bytes": 6560,
+        "sha256": "c8081d03f04f7481080c46c7dcf728962729bc122f49f9502d50cd1e3d04bcf3"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/references/runtime-and-invariants.md": {
+        "bytes": 8265,
+        "sha256": "afa1f967545021f36bc1b59dbf6aa2c367f4d5ece4f08dce755e2017cf50dbe5"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/extend-validate/SKILL.md": {
+        "bytes": 1731,
+        "sha256": "495679e5b99341bf80daeaacd2d4c2699aba16bc16c447cbd8d2fe17f7559798"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/extend-validate/references/change-boundaries-and-validation.md": {
+        "bytes": 7624,
+        "sha256": "f7fdd2d699f09d96fbf5ca9382f9b56113c870db2688d8e1118eb8163519fc4a"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/SKILL.md": {
+        "bytes": 1812,
+        "sha256": "0ef4d2e3a96cbb277557e169a3fc8edf0f0504aad03259771b432a33ee008f78"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/references/commands-and-measurements.md": {
+        "bytes": 5667,
+        "sha256": "76ab51083b14595c49fa1c255cb7e5ebb996f14930a51ab3cbe864cc334b5f80"
+      },
+      "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/references/experiment-decisions.md": {
+        "bytes": 7644,
+        "sha256": "242593c0d386e57104b57a09bc7db97c2bbbcdb905bb68a397cfe612e07fb083"
+      },
+      "src/inferrsi/data/skill_library/skill-index.json": {
+        "bytes": 7514,
+        "sha256": "8ce35e18f7629260d67a95617b9918eecff73ed8d9370dd3d4595c725ada9aa1"
+      },
+      "src/inferrsi/data/skill_library/workflows/cite-check-downstream-pr/SKILL.md": {
+        "bytes": 1093,
+        "sha256": "485932cdf3b1fba5dc69be73e75d38f86a7b79442d26e82a296c963a7598d226"
+      },
+      "src/inferrsi/data/skill_library/workflows/claim-contract/SKILL.md": {
+        "bytes": 1075,
+        "sha256": "3af1dbbe81d91cb05fc2ef15a2ae25d7592a08558e1a5535f49c52f33a488c55"
+      },
+      "src/inferrsi/data/skill_library/workflows/competitor-scan/SKILL.md": {
+        "bytes": 1088,
+        "sha256": "0ede58ace65a9324a538b4816add1684f230ba536d4364fe5a5dd2c0bbed3776"
+      },
+      "src/inferrsi/data/skill_library/workflows/draft-acceptance-issue/SKILL.md": {
+        "bytes": 1230,
+        "sha256": "ce983acfb7d7d3272775e7cf00c92e4ddadd97511b602575a76e5ab7ade01b56"
+      },
+      "src/inferrsi/data/skill_library/workflows/inference-engine-research/SKILL.md": {
+        "bytes": 5681,
+        "sha256": "364beed2ca617e71f4f955bbd024740d411b94eeb9557d7e055d4463cc22caba"
+      },
+      "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/cross-repository-mechanisms.md": {
+        "bytes": 17050,
+        "sha256": "28efa1b25f548763180b70d8d0ddebb7d4e7227df189f871fd233fe1d4881303"
+      },
+      "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/import.md": {
+        "bytes": 3463,
+        "sha256": "c8a57db94768cc6cd4193932def75ba67ec453e56ee0e54f176fb82f1af53091"
+      },
+      "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/model-aware-preparation.md": {
+        "bytes": 9056,
+        "sha256": "787c017aade68ee1133b979d636fa24062302cef6174c001171c963947ee8da4"
+      },
+      "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/overview.md": {
+        "bytes": 8099,
+        "sha256": "7ed0a966239b9481b82dd349245cfbfe966c030fc916771c91fd0692464ffc56"
+      },
+      "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/repository-source-scope.md": {
+        "bytes": 11967,
+        "sha256": "85202ca209171c50573c2fc209acb1e937d8211e1652e4ca9c6cc4bce04d03eb"
+      },
+      "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/source-manifest.json": {
+        "bytes": 1613,
+        "sha256": "0ce1678803d1ede6b485fb6b4b1c88aba909056ee6e3b5af691899ceb2056240"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/SKILL.md": {
+        "bytes": 6502,
+        "sha256": "74cd863df2e80dc2bda1713d24207809173a3116f30325e3487e1e3517cfec1e"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/metadata.json": {
+        "bytes": 3236,
+        "sha256": "12fc9272bd5fbf02dd9b897d8ca40984b72ae54f20492ace5304711bbc4d71c3"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/COMPARISON.md": {
+        "bytes": 4267,
+        "sha256": "38587678a96581f5c2dc697d982127343565504bffa191352811dffcc165aa23"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/INTEGRATION_NOTES.md": {
+        "bytes": 5557,
+        "sha256": "fbc4a96b6f6c477706eed83eaa4fcc29a14f881637629d6c004601ed049f2523"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/QA.md": {
+        "bytes": 1853,
+        "sha256": "3f3785ec39a6938a5d1f2e874e99153fe92c0e666dd3e78a499c8b47d891346a"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/README.md": {
+        "bytes": 2151,
+        "sha256": "221785e91c45588f82b3caab6d52f1ea1d102b16139c20a0802921e2789e6d28"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/UNREAD_LEADS.md": {
+        "bytes": 1038,
+        "sha256": "6c1d87ebe491d27666703149eb05ad582c7e5aff77f0ac5c11a7b89c946f96ee"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/cases/methodology-counterexamples.md": {
+        "bytes": 5213,
+        "sha256": "f9fc21753ea41eef77ca020d78d031e945e5dd835b5d09c5ff0e53749ceb2ca4"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/knowledge/evidence-routing.md": {
+        "bytes": 5282,
+        "sha256": "406d5d74c3fef4c1a23d495492e21a084c3c46495a1a84b385703ed104358874"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/learning-to-commit.md": {
+        "bytes": 2847,
+        "sha256": "9c2d1d738e896d05c00f53e6adf14c9df2501ecc3f0020e71b5a1c2192547086"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/protocol-details.md": {
+        "bytes": 6520,
+        "sha256": "41298c6bde3a28f0712fa06a18bbc4ac59aea6d17cc9eb1318744f4121e3bab6"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/readset.md": {
+        "bytes": 7086,
+        "sha256": "6f953d20d8958c68339c9ceeae348604d0fea3e9c6ccb9b64a63d3b210426a22"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/repo-to-skill.md": {
+        "bytes": 3151,
+        "sha256": "5ebdac2a2bbbe6614f9d4096bc9895bd685acdf76414e2c3c1fb7b50d57c738d"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/skillopt.md": {
+        "bytes": 3040,
+        "sha256": "148075daa67e19f4b4351710498438d17c279a9bce56b2d01cab26ff3be33e99"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/source-manifest.json": {
+        "bytes": 2431,
+        "sha256": "716cd49ed01824ee9f6b6fc55ebf2b24fdd5d5a3e677f6e28e34a472616bbe25"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/trace2skill.md": {
+        "bytes": 2917,
+        "sha256": "230d13b77ba819b5e3edee147523e8cad6bfce95afccc140fb4946d6c8747a2e"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/updates/bounded-updates.md": {
+        "bytes": 7630,
+        "sha256": "d02c5bd0109846d1e1987de074810321cb4af277cd3f66ab2ac268aadd5ee142"
+      },
+      "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/skill-factory-design-20261014.md": {
+        "bytes": 9465,
+        "sha256": "fb8ce1ad83b7a6ddc657e89ee2eebb710873b213cfd85bdd6d823c8691b3cb9f"
+      },
+      "src/inferrsi/data/skill_library/workflows/rsi-evidence/SKILL.md": {
+        "bytes": 3680,
+        "sha256": "a55610874b658b86a39a1a049efb4e4eab0140f6579ed74dd7718a4ca8cedaca"
+      }
+    },
+    "sha256": "6e19365a0ad8e2951af0f3106f93aa545b4ad00b6e0f0317a20f32065b6a05fb"
+  },
+  "environment": {
+    "filesystem_cache": "not controlled; process cold is not OS-cache cold",
+    "implementation": "CPython",
+    "logical_cpu_count": 8,
+    "machine": "AMD64",
+    "perf_counter_resolution_seconds": 1e-07,
+    "python_executable_sha256": "737a7e3b71e3578f8432acc7dd88c452e593622c544bc13da4789d69c63da5ae",
+    "python_version": "3.12.6",
+    "release": "10",
+    "site_initializers_disabled": true,
+    "system": "Windows"
+  },
+  "git_head": "4817f92121df73fd9cd5fb4a319e3c56a68b4a5b",
+  "knowledge_record_count": 108,
+  "library": {
+    "bytes": 2294958,
+    "file_count": 342,
+    "sha256": "be8726296e6186601692dc2d95d99bfbba76f0b4c758de3a88a9b2a523eaa37e"
+  },
+  "modules": {
+    "inferrsi": {
+      "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+      "source": "src/inferrsi/__init__.py"
+    },
+    "inferrsi.catalog": {
+      "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+      "source": "src/inferrsi/catalog.py"
+    },
+    "inferrsi.cli": {
+      "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+      "source": "src/inferrsi/cli.py"
+    },
+    "inferrsi.knowledge_operations": {
+      "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+      "source": "src/inferrsi/knowledge_operations.py"
+    }
+  },
+  "packaged_skill_count": 60,
+  "python_sources": {
+    "bytes": 274444,
+    "file_count": 53,
+    "files": {
+      "src/inferrsi/__init__.py": {
+        "bytes": 102,
+        "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47"
+      },
+      "src/inferrsi/__main__.py": {
+        "bytes": 48,
+        "sha256": "935a1c1166b0c1ea35a82256345000bf2c73ded718d77773bc27a71ecce28f7d"
+      },
+      "src/inferrsi/_skill_selection.py": {
+        "bytes": 5790,
+        "sha256": "4640399f4013425589f6702e2db844820619ba35967d1c3519978eb307168440"
+      },
+      "src/inferrsi/arguments.py": {
+        "bytes": 12200,
+        "sha256": "4456600cf4a362cf069594fe36db46932cf5cb14ddd37683aaf3538f5cf76b4d"
+      },
+      "src/inferrsi/catalog.py": {
+        "bytes": 6855,
+        "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5"
+      },
+      "src/inferrsi/claim.py": {
+        "bytes": 805,
+        "sha256": "5f24c4fa91b48f0bb4d4554531607763fc8fc108e90f332816f06cfb520b5cc1"
+      },
+      "src/inferrsi/cli.py": {
+        "bytes": 14205,
+        "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88"
+      },
+      "src/inferrsi/contracts.py": {
+        "bytes": 119,
+        "sha256": "03000cb9246ce8c0967147533f7ad25df918ba2674839801453ffb870f496490"
+      },
+      "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/scripts/roofline.py": {
+        "bytes": 9004,
+        "sha256": "524de48a6764401695fe1444f1fa3d38f363069b0cd2fba4715a86ca4b28a5d4"
+      },
+      "src/inferrsi/data/skill_library/repositories/strata/scripts/summarize_trials.py": {
+        "bytes": 5170,
+        "sha256": "6a243a08da15659ced6519f49f52a49a675bd9e952740c3ab03dd21964eefbe2"
+      },
+      "src/inferrsi/evaluation.py": {
+        "bytes": 183,
+        "sha256": "482c40710d428206c59fe3696703248cd86c4b3c7d19ad41017c1369bd6d780e"
+      },
+      "src/inferrsi/evidence_paths.py": {
+        "bytes": 859,
+        "sha256": "cc46231e0b8334ef5742defdda20e0093e22b30c86af59df1fb3ae4b5caa70fa"
+      },
+      "src/inferrsi/execution.py": {
+        "bytes": 755,
+        "sha256": "f43d8cbb4d43a72a7ff8a10b29a5fb9cf4cb22af54dd35943283c843fba0da92"
+      },
+      "src/inferrsi/harness/__init__.py": {
+        "bytes": 177,
+        "sha256": "1006e4713003f44bfb32b2e385a918e1652b9e9833801e4555674a71e57fb0d5"
+      },
+      "src/inferrsi/harness/evaluation.py": {
+        "bytes": 4160,
+        "sha256": "9b00b2e93959f16b7b76bb4b257df7dcf48520a06fd6adae96e9a165a05e6e80"
+      },
+      "src/inferrsi/harness/improvement.py": {
+        "bytes": 5777,
+        "sha256": "3faa0f5928a707a98f35f45d1ca18e226c898941e6a07b3829eebeb53e70aeb8"
+      },
+      "src/inferrsi/harness/reporting.py": {
+        "bytes": 3566,
+        "sha256": "38c58f02c44e8c8aeacf46527e142be0e3142a83e4b2693246d0ca76c0ddd7c3"
+      },
+      "src/inferrsi/harness/runner.py": {
+        "bytes": 6567,
+        "sha256": "a566307b9b586655941b627f428b5121ed38d7ac80bccd1e1799e443fb0de911"
+      },
+      "src/inferrsi/harness/scenarios/__init__.py": {
+        "bytes": 79,
+        "sha256": "4e97913c806077f0cadfeb29d986d2fa75b55ab8766cb86e6e3e9a195a9a1f5a"
+      },
+      "src/inferrsi/harness/scenarios/external.py": {
+        "bytes": 1016,
+        "sha256": "84d5f6920501a9d882cc4fffbc721d2051fdbf7c90694bde9cbfba3c2b866304"
+      },
+      "src/inferrsi/harness/scenarios/media.py": {
+        "bytes": 2914,
+        "sha256": "67d979a3ed6f13a8e3e1e40a304f78e5a3e6c2556b3aa8a11c41156745d7050a"
+      },
+      "src/inferrsi/harness/skill_evaluation.py": {
+        "bytes": 11093,
+        "sha256": "b710b8e53758757efd68c703449eb4bab7eb7ec50ae3dfc52a010cdebd8c3bf5"
+      },
+      "src/inferrsi/infra/__init__.py": {
+        "bytes": 85,
+        "sha256": "8779897733779266841b0f79b84b1dc43f0fa57c013855c906c809829ecfc77c"
+      },
+      "src/inferrsi/infra/artifacts.py": {
+        "bytes": 1040,
+        "sha256": "e72e1bebdbb88e42de430238489252386851f9e6645ed7780f2aad2e90338120"
+      },
+      "src/inferrsi/infra/campaign.py": {
+        "bytes": 1028,
+        "sha256": "be505fba1592611bd84ceaca55afbac027205413a2639baefa0c3bb343230209"
+      },
+      "src/inferrsi/infra/commands.py": {
+        "bytes": 811,
+        "sha256": "1890f115b3a2c5c6832173bf60432ad595bd6bf1c69783ecd460c6f8961194f1"
+      },
+      "src/inferrsi/infra/environment.py": {
+        "bytes": 2067,
+        "sha256": "60e46c2aefc216db202b41f44542e7b893e08e0721c667b9ccc50411d1ac84ee"
+      },
+      "src/inferrsi/infra/executors/__init__.py": {
+        "bytes": 64,
+        "sha256": "327c3cd5c821c832ff260a41dba0ee1e439cdabefc9edfb9cad1cd38d14bc87a"
+      },
+      "src/inferrsi/infra/executors/local.py": {
+        "bytes": 3654,
+        "sha256": "37bcbca9f3261c8c58a8bf806c9da40720b18d0676c82e8ed24c3901ac9bbe77"
+      },
+      "src/inferrsi/infra/http.py": {
+        "bytes": 634,
+        "sha256": "f65d6c2b55b6a89694caaf6dcdf7d74f417e9ced103eca1b3f218e305c6b6cb2"
+      },
+      "src/inferrsi/infra/inspection.py": {
+        "bytes": 2415,
+        "sha256": "d1c66deb0cced529bca996d2f88b60f51fa3417ff1cae365c650fec808946ecd"
+      },
+      "src/inferrsi/infra/platforms/__init__.py": {
+        "bytes": 1494,
+        "sha256": "9e88e8b9a396ed90c221e1585135c0a1b58da41c1bb73ed9237c53b6684abfed"
+      },
+      "src/inferrsi/infra/platforms/cuda.py": {
+        "bytes": 3309,
+        "sha256": "9e0e4276e1b2984605493b14b82fba3e86f6635204719c318302e624f66f8b6f"
+      },
+      "src/inferrsi/infra/platforms/mlx.py": {
+        "bytes": 1971,
+        "sha256": "f3b81470db28bd6183ccca2e763c5e079686bba6cdfe6c08ebe2cdf64216f2e3"
+      },
+      "src/inferrsi/infra/processes.py": {
+        "bytes": 8151,
+        "sha256": "3b96ebc7a50913e78e86ff3dcba74c8ee8d03d547d9591ca6488f390ac1e6189"
+      },
+      "src/inferrsi/infra/targets/__init__.py": {
+        "bytes": 186,
+        "sha256": "1f3ea75dded7c6cc967754c7b6d4f30ef00da8191a749938aabc3c6bb8f86eb3"
+      },
+      "src/inferrsi/infra/targets/git.py": {
+        "bytes": 2910,
+        "sha256": "6327c4f126086f310b6de309fc7659c9759da28ed76efaeeb94a3d2fba29a0c2"
+      },
+      "src/inferrsi/inspection.py": {
+        "bytes": 692,
+        "sha256": "afa44bf4cdfbb58a318c21f9b292ca142d1bbb22750e35a4cc8244da5350ebb3"
+      },
+      "src/inferrsi/knowledge_operations.py": {
+        "bytes": 27444,
+        "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b"
+      },
+      "src/inferrsi/knowledge_store.py": {
+        "bytes": 13700,
+        "sha256": "5e794e28906ce4687aff7f14a507e267184640fc8a4fac19e0ced279ce6a6cb6"
+      },
+      "src/inferrsi/model/__init__.py": {
+        "bytes": 77,
+        "sha256": "3c1dba4f5159fd9fbc8d4d96e2a83c8441491c1b1dc7b42031a3fc1782ed9d9f"
+      },
+      "src/inferrsi/model/anthropic_agent.py": {
+        "bytes": 24696,
+        "sha256": "b0ef9cd542544292eeb90ab05ef410be9ed6e5d1a40d729890f49562c28c12f0"
+      },
+      "src/inferrsi/model/protocols/__init__.py": {
+        "bytes": 85,
+        "sha256": "5a2f54483b164de04e309dbebb87a700b213682b0f44a50d89a752dd71a4ce47"
+      },
+      "src/inferrsi/model/protocols/image.py": {
+        "bytes": 2047,
+        "sha256": "251a1f8f885da2cd8e9d3baba766c525afe293091606b4e8668c43ba2cb7cb5f"
+      },
+      "src/inferrsi/model/protocols/metrics.py": {
+        "bytes": 610,
+        "sha256": "17db16d6ef15b41fb60d8f09507893accc43f9cb3ad686b8dc4e9839713ebc56"
+      },
+      "src/inferrsi/model/validators/__init__.py": {
+        "bytes": 65,
+        "sha256": "1d4b6ffa968e283219ddcc8a09de9280238a8d65ae2dc8cd2b60889877a8eb0b"
+      },
+      "src/inferrsi/model/validators/flare.py": {
+        "bytes": 19975,
+        "sha256": "3696b0609c772f2222b448833608565a033123d7e92b11fbc545dd356fec590f"
+      },
+      "src/inferrsi/model/validators/image.py": {
+        "bytes": 3204,
+        "sha256": "79470b6b06a22d53ec6c40636a0b3d821a2e6ee2f7bb7aed3dac7241504add1a"
+      },
+      "src/inferrsi/reporting.py": {
+        "bytes": 127,
+        "sha256": "1f771890ab3e69b1aa9246b548a7825596bdf904f6c395dd218208084266853d"
+      },
+      "src/inferrsi/research.py": {
+        "bytes": 5213,
+        "sha256": "9bfe59dd17b5476c18ce50d671711053c62800521c50d6b710e8140a2b4dca04"
+      },
+      "src/inferrsi/runspec.py": {
+        "bytes": 4398,
+        "sha256": "dca0952425caf42095ef0c7e2feb8ef4740cc422ecb61e777f203962d9b87709"
+      },
+      "src/inferrsi/skill_catalog.py": {
+        "bytes": 9938,
+        "sha256": "65f5a1f3f6894009abb98ec2fec30a63cc1bb2f6ce6aeb01f9f7a3dce0439f2d"
+      },
+      "src/inferrsi/summary_chart.py": {
+        "bytes": 40910,
+        "sha256": "d29023d60f6be804c1b01c9eb73765d4824d6c3bb870fe22c99a24544bc1d8bc"
+      }
+    },
+    "sha256": "4faa529dc47f23c24f8c5b8e0110d8e413aab20df5ad61db591fb751564a6448"
+  },
+  "tracked_source_dirty": false
+}
diff --git a/evidence/knowledge-retrieval-20261027/results/dry-run-quality.json b/evidence/knowledge-retrieval-20261027/results/dry-run-quality.json
new file mode 100644
index 00000000..557db9ff
--- /dev/null
+++ b/evidence/knowledge-retrieval-20261027/results/dry-run-quality.json
@@ -0,0 +1,6344 @@
+{
+  "started_at_utc": "2026-10-27T11:11:53.933313+00:00",
+  "dry_run": true,
+  "verdict": "INCONCLUSIVE",
+  "freeze_content_sha256": "f2aa423228a8802c83c52631600f9b95551f97068346f4b8ca5206977f28b385",
+  "identities": {
+    "A": {
+      "corpus": {
+        "bytes": 2883003,
+        "file_count": 458,
+        "files": {
+          "src/inferrsi/data/knowledge_packs/README.md": {
+            "bytes": 14172,
+            "sha256": "b877e67d9cccf0dcbe57419eeb3ff9374e5bd03ff8eec63adef1025c446140e9"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/diffusers/cuda/diffusion/group-offload-transfer-overlap.json": {
+            "bytes": 3277,
+            "sha256": "9bec3d912192efa2f2c853d3fb70bdae21a75875630a02294fcd084b18c30414"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/flashattention/cuda/attention/generation-specific-dispatch.json": {
+            "bytes": 3085,
+            "sha256": "90290d7bc80b1626863e0520c91bb196d93ebfe08190a450e9de59c6505ae8ad"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/flashinfer/cuda/llm/paged-decode-dispatch.json": {
+            "bytes": 3745,
+            "sha256": "0f2986494fcc9b751df71e598fd61caa68e28bc4b0ac401e4cc763bc3f75283b"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/cross-stream-dependency.json": {
+            "bytes": 3809,
+            "sha256": "b829315e55adf8e11f0bd581d5f45dd130d7eb3f81ce7de532f63c7b9c27cbb5"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/evaluation-completion.json": {
+            "bytes": 3841,
+            "sha256": "19d2e506ac4f50895776a83a062059aa7a052175c8e4b21dd0d46e1a5a6716a1"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/post-detachment-buffer-lifetime.json": {
+            "bytes": 4663,
+            "sha256": "18d78c0e542dacb42ca33e5264d92aef80fa06d163fd16d4feafd5ed0827ab99"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/pytorch/cuda/compiler/cudagraph-shapes-and-lifetimes.json": {
+            "bytes": 3623,
+            "sha256": "7fef4122ff794967e61f55d02a4a6b558d7e498d9513607f5ca86db6386928a8"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/transformers/common/llm/kv-cache-strategy-contracts.json": {
+            "bytes": 3072,
+            "sha256": "a2df26052112a9f3fc6242b2bd636ab13991442db57e56646168d6c52032d38d"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/triton/common/kernel/bandwidth-fusion-resource-limits.json": {
+            "bytes": 2941,
+            "sha256": "6e9673d1a600a6ad34820e4325f2a6417f44f0f96d8ed0bcf21466d998f19ee4"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/cache_dit/common/diffusion/residual-cache-calibration.json": {
+            "bytes": 3841,
+            "sha256": "bb9509ad0225a91b52bc6787477581d05f0ecd5e44fa59f08f718eea36a2ac45"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/bounded-windowed-replay.json": {
+            "bytes": 10959,
+            "sha256": "43613225e9ccdbd34c0f7e3e0767887b9bbfd0f38b2c21684677d99e31a432f0"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/speculative-verification-budget.json": {
+            "bytes": 10177,
+            "sha256": "a729e1fc00be44e1480a5dc65bbcdd58df8599681df74dbc651a5a49f27e9028"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/zero-overhead-deferred-scheduling.json": {
+            "bytes": 8191,
+            "sha256": "ed5426af145f066f6183bd316c4073c46f324f5ee83aa9bd616387c0d0e7149c"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/lmdeploy/cuda/llm/kv-quantization-capacity-cost.json": {
+            "bytes": 3366,
+            "sha256": "873997b38981ccc03ff73e6fb84e96022f293a019843efd8c4aa50f27878ac9a"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/codec-constants-and-state.json": {
+            "bytes": 4250,
+            "sha256": "1cecd7905fef5af3ef8952a4d803900a7df91d662ebae34db47cc0dbf352d4e7"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/fixed-customvoice-prompt.json": {
+            "bytes": 3660,
+            "sha256": "ea5e8fdf54cb731b255b73aa0cf47ef53fac103baa2bd5246be45f95d54cb593"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/graph-batching-state-ownership.json": {
+            "bytes": 3950,
+            "sha256": "130741a9f7ac9d366784d6c4dd42860262235caff0c277de1e8a8f8ec933b37f"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/sglang/cuda/llm/hierarchical-prefix-cache.json": {
+            "bytes": 3641,
+            "sha256": "b50ca58df28d4343545d07adf0da23f8f181297b30aba7870e5690630e7d7315"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/tensorrt_llm/cuda/llm/prefill-decode-disaggregation.json": {
+            "bytes": 2967,
+            "sha256": "349856dd98628f94987e014889cd1b15d0e3ae6ab6c576cdde917f28fbafb102"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/xdit/cuda/diffusion/parallelism-freshness-and-topology.json": {
+            "bytes": 3236,
+            "sha256": "3674b5916cb7b384951bf3e449b6796b161878daabcc1d7e7f6d6f8ff861cc17"
+          },
+          "src/inferrsi/data/repo_packs/afd_plugin/pack.json": {
+            "bytes": 502,
+            "sha256": "00042abeabdc9c0e24f1a4edaba47e233df398aef0ad4e181fc70191d383da5f"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-aclgraph-capture-replay.json": {
+            "bytes": 4311,
+            "sha256": "327414443ea9537ae320aa405912c338351bc06cefdc0b99352e1eea36f61dc3"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-device-health.json": {
+            "bytes": 4115,
+            "sha256": "633fa5fdcea1e1e9edda8a218d6a3fde0d12637848348e1e2fe7d1b0bd744a16"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-distributed-parallel.json": {
+            "bytes": 5841,
+            "sha256": "709ca0d061e4d6917fe8fa049fe93165e71dfecf96c7bf264ced25346381f603"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-graph-mode.json": {
+            "bytes": 4135,
+            "sha256": "bb82ab3d628b91be957e557545810ef5c20790be7d43296096f20a21f45ec391"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-model-adaptation.json": {
+            "bytes": 4137,
+            "sha256": "7ea1ab0f0044afff935a91071b36ac09bc46b5ef044fe2a68ab0608256ddd853"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-profiling.json": {
+            "bytes": 4186,
+            "sha256": "0d27058fd8d0ef8b524b7a60d1cb76950622746c94b5a03915e64e6edb4ab3d4"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-tuning.json": {
+            "bytes": 4439,
+            "sha256": "fcb49d71f59d8b39c1eaa03d24e95e2cbfe5f9d7f6e0f22df03ee0d60f6a568a"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-optimization-casebook.json": {
+            "bytes": 4371,
+            "sha256": "05783630bfaf3c7ef99b848865f4bbab50d0b52e80ad04fcd028c624d7638e9b"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-performance-analysis.json": {
+            "bytes": 7200,
+            "sha256": "b6eea897c12d0c6ac9d172089535c82c25eb71c09d7a4972e64812cc4682b5c9"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascendc-tiling-design.json": {
+            "bytes": 4505,
+            "sha256": "e69b6e89dd342ba55ae54b5fac14aa1b91ee2a361da40cb669c287ce519f4517"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/cann-attribution-overlap.json": {
+            "bytes": 4915,
+            "sha256": "8ac01b60f5cd37fd594d6cb5119e4e6525199d017296fc8779dc68a576f5525c"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/torchnpu-device-memory-lifecycle.json": {
+            "bytes": 4349,
+            "sha256": "7a79d0eb2c50381d40086e6db72d4bdd2cec5d2261bfb00cf73a60701a94da2a"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/torchnpu-pinned-host-memory-lifecycle.json": {
+            "bytes": 4431,
+            "sha256": "8bb34dcfbebb48667a94434b036a1a408b970af9890b047a3de325f364ed5f7c"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/assess-serving-preemption.json": {
+            "bytes": 4198,
+            "sha256": "e3d2a1a72d0d980f31fc85deee27db26bc0daa0c709e8ae99e42deb814a4a7f4"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/assess-split-backward-schedule.json": {
+            "bytes": 3915,
+            "sha256": "a8561794786eb0a897d831f00c92b4fdd3b0492316c1b1d8ad13c1f8f75ac815"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-h3-runtime-residency.json": {
+            "bytes": 12169,
+            "sha256": "602dd625874bd4a89861d5bb25881cc63e4b5a7fdedc2867089f0952943f6eff"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-llama-runtime-state.json": {
+            "bytes": 5816,
+            "sha256": "64f6ab8a95172378350323cead08bc6c201c624b711e6242938a6a21a8e83ad7"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-sd-cpp-residency.json": {
+            "bytes": 8349,
+            "sha256": "95ed97b717255be3b817e5c69debd6d41197559dbf21771507d5da79134a72e1"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-sglang-diffusion-lifetimes.json": {
+            "bytes": 7276,
+            "sha256": "5e5889223878c6cf42a49c6f09c60e9d9add88cb7e7d0df050c4e75d88a05aa6"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-vllm-kv-lifetimes.json": {
+            "bytes": 4275,
+            "sha256": "00424005a5eeb8b15f239edd2cb2c71047787e3ecf55924860cb7a78d137b586"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/bound-windowed-attention-replay.json": {
+            "bytes": 7877,
+            "sha256": "98b2291147a796ef59cf20a075ff662998ee6a529f85b49e9c6fb62fe41fad9f"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/budget-speculative-verification.json": {
+            "bytes": 7490,
+            "sha256": "66448f59460c9dffac46e74160388fab650cf8bbbd5bdc95c1465b70daf0c2dc"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/calibrate-serving-simulation.json": {
+            "bytes": 3992,
+            "sha256": "055b70e0b86f65156775b519e4b8cbf06a956a8993b06a66175167cedb78f2cb"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/cgroup-resource-controls.json": {
+            "bytes": 4801,
+            "sha256": "ad406b6dd5d9415bd28d448e64269a388b50e11f9eb7f5a9adb6b0a6246fd303"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/checkpoint-recovery-consistency.json": {
+            "bytes": 4740,
+            "sha256": "ccb85314a831bde58f8cc515580d001b174c11dca0bbcbc6221303acf629046d"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/choose-parallel-training-plan.json": {
+            "bytes": 3901,
+            "sha256": "40afcb41c59988fc1af8fdd00d2340800c54e96847dd0de16ecddf7ac221d1b4"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/cite-check-downstream-pr.json": {
+            "bytes": 2533,
+            "sha256": "dabff14684d5ae9a2b05a9743ff5b76d00a90fdb29319588aba67f5087aad706"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/claim-contract.json": {
+            "bytes": 2469,
+            "sha256": "78f15f6533434fc29adf1df651f7a15eb8ba81733f8b02c710d90102541d707b"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/comfy-ui.json": {
+            "bytes": 6919,
+            "sha256": "1496807d958df055f720ca93e6dcf7f7e8f506f3f12c94c9a2166f789b00058d"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/competitor-scan.json": {
+            "bytes": 2449,
+            "sha256": "c2e97e49e5185710cdfaa08ac1c2b1c19b958e68ebfca885d5aee0ed8c9f8c8b"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dcp-async-save-performance.json": {
+            "bytes": 4538,
+            "sha256": "359a5acf8304a78fc074688676f9a0be62f56f463d0960a544e409f49843bdaf"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dense-performance-evidence.json": {
+            "bytes": 4227,
+            "sha256": "d87f93a9dc9ddc583e3929baaac85f7a425666dd6dd71665933f18344ccec834"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/diagnose-vllm-admission-pressure.json": {
+            "bytes": 9789,
+            "sha256": "a1de40def775b2e840c27330b67e6f37c3c244db1731fab2bd3e8a787143210e"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/diffusers-invocation-state.json": {
+            "bytes": 6109,
+            "sha256": "914d2ffe9776ba931bed780a2a270c1525e3c6cbaf0735b8531064d3180f3233"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/draft-acceptance-issue.json": {
+            "bytes": 2549,
+            "sha256": "6aaa63dbfe6cda982ccd3a4e368527b80fd2873672c000c66d06d28e40a2bf14"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/evaluate-elastic-serving-placement.json": {
+            "bytes": 4334,
+            "sha256": "fab1fa814e789ec2d86587fcf7de7c38174464b9e9ba0cf53ef3fb470053e7c4"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/generation-cache-lifecycle.json": {
+            "bytes": 5972,
+            "sha256": "e5679cdecb722bbd17280913913da8473052ad09e9e7a13b3a9b95463a139ef6"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/host-memory-management.json": {
+            "bytes": 6530,
+            "sha256": "a36aea21169281866b2468860d3ffeb912e61edd571308b02f82939f17f694a8"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/host-overhead-optimization.json": {
+            "bytes": 4243,
+            "sha256": "05fd7f054966b96892980d585352bc1c3fcd47f439cb34268fdf07feee58a998"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/inference-benchmark-design.json": {
+            "bytes": 10735,
+            "sha256": "091bb086c6f0123dbb5db78968c13f846cad135e8969982efe557644b7c7613b"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/inference-optimization-casebook.json": {
+            "bytes": 6531,
+            "sha256": "c512025d7a59934c91665e24752d9393ca9b6a9dfb624c72789294defae93c0e"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/kernel-writing-guide.json": {
+            "bytes": 4545,
+            "sha256": "f1f0ca1d11d3550c9b8c5fdbed8f6cd09b877ac14341f973d58dca02b5d8995d"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/kv-tier-management.json": {
+            "bytes": 4609,
+            "sha256": "6141f61fb6b0d147b704ab7a2627e9babdb53a90e15b659152a3d565967afddf"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/lazy-evaluation-stream-lifetimes.json": {
+            "bytes": 5934,
+            "sha256": "34f47009a8a33f8907bd78f110b95fcd62409bef1b38e1055acb46b2eab285ff"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/npu-skill-factory.json": {
+            "bytes": 11868,
+            "sha256": "24c1d3bff288e86b99ea674e3252ec6f54d9ba944274aef6a7666197afd6788a"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/optimization-search-policy.json": {
+            "bytes": 7101,
+            "sha256": "d7bf71d5491765f62eec02fe5013dc1a5ef56e1516892da48593c6165a86366d"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/os-process-placement.json": {
+            "bytes": 6304,
+            "sha256": "e6e8fe5d7d66f3e98f05762d94441726dfa757ce7fbffd2b35402b3f327f8fb3"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/performance-ceiling-analysis.json": {
+            "bytes": 13695,
+            "sha256": "47c034d96026e1c2a02848d9c9c3caf4024a5033664bfd089ad30d9f23a80deb"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/plan-activation-residency.json": {
+            "bytes": 3871,
+            "sha256": "436e364020cc854fd52a10ee76e4593c7519bb0546c8aa5238fcfd89158e8ab0"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/repository-evolution-analysis.json": {
+            "bytes": 5865,
+            "sha256": "e0af402c3320ebd9bd36c774cb52c0708721a86ca446c05f6e94306888838e30"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/rsi-evidence.json": {
+            "bytes": 2475,
+            "sha256": "c8b4bb5a8c88009ac5b1526e82639715381c940d3c3eb295f0333cae6c2d3ca9"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/skill-evolution-playbook.json": {
+            "bytes": 4091,
+            "sha256": "0e063c2e7686de0c27c061e1aac0eb2d8c4fa83246e32ddeeb0e28126c176425"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/storage-io-measurement.json": {
+            "bytes": 4783,
+            "sha256": "80e589e3a0b67d3a045e03feeadc45e720ceb277e6124c917eec04863eb67419"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/training-input-pipeline.json": {
+            "bytes": 4296,
+            "sha256": "ae696bea8d2b7bfdaa29b6e5bc0bd861abcc0e42e7d9bec6ae20c6e7d6f27669"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/training-resource-budget.json": {
+            "bytes": 6944,
+            "sha256": "ea5e5e864e3aa34bd00be67c8e741877ae01480f039652073b65b71dfcd54c97"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/weight-loading-optimization.json": {
+            "bytes": 8139,
+            "sha256": "7f2d1fbc6b1cfd80c3d6ed8df107d4ca9b97655626ee078c345445a303b1b367"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/audit-afd-p2p-exchanges.json": {
+            "bytes": 6640,
+            "sha256": "9ac45aa784b430ad1ca9ffddecce8dbae9740cfe9c98c10b330e01d350015a38"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/diagnose-attention-bottlenecks.json": {
+            "bytes": 7602,
+            "sha256": "e525d75578de1cc164e6553d01efbd0a014e74646200580f59b1161c030413ad"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/host-overhead-profiling.json": {
+            "bytes": 3965,
+            "sha256": "e9e6ae8c5b9986e83cade3bd029baae112859e01aa6c96f741a264f3bee44adc"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/reason-hopper-kernels.json": {
+            "bytes": 15086,
+            "sha256": "758a649e651d8f9635ac2df45d18043e156c8d54b8f8dbc95f8e68ebcb866a28"
+          },
+          "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/parallelism-and-extension.json": {
+            "bytes": 3399,
+            "sha256": "1253f5f01bcf1498703c7f34960f0c50580df5ba8f140af557a903e296ad81d3"
+          },
+          "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/runtime-radix-hicache.json": {
+            "bytes": 3317,
+            "sha256": "5a0b2aed51649af214cdc1034f57650b7ce195c934a17452ee67a35131f80cdd"
+          },
+          "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/tuning-and-validation.json": {
+            "bytes": 3385,
+            "sha256": "ef4d4f7182da4433d59d72dda0d21b4aeafadd73260be11ab8d4cb36c20fafc6"
+          },
+          "src/inferrsi/data/repo_packs/sglang/pack.json": {
+            "bytes": 620,
+            "sha256": "bce2e0c42796a9ef499d459c37b56d1d56f4b19344841359c534fca29c0359ce"
+          },
+          "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/audio/serving-and-streaming-quality.json": {
+            "bytes": 3440,
+            "sha256": "97fc1d5a11a39a0634dae9e9602a4ec47a4d2348e121076c2d5c836cf0692114"
+          },
+          "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/omni/extension-lifecycle-validation.json": {
+            "bytes": 3482,
+            "sha256": "7da8dc5d2655099de2bcd5f0fa922d9e1bcb0bfd2f571f026a004a7f402a75b6"
+          },
+          "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/omni/pipeline-state-and-transport.json": {
+            "bytes": 3404,
+            "sha256": "0c57692772019504264e39bf0444a69e20cff86f6e112b072c53cf7a4299cc82"
+          },
+          "src/inferrsi/data/repo_packs/sglang_omni/pack.json": {
+            "bytes": 630,
+            "sha256": "8e70336bc2aac0c74732a026337813652b5d5c6e1fc60f5571e9442345077b96"
+          },
+          "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/architecture-search-space.json": {
+            "bytes": 3294,
+            "sha256": "2c5ad72d5bab4f7082fc3be1532c347444cde38f5a84639fe16b22134b68685e"
+          },
+          "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/controlled-execution-tuning.json": {
+            "bytes": 3255,
+            "sha256": "6a2bf611a432213919e1c237687507ceb0ab753cccad0e934dc77f86604805e9"
+          },
+          "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/engine-validation-contract.json": {
+            "bytes": 3295,
+            "sha256": "ef89775bc82965fe87a832c7ec8a5cf55cb9eba4ecfc6279caea9ce3b0410c41"
+          },
+          "src/inferrsi/data/repo_packs/strata/pack.json": {
+            "bytes": 617,
+            "sha256": "643b32defb8db9aa230f624856f4fb4b595b5c23db58d9453091f5420cff71c3"
+          },
+          "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/bounded-serving-measurement.json": {
+            "bytes": 3239,
+            "sha256": "fb00ddac14ef6b06f48da9ba335529bd55d5d83c2a2fe331f0aa1ca16afceb37"
+          },
+          "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/extension-correctness-gates.json": {
+            "bytes": 3343,
+            "sha256": "4ea7083d277df205ea4c361bee87a8207951596e9ef6cfa136cc7d2aeab02f46"
+          },
+          "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/runtime-and-kv-ownership.json": {
+            "bytes": 3280,
+            "sha256": "e6142c1065754716a5f26f42119722908dc8fff63e667c986d57f1227b8610bd"
+          },
+          "src/inferrsi/data/repo_packs/vllm/pack.json": {
+            "bytes": 508,
+            "sha256": "1705d96710249c4514c15d8d011b43c1a1ae7c15fbc9ddf11ef2d0e6b2ab7d07"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/minimax-h3-reference-kv-reuse.json": {
+            "bytes": 4047,
+            "sha256": "80b853fc594d387ab48bed04dca245e0a22b8cdaf7c9b853551ab3bafa302e84"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/minimax-h3-sp-head-buckets.json": {
+            "bytes": 3742,
+            "sha256": "c4d29d1bceb5e2cf86151bf53c9dafb3d83428fb1e9d2a848d852bb8824bf857"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/qwen-image-21-eager-t2i.json": {
+            "bytes": 5268,
+            "sha256": "ef107d43cc448219eda4d64ab961b61cda6350e20d30d94de1ff7c4ffff7e775"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/audio/streaming-serving-measurement.json": {
+            "bytes": 3389,
+            "sha256": "2680f32f28907f6abc7a7e1e3ecab4f46ee412f40f4898ed2f371ffadf47ae4d"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/measurement-and-profiling.json": {
+            "bytes": 4472,
+            "sha256": "c8b25583cd160024c7472a20debea34cb6d4bc79714b4bad2953555adc9f794e"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/production-integration-gates.json": {
+            "bytes": 6112,
+            "sha256": "649b49ff64d3e9e88bd620027dabf9d126ca69bbe91b448f28e845ecbe1f8e2f"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/step-state-and-output-validation.json": {
+            "bytes": 5090,
+            "sha256": "e1ffe28c4fa54af739cdef2fedf4755bba18e078f8380d4fde8138ea712e5778"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/upstream-review-gates.json": {
+            "bytes": 7614,
+            "sha256": "5127709bb85ce31008d19468e1b4fcb5454a71776f69f991841cd9ef7558f9c4"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/omni/stage-ownership-and-search.json": {
+            "bytes": 3427,
+            "sha256": "e75d6c2829ae22881893f5da2ad7edba099fd64ccfb0fc93d30473e06e9d4d80"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/audio/concurrency-scaling-state-and-overlap.json": {
+            "bytes": 25324,
+            "sha256": "cb47fabb232db81ed525cbb54371becc7e4a28b2bae9d8492f60dfb6f31563e8"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-adaln-schedule-cache.json": {
+            "bytes": 4253,
+            "sha256": "ddbba3bb72b5789938595a82d3280e861f75f6ed54f05a1d2b7dcb1089d30f63"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-cuda-mxfp8.json": {
+            "bytes": 4468,
+            "sha256": "e4c538f03f4ab90a4acc8e3d8721bbd226aec1ee9bde45757a7c27d3d970821d"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-fp32-modulation.json": {
+            "bytes": 3625,
+            "sha256": "abb8d94a1b025f0e2c8e085275e96ee85d68aeb27cb12d0918e4647f8255b693"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-qk-norm-rope.json": {
+            "bytes": 3706,
+            "sha256": "4c1c4acfb9cc0baa90bc28fed1903020ea1ea38755b84315711795db45e48bb6"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-ref2va-teacache.json": {
+            "bytes": 3835,
+            "sha256": "8aeb3f99ac4227cf8b3a5cece13d938b19082e265a557eddae27b7e495bcc22e"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-vae-stacked-tiling.json": {
+            "bytes": 3508,
+            "sha256": "8cc18eacbbfd8fc4d3a6b4152c4b584d559fc9b15377da56ef072e542d807c93"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/pack.json": {
+            "bytes": 566,
+            "sha256": "58c971e9dcf6ac15752629ed4d9426aeb5b5d857fc37b5bcf921aa2e81c0cdd5"
+          },
+          "src/inferrsi/data/repo_packs/vllm_rlt/pack.json": {
+            "bytes": 649,
+            "sha256": "6097041b8d026740060e527f49b1b4e288315b1c3227f7583805b86d8fa415f7"
+          },
+          "src/inferrsi/data/skill_library/README.md": {
+            "bytes": 18819,
+            "sha256": "ae583f3578425df1dc91ba32b8b6543c0469c9a817ebea873a77051d6b13e1ce"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/SKILL.md": {
+            "bytes": 5494,
+            "sha256": "7581b56336d07213a87d01477793ae1a8acf0b7cf876fce36c9a761ee32c1ff4"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/metadata.json": {
+            "bytes": 1429,
+            "sha256": "68ac7f64b7ab947fc2db664bc3990f102c1cc5539b42350aa5818d0bbf17d86c"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/ownership.md": {
+            "bytes": 2081,
+            "sha256": "40688cc44183027efbf2fca8d3548c299cdf2e33fef54f76ab3549353eb9bd1b"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/source-map.md": {
+            "bytes": 4635,
+            "sha256": "10562a75683bd9bd07ec2e37a1cbd0f93f2534d6f85b55a566adb2d97485215f"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/validation.md": {
+            "bytes": 1477,
+            "sha256": "5a0b854b87d091d7d505b0bdd5c03330052507a851ab6edf35b93a6675f2d4f7"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/SKILL.md": {
+            "bytes": 4010,
+            "sha256": "a95569bcc0da6e715fd74783a6a1e0489e92b680c582f15dafbf98df589d505a"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/metadata.json": {
+            "bytes": 1519,
+            "sha256": "9c2293c8956508de987c886c7224d673d22912e7e915b026b05abdd287e97401"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/execution-lifetime.md": {
+            "bytes": 2147,
+            "sha256": "c4bc41e08aca68f661325f61e6f0db1760a079dc362f295c847bdc374fee3bf6"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/source-map.md": {
+            "bytes": 5918,
+            "sha256": "798b9cb89ed6b885eb47464ca9f137d2802a54571b9327e631d7742d3e5abd5d"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/validation.md": {
+            "bytes": 1608,
+            "sha256": "ebfc00bf7d6ab0a1dc415953ae84d56ccdc713d5614aff85e433c4a843793bff"
+          },
+          "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md": {
+            "bytes": 7322,
+            "sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845"
+          },
+          "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json": {
+            "bytes": 587,
+            "sha256": "db7ad368394ccbfdeff6a16e3423c300d51305b1f95ff5c3936df91738a8cc35"
+          },
+          "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md": {
+            "bytes": 5314,
+            "sha256": "22afe4f4d77f441096cc657d55d763473bfcbea8fb531c8e493fe49ab84955f7"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/SKILL.md": {
+            "bytes": 6745,
+            "sha256": "d494fc7e539cbb7388189bd7b3f04bb5c4983fdd7a82484348aad1cee7f741f2"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/metadata.json": {
+            "bytes": 558,
+            "sha256": "71c68431663ed2336cccc985a354adab6b13b4cfd6123a18a5d8d53c28268509"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/references/source-notes.md": {
+            "bytes": 13004,
+            "sha256": "6ffbfb68cc7e04923b4fbca2a73a7555b7c3b45a6f40819bb5f3213ef05cf88d"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/SKILL.md": {
+            "bytes": 5977,
+            "sha256": "03829039860ece35519f2e4c339a48b5014ec25e07a2c150a3efb6a3a4e35b4a"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/metadata.json": {
+            "bytes": 559,
+            "sha256": "80d80c15514919826fa72d6f6cf26b76db8accd85d21187fc239232ca454e7bf"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/SKILL.md": {
+            "bytes": 14703,
+            "sha256": "4ce9dcadd1ae3c7a8cc633336760d0a3c382f354f1b7028e3a10d28fa0d0c9fa"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/metadata.json": {
+            "bytes": 1403,
+            "sha256": "b6cb25879b1bca5f5af08b48776f2e664cf5dcd11a0efb1c4f56879e2afa9006"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/references/awesome-ascend-communication-tests.md": {
+            "bytes": 9388,
+            "sha256": "0cb27a3d24409171bb414d7ebf0c49ad62be5f46482f29624695496ba9eb3e63"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/SKILL.md": {
+            "bytes": 5839,
+            "sha256": "9203cc5a14a1d65df5fbf275734b93be023e06527459b2f79864f28a338cb44a"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/metadata.json": {
+            "bytes": 536,
+            "sha256": "72b41e7845915d34b09219624e82976386f3125816293540c2c38143fe6b477f"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/SKILL.md": {
+            "bytes": 6440,
+            "sha256": "8fdf46fb2743b67a66bc069d22a45305a222debfb606777bc451b4b8f31d9cc6"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/metadata.json": {
+            "bytes": 536,
+            "sha256": "be681a3206e5e183b4c9670c9d7784bcffce18c891449e5b15f94ca5f1324b9a"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/SKILL.md": {
+            "bytes": 12988,
+            "sha256": "401981cf7828edcab0c9ea882f7ecd83badfa3c2a95006e912cf99d67b3d6a12"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/metadata.json": {
+            "bytes": 760,
+            "sha256": "efd37bb23acb500dad4ec8ebe7ddfe5c3f196374052825f923ac85da8445f06b"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/SKILL.md": {
+            "bytes": 12236,
+            "sha256": "38b151bc6b4e9734c6fa41efb7a30343559400c0c75bbdd26bd8a968b8ea4592"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/metadata.json": {
+            "bytes": 557,
+            "sha256": "04d2002fc8a0a7becf86b0b73daaf2da8716ab466d7f988da43d41a10f26680e"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/SKILL.md": {
+            "bytes": 9172,
+            "sha256": "647f5e9a288f2c4e1617ba53df59d9d827df749eb4fe86182d21a9c41175caf8"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/metadata.json": {
+            "bytes": 536,
+            "sha256": "f425f7c1abbd616387273e8932c6bd02f4b1da012f099286ad4171e0f6037b22"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/SKILL.md": {
+            "bytes": 9880,
+            "sha256": "a057b7f86a24f34a8f84abbbb4910d58d4e50ff47060e1c404600017e3febe89"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/metadata.json": {
+            "bytes": 2297,
+            "sha256": "9e705625c8308013c287576b0247da6f6116d817a4a535cbecff80abd6ea35ee"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-CANN-2.0-LICENSE.md": {
+            "bytes": 5262,
+            "sha256": "89b1c114df62cfd183004f38d18d07f186436c3b37203dd69c3e84becf9d2282"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-stage-evidence.md": {
+            "bytes": 11789,
+            "sha256": "f56e4125325c9c1b9235a1d64fd732861c7686444492422daff44ba8dd648be4"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md": {
+            "bytes": 11860,
+            "sha256": "93aa1b775ced304cebfe6c4dec0503ddb2260ecb3cb01edc17ce329446c5ce2b"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/metadata.json": {
+            "bytes": 552,
+            "sha256": "f6d94f104cb70969024b2f4542078a240e200712c66279eca8b94f60bee1dd71"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md": {
+            "bytes": 8972,
+            "sha256": "0e6f094dc554abea12e07def50f20b233b47127779a8aefad595b4ab1efd24c2"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/metadata.json": {
+            "bytes": 569,
+            "sha256": "acccc8eb86de0df6caf78a9b3c6db76a2a686adc4a5ba0029a4d62d50ff9a8f8"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md": {
+            "bytes": 5989,
+            "sha256": "8428df7157af024fa17cfe0b3818ed54bd9f468306dda46eb8d4f0edf4f3f230"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/metadata.json": {
+            "bytes": 579,
+            "sha256": "c650dce88c07b636d68e063dbda4d319f6f7f3ecbcdd5e47f393cbfc3af7c71d"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/references/source-notes.md": {
+            "bytes": 11470,
+            "sha256": "80cba4d3a884e7ee2618326cf74b9f9c3a34e4d9cf443707f6bbe673f2d5ba65"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/SKILL.md": {
+            "bytes": 6159,
+            "sha256": "23e93ea7c98f99743509fdc04139d78ed2e7123c63e80b366c6beaba7eb70d0d"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/metadata.json": {
+            "bytes": 581,
+            "sha256": "97836a38465106d816a9ace894a7b277d6d330f9accade8f8493ce6f0fb57ef9"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/references/source-notes.md": {
+            "bytes": 13493,
+            "sha256": "435193590b4a2ba32e2e4c6fa9d7e8f889528d412cbbb48e35ec5433f7dab384"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/SKILL.md": {
+            "bytes": 7137,
+            "sha256": "a25738dbb3b9daa15c5827cb536a20fe13f7c8b9dd14aa3fb7cfa7e132879431"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/metadata.json": {
+            "bytes": 1007,
+            "sha256": "3abe626a6e4c73c943429f609ce2c62aa77aae7e1726475ebbbdfbe1794f6f7d"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/mechanisms.md": {
+            "bytes": 17376,
+            "sha256": "fe08b2292fa587be3874774ddf41063b84c579a0085df422ec6710d8aa90b8b7"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 4024,
+            "sha256": "371b01ee201ef51fecbfd6da892d517aaba295930b9395025f03f34e21c030a1"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/NEXT_GATE_LEADS.md": {
+            "bytes": 1623,
+            "sha256": "35b8b58ae5db02b2df51ccbe5cdf870a7383d17f345f9ee5eb7b0a4146dfc74b"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/README.md": {
+            "bytes": 2888,
+            "sha256": "e0b1ec1b39c10067f5ddce58320851ad089568c4b2296a24777389097cb5b388"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/io-reduction-can-plateau.md": {
+            "bytes": 2671,
+            "sha256": "5cc35f514224ac6e6efd2da6db221bba468b2f4db66807295e67b240bcdfdba5"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/overlap-and-fp8-counter-signals.md": {
+            "bytes": 3692,
+            "sha256": "80c8a60a6df55f3d84fab23e03567d8b38e23bcdbd63cc09881e46db1e2888ab"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/partitioning-after-fusion.md": {
+            "bytes": 3132,
+            "sha256": "4d4cabe0991eab554dcfbd4fa9c8199e61fa4767709b0c4cd646e2588bb398a2"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/knowledge/attention-cost-model-boundaries.md": {
+            "bytes": 5223,
+            "sha256": "5ed8f888443e38ccb72545b6025447aab7b92936f78c151fed680e582126239b"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/sources.md": {
+            "bytes": 9069,
+            "sha256": "53436e0de4e6dda2d29426591bf61f8526378cdc6f9ec7967b88b1bcd4c61e6f"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/host-overhead-profiling/SKILL.md": {
+            "bytes": 11199,
+            "sha256": "326e566fbe30d3ba04e4b50c3e87aad9df037f0ac4ad6fe129165ae97dc5f91d"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/host-overhead-profiling/metadata.json": {
+            "bytes": 527,
+            "sha256": "76e34c1c64aba473a9e126b032a776417caa0f913f6ac3cf4dba3a5d123b1ee0"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/SKILL.md": {
+            "bytes": 4905,
+            "sha256": "cec4e3aa1eb64c31c54b5f5ca40af2b61b7919d94be4174968581d5f752c0538"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/metadata.json": {
+            "bytes": 7453,
+            "sha256": "7520da1cc97412631a600957ecb18aa5a7a4461ea21e76c237330180193ed898"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/fp8.md": {
+            "bytes": 6287,
+            "sha256": "31d9c211c11c2224007abbdcd4059913c4a54833fe2d2f926411c5238fcd3501"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/cutlass-bsd.md": {
+            "bytes": 1547,
+            "sha256": "42fec630f410aa308f70a51a89fadcd19586fa620f9831a32bee528a9a10000e"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/deepgemm-mit.md": {
+            "bytes": 1065,
+            "sha256": "6c73a30c14a2938abaa6afe98ebb19e6d81dac5bd4ace0cb05c61c20afa37b85"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/kernelwiki-kda-mit.md": {
+            "bytes": 1063,
+            "sha256": "84aab57158a914f448b38d8be29dd842b634a08c192d844f416a256beed4ae94"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/notices.md": {
+            "bytes": 3240,
+            "sha256": "876ec016644bf5331b3f257985e126c0e61980e3bfd82569b7b0c3f6b2c6c64c"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/resources-and-scheduling.md": {
+            "bytes": 8919,
+            "sha256": "1ce25b426736072435a1ead9a2a8e1f56df5d0abb8198e8c4fdcb34321db4964"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/source-inventory.json": {
+            "bytes": 120506,
+            "sha256": "1dc09c895c970a0c9708158bcc6dd65cdb736274fd2c35d44542c70407d3849d"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/sources.md": {
+            "bytes": 9004,
+            "sha256": "6b74b9e774f2a80db0860510d9e6940d36284430f29c1e124f115a9eaf25833d"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/synchronization.md": {
+            "bytes": 6843,
+            "sha256": "4322f6b3c0e4368bf8c192b4ec752529767f50d19c0c38dad8fdf9cd781e41d9"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/SKILL.md": {
+            "bytes": 7348,
+            "sha256": "fd1f5a7e6a55714730044155f5b38c1d700383d4e16c3e91963bbfbe1c773af5"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/metadata.json": {
+            "bytes": 572,
+            "sha256": "19e40763fedc20990651bb989c32a7693306c5c54af46c3d4be77fd4f87b6c03"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/SKILL.md": {
+            "bytes": 23213,
+            "sha256": "2c807f3df2da2f533d25183de5771f3d412058d16410994292ef8f48c49f9b3a"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/metadata.json": {
+            "bytes": 1684,
+            "sha256": "9ac6af135f0b6d76a76bc201b6bbd8ac9afe5201dd1e2ea82c0ac4826de401da"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-LICENSE.md": {
+            "bytes": 11357,
+            "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-native-offload-readset.json": {
+            "bytes": 6267,
+            "sha256": "cc6c2011be746002ba3ff96ccefab1809863748ddd17eb102b75e7b0e134fe55"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-native-offload.md": {
+            "bytes": 8966,
+            "sha256": "6c2a72b5ba3708a01d229e0576570837a617e6a42dbb0b5312a26a65a6c31542"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/SKILL.md": {
+            "bytes": 16072,
+            "sha256": "36ef88826a32ac3147b869001004ded17a8d5a491e25bfe105005ba39ea9d1eb"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/metadata.json": {
+            "bytes": 539,
+            "sha256": "ae71c389dfdb1a19d267feecb4c6f1fbb4c9afad501ff6d8fab7e38b29bcc37a"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/references/deepinfer-zero-overhead-scheduling.md": {
+            "bytes": 10894,
+            "sha256": "f452bf40132c9195d739f1c86b3aa18857138c7ed0f1861af5853d93b9955924"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/os-process-placement/SKILL.md": {
+            "bytes": 10327,
+            "sha256": "2003492e4d89fde0443b180d74933e764bd1979435e109c8d26e20eb8f6e5053"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/os-process-placement/metadata.json": {
+            "bytes": 1399,
+            "sha256": "ec02995c075f0a540921f57b4846f2cafe334eb90f44389f17c3c50457c3e9c5"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/os-process-placement/references/llama-numa-thread-readset.md": {
+            "bytes": 6617,
+            "sha256": "748b254b3b043a5ddb0f480eb4a566854d27eedd7b111499863e9bf0a5a6951e"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md": {
+            "bytes": 8813,
+            "sha256": "63b883c10a8d3e7d918abfc7175b72e2b401ff1e02697f94b8382ec1c8f369b7"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/metadata.json": {
+            "bytes": 558,
+            "sha256": "0a41c7b00978332b54ee1e4a7c1ade88ecde3e961128f24d1b4f7708caff5ff5"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md": {
+            "bytes": 6402,
+            "sha256": "0a5f48d9352c5bf460775ef47f9136829978811f4a16e84983e299782cae305d"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-serving-preemption/metadata.json": {
+            "bytes": 537,
+            "sha256": "77ce5ba16245cbd80d46d0abdf80aaf26fc0924fa2d2a5624ff430de68a1daf0"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-serving-preemption/references/source-notes.md": {
+            "bytes": 30405,
+            "sha256": "f29fcecfba0490008a354c11b42403983fe165e56ff907ad61cc6040d2b6407f"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/SKILL.md": {
+            "bytes": 3181,
+            "sha256": "1e64f28ea71672fb53c56f6108b87864738c6d03d6dcd32eedaef9e0625e96e3"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/metadata.json": {
+            "bytes": 478,
+            "sha256": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/references/source-notes.md": {
+            "bytes": 10548,
+            "sha256": "61b1609f71c37b61d136b8c4d39f793de87cffb1cf31c8ec737a1c3fa6fb1b4e"
+          },
+          "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/SKILL.md": {
+            "bytes": 3884,
+            "sha256": "ac773c6f1c110a664dfca28a3ae9129b10c07614e60ca3640dea776220e7b352"
+          },
+          "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/metadata.json": {
+            "bytes": 648,
+            "sha256": "261ce9f644ea6eebb90c830f41fb8b320d2b401456e5bd8d494095a44b8620d0"
+          },
+          "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/references/source-notes.md": {
+            "bytes": 8415,
+            "sha256": "9696d3c0ef96b6611787446bd6f96d4d6255ca8d107105de8c1d9af83edb9fa5"
+          },
+          "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/SKILL.md": {
+            "bytes": 6416,
+            "sha256": "6b7db1eb2780dec27858cb7aa0a3c093c5b5dfa4c30261aebfeaa255213ff688"
+          },
+          "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/metadata.json": {
+            "bytes": 582,
+            "sha256": "4dc8659806539b4784ea9348a01d4194ac70d0bb297a916f90ff6419efa5c328"
+          },
+          "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/references/dlrover-persistence-replicas.md": {
+            "bytes": 8398,
+            "sha256": "055f4357f79026fcbcc766bbe77db610376ec353623f8898f5d14a30ee69e7bc"
+          },
+          "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/references/sources.md": {
+            "bytes": 7984,
+            "sha256": "9425b1aab963a2c8d35f3db897bd15ee69699bf5b0d0c8425c3c8bba3e9a0233"
+          },
+          "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/SKILL.md": {
+            "bytes": 3068,
+            "sha256": "25d04aed9782b2a8ff1186828becdc7fce93c4418039ceaef9b8dcc5ec9bcd24"
+          },
+          "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/metadata.json": {
+            "bytes": 478,
+            "sha256": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7"
+          },
+          "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/references/source-notes.md": {
+            "bytes": 8521,
+            "sha256": "9f9184d46b19312138a6e805e439e43dc0594a0cc6b46675255a582dc946ca5b"
+          },
+          "src/inferrsi/data/skill_library/methods/dense-performance-evidence/SKILL.md": {
+            "bytes": 8305,
+            "sha256": "53cc1867a13f54d1006d50517dea8848869ca50df3f0cd610b2424cc8d18c933"
+          },
+          "src/inferrsi/data/skill_library/methods/dense-performance-evidence/metadata.json": {
+            "bytes": 707,
+            "sha256": "a0d1e3b17140295510a764045d1b222e19be3a3406e6d1e7dddaad49ff474490"
+          },
+          "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/SKILL.md": {
+            "bytes": 6667,
+            "sha256": "82d0099d3c33706f46c657d4fa7f76d3d486e019b36760d1de001921328a60cc"
+          },
+          "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/metadata.json": {
+            "bytes": 547,
+            "sha256": "37379a4d4a058ca46da2b20a920dc01be9ec11e557c32b2332702ec76027f4b5"
+          },
+          "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/references/source-notes.md": {
+            "bytes": 44796,
+            "sha256": "ea416831fd7d0874e06d0a76fb0c7196df53072b717673a0cdf635db6bfd917b"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-benchmark-design/SKILL.md": {
+            "bytes": 8974,
+            "sha256": "19ebc1f567aadf49f7a534a18fdb629c72c0e4ce043d360983136491cb72bfab"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-benchmark-design/metadata.json": {
+            "bytes": 5923,
+            "sha256": "18310402a8ade03fd48ac687f79ba9d447871e90337de84c0e1bc15c0b827306"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-benchmark-design/references/benchmark-commands.md": {
+            "bytes": 8466,
+            "sha256": "0898424e0bff3a6246e1eeeda6abd6c5bd974c047fe32f3bbd5ab9d31ce17cd3"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-benchmark-design/references/task-benchmark-selection.md": {
+            "bytes": 25086,
+            "sha256": "0d4bb080a754e26846e81680023db0dff5e6a4c96be4327c411c3b72ac4ef8e2"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/SKILL.md": {
+            "bytes": 6644,
+            "sha256": "89a6a82c692f7f5192f4d7f4c3a5ff08a4df3b1c4bdaa0aafb7e04292d8aa5b4"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/admission-drain-occupancy.json": {
+            "bytes": 1354,
+            "sha256": "3319e23bbe2bd40e2d45fb89dc87bdc92d4df2b5ddd8086fcf84813b43885b7c"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/admission-drain-occupancy.md": {
+            "bytes": 3134,
+            "sha256": "e88bbac5d253a8e7e3a3d5e21047f23f2970c11ac23887d4303ab94fca52dd27"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/drama3-rejected-directions.json": {
+            "bytes": 1459,
+            "sha256": "2cc0a526fe0509551607a03e7626dd3d7844e643c4afb194837f75cddead8b12"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/drama3-rejected-directions.md": {
+            "bytes": 3557,
+            "sha256": "6f143bc363a3290fc008c08e57cb075f1776d1c8cca8a6a59297d5b1fbbf61bf"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-ascend-rejected-directions.json": {
+            "bytes": 1293,
+            "sha256": "b3d538f79c56724ccb2895cbf50cf2e6b0fec3979be62a684a8230cd932dff55"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-ascend-rejected-directions.md": {
+            "bytes": 2564,
+            "sha256": "be396983812262515899a3e2ba4f01b23ff92e1b836881882c7580372b8ae552"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-b300-rejected-directions.json": {
+            "bytes": 1196,
+            "sha256": "146ea1564a46ed03092c7dc5796b318fc05b04b1821105ee00eb7fa7fa1333e2"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-b300-rejected-directions.md": {
+            "bytes": 2656,
+            "sha256": "cce2930f544a0f78f21797fba4d8ded298e4a81f76c34d7977b0e31d361bc738"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/metadata.json": {
+            "bytes": 535,
+            "sha256": "fb55067fb45970319787068a3206110641cceaf0ccbabd706be6d0b3d2afc2ed"
+          },
+          "src/inferrsi/data/skill_library/methods/kernel-writing-guide/SKILL.md": {
+            "bytes": 10029,
+            "sha256": "01fb1712d7eb988c5e412f02340d2e48b4f8564fcf4ab232dc5897869d0d50a0"
+          },
+          "src/inferrsi/data/skill_library/methods/kernel-writing-guide/metadata.json": {
+            "bytes": 653,
+            "sha256": "0189d184337662056fc6952c7beffc1bf6154c4fa173ae3e3e463bd22d70f36c"
+          },
+          "src/inferrsi/data/skill_library/methods/kernel-writing-guide/references/akg-affine-tail-guards.md": {
+            "bytes": 11941,
+            "sha256": "113deb2db10b171db3425bb3f1792bee763706b794c8cc2e222b4cc0d0114118"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/SKILL.md": {
+            "bytes": 12914,
+            "sha256": "882e7850b9b49ad692a2d11618701afe0f4e83727ea78df114be2029923c725a"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/metadata.json": {
+            "bytes": 1403,
+            "sha256": "9f1fd8a2af52bc964021e21f16ecbdb891e08f2057370e700d67f86605d333db"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 3597,
+            "sha256": "157e74ba180426b279b548a729769f152bf18f9a78ffca4ebfcaf26b7d0c5292"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/MECHANISMS.md": {
+            "bytes": 5875,
+            "sha256": "18bb2f5d7e52773c516420a142edfb531d0b9c822a2934d60f3234ac1cb9d40e"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/NEXT_GATE_LEADS.md": {
+            "bytes": 941,
+            "sha256": "16a264c7159f9c2c700ce8ba46e63f3e636b277c987da4970f6d2fa085fcb053"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/knowledge/adaptation-interfaces.md": {
+            "bytes": 3893,
+            "sha256": "5e761957cb5ecd93926e6ffc13deee7c830c6e8f0ee3c3846e9dd3b50dc74446"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/references/source-notes.md": {
+            "bytes": 13370,
+            "sha256": "54d83c3713045d59a48821bcf93598374ba4a630e3891ad39498f919317c4fb8"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/updates/optimization-search-policy.md": {
+            "bytes": 4238,
+            "sha256": "d3db44fa9a6831b86a02c638b323ed943e69fe256866213f10d68c72d762879e"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/SKILL.md": {
+            "bytes": 15324,
+            "sha256": "c5f78961c23e3c8d659f5675b522518ab06e1f2a8ea8bc74c5d428ffb41693d5"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/metadata.json": {
+            "bytes": 3669,
+            "sha256": "60b5b24ead590b0fb222398fba47822bf6723690da683e256a450bb5c0eeb806"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 7349,
+            "sha256": "083e462e0a204dce599362d46712fcfb392d053c5cda56216e22e28d4b8fa0fa"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/MECHANISMS.md": {
+            "bytes": 4003,
+            "sha256": "742c249082ceca230326f4407a40130fe624553b796ae5a157a8a4c15a7e1398"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/README.md": {
+            "bytes": 1204,
+            "sha256": "17b2d1c7c14124d59ac8a0d5ad4200ad2803d991e784a1a98cd97188cc364999"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/READSET.md": {
+            "bytes": 3937,
+            "sha256": "66459ce393858b2b2160f1e2973373bfc4b9902a12644631934b4ebe3140296a"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/STATIC_CHECKS.md": {
+            "bytes": 1111,
+            "sha256": "cb0aa7d148352ebb7b1678fc42a557a099575054716a341dfa25b6a998029064"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/cases/decision-cases.md": {
+            "bytes": 5491,
+            "sha256": "781a675a6533e0a18c044aaaee0d010da7013b6504654b8b8292a35176db801e"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/govern-optimization-experiments/procedure.md": {
+            "bytes": 3823,
+            "sha256": "ce036a4d8a0908da204515c47e2ff9623548d09de264fa65bd58fedce52a78a0"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/govern-optimization-experiments/references/source-notes.md": {
+            "bytes": 21659,
+            "sha256": "f9a8254e45473ddb2b73e576957b6d3db83c62b99dff1922f4d1e606e569559e"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/knowledge/bounds-predictions-and-decisions.md": {
+            "bytes": 5297,
+            "sha256": "5780d6eb5257612ec45deb94f85f334a6ecabf7e466d1678706041f6549bda79"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/redirect-cross-layer-optimization/procedure.md": {
+            "bytes": 3648,
+            "sha256": "6593a63fce9c474b515663a560c94b7df048dc5f6e77a95b8e54e4a64cd10a09"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/redirect-cross-layer-optimization/references/source-notes.md": {
+            "bytes": 26469,
+            "sha256": "6ff3a7bc31f1172db92aedaf7df1f1aac688bf769227b583a4cd82bfad52da67"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/agentic-kernel.md": {
+            "bytes": 41353,
+            "sha256": "7190573b85f94a4391bb0d887ccbd37a31b3f08bb0531114ad7afd72fe1567a7"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/flashvector.md": {
+            "bytes": 9830,
+            "sha256": "3926369acd88ff3c370bee5718e74dd15f3219c3e4f6311095e30e917bffa4cf"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/inferencebench.md": {
+            "bytes": 9294,
+            "sha256": "21f47111197a0b7f8d6384ea918d139dc8318a62ca3313ce060048271f50b639"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/prior-coverage.md": {
+            "bytes": 3254,
+            "sha256": "35d47ac078591221fcb666219bcd824f9400fc17aafecdc0499d8336f7178238"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/vidur.md": {
+            "bytes": 7736,
+            "sha256": "d38602af5219f5feeb49e6cd6bf53d66098c77fad141dae676f1d8a6743cf25d"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/roofline-contract.md": {
+            "bytes": 7091,
+            "sha256": "419ec4bcaec45ac579a133bce083d4f48779f26e664d7a86f720fa65f343b1d1"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/roofline-example.json": {
+            "bytes": 2015,
+            "sha256": "597cd3205df043cff3ae9d15924c77934e8d8eb300c00c4840f6bb2921f90faf"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/scripts/roofline.py": {
+            "bytes": 9004,
+            "sha256": "524de48a6764401695fe1444f1fa3d38f363069b0cd2fba4715a86ca4b28a5d4"
+          },
+          "src/inferrsi/data/skill_library/methods/plan-activation-residency/SKILL.md": {
+            "bytes": 3507,
+            "sha256": "542d8a1e6963f5019986e1fbd79d3a247a045a51487c09313026087ac457c15e"
+          },
+          "src/inferrsi/data/skill_library/methods/plan-activation-residency/metadata.json": {
+            "bytes": 478,
+            "sha256": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7"
+          },
+          "src/inferrsi/data/skill_library/methods/plan-activation-residency/references/source-notes.md": {
+            "bytes": 13907,
+            "sha256": "5ff52655cfd06fcf970ed2a7d4d4fe1c31115b102af2d41497de0c74848ed2b3"
+          },
+          "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/SKILL.md": {
+            "bytes": 13800,
+            "sha256": "7e20d3390f2501263e5e298e026c47cac16c234f2397833131e15ae5d34c0b97"
+          },
+          "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/metadata.json": {
+            "bytes": 1903,
+            "sha256": "d0e4d6950250e4cec43d82e23324c9301a925d4c22be30b4d2713d790cb6504b"
+          },
+          "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/references/capability-presence-audit.md": {
+            "bytes": 11461,
+            "sha256": "c21eb4a6aabe811a49a6540a34c54dbcfdcf7efae3052a79b54034c7838026a5"
+          },
+          "src/inferrsi/data/skill_library/methods/skill-evolution-playbook/SKILL.md": {
+            "bytes": 9448,
+            "sha256": "9cd37f21fe0d862069e4798edca8a04d59f8b840e55d521901594a63dddd9ba3"
+          },
+          "src/inferrsi/data/skill_library/methods/skill-evolution-playbook/metadata.json": {
+            "bytes": 527,
+            "sha256": "92b36338b9257e2e41b95e4370694d7de52ed84bc7971056b3bfced07b0993d8"
+          },
+          "src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md": {
+            "bytes": 6865,
+            "sha256": "64bd2c6b795d966045d1d3cca55976d3d49b08b06df7c7e06a70c2824cf55b7e"
+          },
+          "src/inferrsi/data/skill_library/methods/training-input-pipeline/metadata.json": {
+            "bytes": 587,
+            "sha256": "80659099fce6a6ec05e9084b75452c877a5812e30f60a78f31ef8c7efcdc1bac"
+          },
+          "src/inferrsi/data/skill_library/methods/training-input-pipeline/references/sources.md": {
+            "bytes": 2637,
+            "sha256": "60e0f3e88655d8e1d5eb4797f39ca4370deda01614f265b8d2ea4559d8ac020f"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/SKILL.md": {
+            "bytes": 7200,
+            "sha256": "bdfa5a4ef37d9f303c3a23cd6f98c21137b97ac5ed58e0ed5fac6d71d33e52ef"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/metadata.json": {
+            "bytes": 976,
+            "sha256": "d42b614af9cee6ee0d4a25dc5d57c7b98438feed1e807a7c9f1755109a86e9cf"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/mindspeed-historical.md": {
+            "bytes": 3384,
+            "sha256": "eb97f935ed30517a8fc9edec5e9cddc04f5d679d156c16d75bd29c534593360a"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 7368,
+            "sha256": "c55c6ad2131ef788866411dcc7056f221420ebff30c95c6bc919b09a4818cb73"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/cases/paper-grounded-cases.md": {
+            "bytes": 5649,
+            "sha256": "19bd25b1a944adb2351ebc6b1d795150ea033041a44f6cc261e881677812a101"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/knowledge/offload-delta.md": {
+            "bytes": 2026,
+            "sha256": "f8740fc5c8bbbe5ac13ff1cfec7ca499569d0b46e8cd56cf1f5a44c3a5e72242"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/notes/training-memory-scheduling.md": {
+            "bytes": 18660,
+            "sha256": "9aa6d3e01452dff95c8a8940ff62ead8c146ce215c9175b78c9f2ef508a7b2de"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/references/source-readset.md": {
+            "bytes": 22946,
+            "sha256": "2a5c166767f56fa47e7eba7386f1f36da8cb73702e5b5bbe14d6a456ca9ebfae"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/sources.md": {
+            "bytes": 2752,
+            "sha256": "6d248f58af8b350fc66c22898988d6f451ac60d03bcd805cd6e84df5e55174ac"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/stas-decision-notes.md": {
+            "bytes": 4086,
+            "sha256": "f1a993e6c4ca9a035a2effcaf5500081f227b0cc56cc885a6b22fda0041fc744"
+          },
+          "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md": {
+            "bytes": 25969,
+            "sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213"
+          },
+          "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json": {
+            "bytes": 2418,
+            "sha256": "798cbff6f10b7c24b188080ebf233b188b0e98ef2c6fdb5e00e1fe3eff727c0f"
+          },
+          "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md": {
+            "bytes": 8656,
+            "sha256": "5a39f14b102b0cee304541e011aa31e44e6b13527d1bee9987c3f3b66c064d3d"
+          },
+          "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md": {
+            "bytes": 6392,
+            "sha256": "6c640f28ee3f0253a17b81461c865cb9f947367002cb75bf72550c4572d02e97"
+          },
+          "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/SKILL.md": {
+            "bytes": 7181,
+            "sha256": "a46110891eca860915e314d6921a22856d631453b22b17837fe30262d6780535"
+          },
+          "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/metadata.json": {
+            "bytes": 2175,
+            "sha256": "4e6100dfb9fc7bc00b19915f4fd38ad509704f2a609c658adf4d632d4e05b6c1"
+          },
+          "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/references/afd-plugin-LICENSE.md": {
+            "bytes": 12747,
+            "sha256": "0e8566ea362276878e1f38f97da49964dbea301c15d4aa9d727f97a4c39fd943"
+          },
+          "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/references/source-notes.md": {
+            "bytes": 17240,
+            "sha256": "5073d87450cb60d1b245308dff8807e73749c07de976d50f2fa26694fad81a72"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/LICENSE": {
+            "bytes": 35149,
+            "sha256": "3972dc9744f6499f0f9b2dbf76696f2ae7ad8af9b23dde66d6af86c9dfb36986"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/SKILL.md": {
+            "bytes": 4507,
+            "sha256": "c993bd6ff8e9b5f82f39f175ec8242341136e4ee0808af414e1d94ad3e196c5e"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/metadata.json": {
+            "bytes": 2032,
+            "sha256": "74742851a002a51c208362cdcd2a59fdab5d1734d8492b1f93fbec3cf21e0841"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/references/graph-cache-invalidation.md": {
+            "bytes": 7679,
+            "sha256": "614c9a4e6943d24d7d7be7d632682ce8b1dcc239124a37d9dc1d487113772eac"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/references/model-residency.md": {
+            "bytes": 7191,
+            "sha256": "014c68386e7cdfdb7b4f566836e8b2ff61bd5a5a506e7b4723f83fe80c5d4de7"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/references/repo-provenance.md": {
+            "bytes": 5648,
+            "sha256": "ca996ef515af42a3ce304747fa3f12a31c00533deaace5ba2ce21aefa0a62e2e"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/references/validation-boundaries.md": {
+            "bytes": 3523,
+            "sha256": "1913dedbeb558d872d6a0e015cc11d822910c3cb259a236db29fb60ad8a64263"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md": {
+            "bytes": 12294,
+            "sha256": "4703beb6e5194ef2ff2c0b5ec69f64e35417a4bc0dfb50a10c80a920b77ef3b5"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/metadata.json": {
+            "bytes": 2333,
+            "sha256": "014be24d0e669f4acbc0717dc323eec784026cf185186619d5a19d6d84496131"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/source-notes.md": {
+            "bytes": 12036,
+            "sha256": "fd917bd5ac80bee9c9f4525f8e0300f6d67dc144e2a16632af42a67f3d53d098"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/validation.md": {
+            "bytes": 5402,
+            "sha256": "60e71e70c5519bcdf6c171eec2be2f342e8801bedd746d6005f11adc646dfaff"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/SKILL.md": {
+            "bytes": 11202,
+            "sha256": "f157562d1edc905cb4fc07246567e3c5b844efb9fbf5fe39f0d40553a1b1a307"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/metadata.json": {
+            "bytes": 2156,
+            "sha256": "4d68bf4c1767eaee12ade963bbd551d0e61d1bcd81a984bb02fd3aae46965919"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/references/source-notes.md": {
+            "bytes": 11633,
+            "sha256": "6fa955022f16cd598f44fd6521901d1cb868b06cec545cabe2b82769a641b5d6"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/references/validation.md": {
+            "bytes": 7196,
+            "sha256": "673d93f6700052974b340fab25b3438e0e6344960569ac9abec7d7720f662288"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/LICENSE": {
+            "bytes": 1077,
+            "sha256": "663e51edcd9a3a8f858018ce6924ad2895a24c269c26c39ad75325bcd0f7e5bc"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/SKILL.md": {
+            "bytes": 5432,
+            "sha256": "fdb4d9483c6efa836f9e59617da605fa9c2186e11c2041b05d5c81850ba0475d"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/metadata.json": {
+            "bytes": 4923,
+            "sha256": "fe626a77af3f084f36fd73ffdb6f67976e7dceeafb94ba22078ef03c3a96cfae"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/THIRD_PARTY_NOTICES.md": {
+            "bytes": 1774,
+            "sha256": "a0d77db7d2aca0eda811ab3c386cc32b3143f2c2609a042cf4642954f0f6fb09"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/minimax-h3-apple.md": {
+            "bytes": 12059,
+            "sha256": "57ce01754ba96c378f13a91305ef3d418b926f99b2753f0af99315572bce063a"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/minimax-h3-weight-inventory.json": {
+            "bytes": 10664,
+            "sha256": "fbaacaeec340d3ce893556ee864405a0746af4d13530a38cb3d5aa9d60929271"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/phase-residency.md": {
+            "bytes": 9408,
+            "sha256": "2836a550eb2be7f5a862f51ec7b135ebd9715bf3c4faa84a7ecbc9d9c11f5d8d"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/precision-and-streaming.md": {
+            "bytes": 9440,
+            "sha256": "7e166c89193f4c7aed852ec95346a2de2b5008dac002e2e1862be1b3e57f6093"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/source-identity.json": {
+            "bytes": 7747,
+            "sha256": "730f8dd23648a3f8f47b9726cc613d893a79d1817d515eec609a9b06d8a9ff90"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/source-notes.md": {
+            "bytes": 9448,
+            "sha256": "7304bf9369efc157dc7aa40686dd828cc14db45c49442e908b3d3cc4b329eb86"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/validation.md": {
+            "bytes": 7880,
+            "sha256": "0754873a758e048480938314894e25b504ae0c269549312e7f8e6020d83fe294"
+          },
+          "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/SKILL.md": {
+            "bytes": 9338,
+            "sha256": "95e81b6bc096a954fa73139ea5bc001d1efa6c99747161da62f012b494e5c013"
+          },
+          "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/metadata.json": {
+            "bytes": 1718,
+            "sha256": "1d30b31dc8352fc6a8d7b8abbef8118c840e47e5b90795dae0c758bf9717278c"
+          },
+          "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/references/source-notes.md": {
+            "bytes": 15234,
+            "sha256": "3d12a401e9decb55a2f4f3b2bfa0d360ef5110f63a0469dfe285654e78054190"
+          },
+          "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md": {
+            "bytes": 10812,
+            "sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27"
+          },
+          "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/metadata.json": {
+            "bytes": 572,
+            "sha256": "c221189b8d85fac3bd41ef1c7cd560f1aebc8cfa0381c182d1d571a7169be829"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/SKILL.md": {
+            "bytes": 3788,
+            "sha256": "bf1177cb9985c12a9bf11edb64771018e1e314c4713920de8fd4a989ff70115f"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/metadata.json": {
+            "bytes": 1299,
+            "sha256": "d86cecb615e93ba0f6be09d1ead102cd5a17068c07e1be2549ac5cd8f7519279"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/cache-stop.md": {
+            "bytes": 2026,
+            "sha256": "bba6b5c73261065309cbcf11b1e591eef48916f0c79ceaaf74d98a9fa96b0992"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/generation-batch.md": {
+            "bytes": 1733,
+            "sha256": "f16d3c57d45e63bcaffd3871aa41c76972ee11f7dbf1e12b45d9b170303cad9b"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/source-map.md": {
+            "bytes": 6358,
+            "sha256": "91d26181f246307f2113f94b12ac9587d689b654dcd216e9f14b11cf5ebb650c"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/validation.md": {
+            "bytes": 1665,
+            "sha256": "73b5dd347017baba519ecc85415aead9993c1751d116b9ba4f07689f58c5f783"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/LICENSE": {
+            "bytes": 11352,
+            "sha256": "a54c8553df5d4313d293d1aef3bf2f13952bda8b7c7a1fdd729b2bbe27ee6015"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/SKILL.md": {
+            "bytes": 2304,
+            "sha256": "d75f5b2881b240ec2d9d4f5cc8e3f7d53e4378cb01732a2fb9699da2901ba38c"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/references/repo-provenance.md": {
+            "bytes": 1778,
+            "sha256": "5e8dbe168a8885e209d4b4532944ffa7fead658b23fd6588054d57b22476f63b"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/references/repo-routing-metadata.json": {
+            "bytes": 511,
+            "sha256": "6820707cb9bd877d1635928cf5a0efe38e4fe37dc04036dc70d53e479ed9bf7c"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/references/snapshot.md": {
+            "bytes": 3488,
+            "sha256": "0fb5d776349de845dc314f16ce73a06c9caf5db84dc1c293327d3b3ca1337c80"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-architecture/SKILL.md": {
+            "bytes": 2355,
+            "sha256": "766b2d66d2056dd779abd9e492cb3af6e25899298f44964f565509f130a7cb24"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-architecture/references/runtime.md": {
+            "bytes": 7734,
+            "sha256": "36bba9874fe054acc5217f9c265f5ae309c83943e0d46228370ae1bcdc69421e"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-extension/SKILL.md": {
+            "bytes": 2193,
+            "sha256": "9af38bca840d0004fc2ed049141476fc40fd40dcfa843f38a404a7a314c19425"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-extension/references/integration.md": {
+            "bytes": 5775,
+            "sha256": "6ce9383ca17bacff60a180f8fb8c43ea47eaa37117fb8da6051ad86135d228ef"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-serving/SKILL.md": {
+            "bytes": 2335,
+            "sha256": "990b6bdebfe22612de49dbebcb16810e8abef94d70691aa883b7937c85210eb6"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-serving/references/recipes-and-tuning.md": {
+            "bytes": 8918,
+            "sha256": "94afb762bbe185f58606154b3f0a5af56e7089cb84ea94e26cfcb9f93c95439f"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/LICENSE": {
+            "bytes": 11346,
+            "sha256": "1495e1e757ef4d0925a2350563cf5754bb23c51701a8ec4fb3c5cdcbedae6747"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/SKILL.md": {
+            "bytes": 5466,
+            "sha256": "ae72d1f0cc750699bf79d7a138a97219209bca7b7e5f19cd427d306245eaf841"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/LICENSE": {
+            "bytes": 11346,
+            "sha256": "1495e1e757ef4d0925a2350563cf5754bb23c51701a8ec4fb3c5cdcbedae6747"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/SKILL.md": {
+            "bytes": 5738,
+            "sha256": "b0a4ba4751087126143ef86983798deb0ed00e24b6ff06bf6ef76a9df6e274f2"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/metadata.json": {
+            "bytes": 2844,
+            "sha256": "6ed57dfb4f25ecb34b81f2188b055e05d172ba2581e6a3b605b814b2f733a4d6"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/references/source-notes.md": {
+            "bytes": 19512,
+            "sha256": "d0fcfd1177ae2f351041c5b2abba7f5121cec133006d9e4dca8f74ddb567332d"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/architecture-and-memory.md": {
+            "bytes": 7735,
+            "sha256": "798e747022b7ec135cc4861c2c9ac44fb14b503bb77a0068ee7fbea494b764a7"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/native-candidates.md": {
+            "bytes": 3961,
+            "sha256": "abafcc9a34df196a6365145554887b8d37afb36c6e236c92e50a4a06c8a4ec93"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/parallelism-and-extension.md": {
+            "bytes": 6324,
+            "sha256": "e3be7f4b51264ac68682a64d264e4d96857d5bb9f9d99cae5f327c56984a9e6b"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/repo-provenance.md": {
+            "bytes": 2274,
+            "sha256": "534d95ae8d0b29bcbc8f6707b3d38fb964e31f5b66ba13517a9a6b8f309bbb38"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/repo-routing-metadata.json": {
+            "bytes": 625,
+            "sha256": "b2b879c1cc42ddc2a5b95ae24115eb891562fa9a6e96d151f4ddf6f7719fe73c"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/source-map.md": {
+            "bytes": 8945,
+            "sha256": "c2adad64ceb0c305cdc4859c03012b3e0a4ab26cd0cb289bbd76e192b0cc3794"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/tuning-and-validation.md": {
+            "bytes": 10832,
+            "sha256": "49b3343b3292cc098a787de9d24d968b444d0ef4f73fedb5d8183390e8f39c9c"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-parallel-extension/SKILL.md": {
+            "bytes": 2316,
+            "sha256": "18bd54ebdd629e1ba6dfa589f391c7cc3c5355785c2ae82f741b64ee8022464f"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-runtime-memory/SKILL.md": {
+            "bytes": 2154,
+            "sha256": "33b1a65aba00fca819d2b6cc42202e7708e9155e4d8f4373a2387f4ee2f2fa3e"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-tuning-validation/SKILL.md": {
+            "bytes": 2018,
+            "sha256": "744b2ef9122101d3f50cbbb40fe020e2b3d3888adbe2d86041041e7c7ce424fc"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE": {
+            "bytes": 1062,
+            "sha256": "b53fa08f515cb5a6fff7b9fd8fcd0961a4b80df29d74372d25e1f9171aa042ee"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md": {
+            "bytes": 5504,
+            "sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json": {
+            "bytes": 2202,
+            "sha256": "f7b7581a3f44ba9774480c559408fc66d6a4d6d1dbbeabe4529f1eb7d4dd5ad6"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md": {
+            "bytes": 8136,
+            "sha256": "7b537cb972347dd68e303960f142892346d8b19b884e4d08016ea36a23bdaa7a"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json": {
+            "bytes": 3617,
+            "sha256": "af4830370b6eddc2d176a9ae14da4ab56a90b0fe048a01a89ee051df4365d8e9"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md": {
+            "bytes": 6589,
+            "sha256": "b67affbd7d237d82b7f51be49f1537e9820fe56ec70a71ee76809f9e599ca693"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md": {
+            "bytes": 3394,
+            "sha256": "5dcf5fa7bf734fbe13214f57d2dfb40ff8215d9cd34e072a9e615bf103b82fc2"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/LICENSE": {
+            "bytes": 1093,
+            "sha256": "949b9b0aada7476006469e58ee4ffb0e0f68c2d9137e90a356757b61a09d37f0"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/SKILL.md": {
+            "bytes": 2813,
+            "sha256": "0e5daeb6409c916be45ecd7ce2531ea5f1c60a4a53e2b0ed25dc1e2cdc5c17ed"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/design-space-analysis.md": {
+            "bytes": 27736,
+            "sha256": "11ecafe71530ddd4839268ea1b24b3361a3895012a705859ca06dae29fa72630"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/repo-provenance.md": {
+            "bytes": 2913,
+            "sha256": "05f17d0ae996ad6c12314bdd29f3427574f6aa611ea60effd18598a0f386638b"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/repo-routing-metadata.json": {
+            "bytes": 332,
+            "sha256": "a10ba49b7e6565109b566c105e17551d042075a1ff649c6f371134e3b7402c74"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/trial-format.md": {
+            "bytes": 2291,
+            "sha256": "92d8cf38549268ae4eef382890662d03e18a7dff858ad45ee5eb6f47eb39c8c0"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/troubleshooting.md": {
+            "bytes": 2079,
+            "sha256": "7301cfa5a5f6cfbf9576c3845a96c6b4de8f8a6509f2330970a8e2e4669b3d5a"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/scripts/summarize_trials.py": {
+            "bytes": 5170,
+            "sha256": "6a243a08da15659ced6519f49f52a49a675bd9e952740c3ab03dd21964eefbe2"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/design-space/SKILL.md": {
+            "bytes": 1826,
+            "sha256": "857bea3c64db45dc035728a41523142ac65778b41f63092148b8ab936f971d76"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/design-space/references/design-space.md": {
+            "bytes": 24231,
+            "sha256": "0a428dcd12d5290b554e9da6795a110f94d98f128771602e1ea5f608048b135e"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/tune-execution/SKILL.md": {
+            "bytes": 1145,
+            "sha256": "cdf51c537e7237296d52ef6cda9f2e4c9d9ce3fc87265a369cd9eb7497c53512"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/tune-execution/references/tuning.md": {
+            "bytes": 11310,
+            "sha256": "7b69aa7872d60dd3fb91e41b77a632572070dcf80dd5b5f6dc1722d8540177b1"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/validate-engine/SKILL.md": {
+            "bytes": 4613,
+            "sha256": "90bfd975e25f18895ce0e85f6f35d24149ce026305a4d479aea4924510053460"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/validate-engine/references/validation.md": {
+            "bytes": 16895,
+            "sha256": "dae846f4c47c68731b96ec1d50cb008c03871c72078f1c5b0f4a9423be9ba397"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/LICENSE": {
+            "bytes": 11357,
+            "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/SKILL.md": {
+            "bytes": 4945,
+            "sha256": "df3e63dc28de2dac2aa92824fd88719159138650d96487332e0effca3f0e2a80"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/references/evidence.md": {
+            "bytes": 4689,
+            "sha256": "ae0d51ba56b34ed81fba34de8fd3685bebf2f160b188fa53719936002881d220"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/references/repo-provenance.md": {
+            "bytes": 2221,
+            "sha256": "43bbfe431434ca4ad6157ba35cbb9ce2d89e4065f427abfdd5f372f51ba0c9c9"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/references/repo-routing-metadata.json": {
+            "bytes": 342,
+            "sha256": "7784b4d98d096718c4f4cc42ce86d5ec113c469844e12de4690b0fccca59d329"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-architecture/SKILL.md": {
+            "bytes": 1336,
+            "sha256": "fb3ff50a7b9122a2545a97ef5179011237c4c4ca214dde7236ad7bf79a7c8623"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-architecture/references/architecture.md": {
+            "bytes": 5600,
+            "sha256": "db845eb4207ddeedb8ae381f1b91704a42cb53ba0ee6c7ee581510ba1a412190"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-serving/SKILL.md": {
+            "bytes": 1343,
+            "sha256": "8ab3209b838635395b22deca6aa879172016531d45c2374eb8e498e7f71bb40f"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-serving/references/serving.md": {
+            "bytes": 5049,
+            "sha256": "9bd5eff300d9755137a7600c8f7a7753bb22d09c8853e08062d7e535662ba8a3"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/SKILL.md": {
+            "bytes": 2227,
+            "sha256": "93c83bc4b799b32f50ff54b870cbbde050a790af16d9b7b062039c4987c80148"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/references/abort-acknowledgment.md": {
+            "bytes": 8547,
+            "sha256": "e95ae5b2cc410917ddf5e9066890d0bef9264104529bdca5ac12bdf56fe80ae3"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/references/validation.md": {
+            "bytes": 4632,
+            "sha256": "1542061f9c76fff2acb6529a58238301484b0e23ec0c697d2fdf129e200ea973"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/LICENSE": {
+            "bytes": 11324,
+            "sha256": "50e6751797c50dedd75ef1b8a0d9e42f5f8472e9fbce91f34718e9f97b0c780a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/SKILL.md": {
+            "bytes": 2863,
+            "sha256": "38e522b8abeb0dc140a2194340432ed41dc8b569d33b63d5d4c59aa1092cd2d1"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/SKILL.md": {
+            "bytes": 6664,
+            "sha256": "313733a65cc48442b18dda1c4665139a689963a90e1032a2235bc6f40e3a8e06"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/metadata.json": {
+            "bytes": 781,
+            "sha256": "b43fc3a7ccae59c664fcdf9ff77962edcc07c2a3df8b59324db551642bd23e23"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/references/source-notes.md": {
+            "bytes": 13853,
+            "sha256": "3665b034594fc8cc767c408002f4e16f8803369242405027c7f37ce6c9ecebe9"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/SKILL.md": {
+            "bytes": 6565,
+            "sha256": "5fb9ea1d1373e110d9b506f69963c4857338b7dbf5e38ec9a49b59a0ff683ed0"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/metadata.json": {
+            "bytes": 1372,
+            "sha256": "78459416734f61a8faa819d9a1a3ad7b9a985e8bf8e149813c604d15be6bb92f"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 9006,
+            "sha256": "82b24be82451ee121892f61575d0cf4ecb9920813a25ccb26cce2b45dda6d8b3"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/migration-is-not-free-memory.md": {
+            "bytes": 2145,
+            "sha256": "2ac168f2725cba7fc92dfcc67c01bf2258dff6a2dbe1ddb063bc60c175d82d0a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/parallelism-change-with-state-cost.md": {
+            "bytes": 2069,
+            "sha256": "5accc81086e4500d66f894ef6af9951e1cbb76941f6c631aab5c5561a786d8d8"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/preemption-under-kv-pressure.md": {
+            "bytes": 2630,
+            "sha256": "462200a6d0b193ef01f32374a463e0f2838ca8edf33d9ab898def65ffd0b490a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/knowledge/scheduling-state-boundaries.md": {
+            "bytes": 7488,
+            "sha256": "d4ef29869959b835a5dcb06ddb2fd8e6cff49bb176994dc1d46eb9a852f6228d"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/fastserve.md": {
+            "bytes": 9289,
+            "sha256": "9b4e275843000ef453d508d9c3c9c3f8ea1de5545fbd01d7abe74baa8a4931a7"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/llumnix.md": {
+            "bytes": 10181,
+            "sha256": "f252ce33ac2db017701ca1d49813c5ec1996c4c92900cf35bb6bda7fc59c0f18"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/loongserve.md": {
+            "bytes": 12422,
+            "sha256": "7bc6d1e256d461b04acc8ed856cf99108faceda9cde729ab95bcca596fe019ca"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/orca.md": {
+            "bytes": 8826,
+            "sha256": "f5a3b5870d64226408a8fcda5d69483769d823f8a5efd7f11698c8da483ab195"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/fastserve-readset.md": {
+            "bytes": 7173,
+            "sha256": "9866bd0c78b2b638998b3b632f6bae26aef032462c81abee336d309af5b07831"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/llumnix-readset.md": {
+            "bytes": 9386,
+            "sha256": "3a2819d65ad24087c08530492ef39d8ea046ed1f780d53512f1d23e2b3578487"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/loongserve-readset.md": {
+            "bytes": 12490,
+            "sha256": "f1a383e7bdbbd5a5a472d7f6e1292c8a17d4910ff736ca5a6aa422e1bee74f8c"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/orca-readset.md": {
+            "bytes": 4796,
+            "sha256": "a497ef23b7e60c0fbc15d95b7e07cd20547cfe56f37c24dfe6ed70d79ef7a3cf"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/source-notes.md": {
+            "bytes": 9575,
+            "sha256": "84033fd514773f10c0bddd9914fabb980497ba2b308f226a922c3e1888413827"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/references/evidence-map.json": {
+            "bytes": 14384,
+            "sha256": "dd8bbab5d285ed3a1a19d2f06ad84e8c444251690d8f673def3e2681765ef6c1"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/references/repo-provenance.json": {
+            "bytes": 1313,
+            "sha256": "4e705a964697712e950e02bedc41320806a41311b7f4dcf32f923d60b1d2e919"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/references/repo-provenance.md": {
+            "bytes": 3261,
+            "sha256": "6e841c38c35ff2cd6ff8139cb1a6c8f75701647897f1089a94a5c62928609941"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/references/repo-routing-metadata.json": {
+            "bytes": 432,
+            "sha256": "a20b54f682eb52311e445ac62b95d1935c001296f5cf23b7764cb04477d535c9"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/SKILL.md": {
+            "bytes": 1835,
+            "sha256": "2a945b5fddacfb90d1cd7cb2b2f767067d641061423a65e1c19afecec7f3c06a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/references/capability-boundaries.md": {
+            "bytes": 6560,
+            "sha256": "c8081d03f04f7481080c46c7dcf728962729bc122f49f9502d50cd1e3d04bcf3"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/references/runtime-and-invariants.md": {
+            "bytes": 8265,
+            "sha256": "afa1f967545021f36bc1b59dbf6aa2c367f4d5ece4f08dce755e2017cf50dbe5"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/extend-validate/SKILL.md": {
+            "bytes": 1731,
+            "sha256": "495679e5b99341bf80daeaacd2d4c2699aba16bc16c447cbd8d2fe17f7559798"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/extend-validate/references/change-boundaries-and-validation.md": {
+            "bytes": 7624,
+            "sha256": "f7fdd2d699f09d96fbf5ca9382f9b56113c870db2688d8e1118eb8163519fc4a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/SKILL.md": {
+            "bytes": 1812,
+            "sha256": "0ef4d2e3a96cbb277557e169a3fc8edf0f0504aad03259771b432a33ee008f78"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/references/commands-and-measurements.md": {
+            "bytes": 5667,
+            "sha256": "76ab51083b14595c49fa1c255cb7e5ebb996f14930a51ab3cbe864cc334b5f80"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/references/experiment-decisions.md": {
+            "bytes": 7644,
+            "sha256": "242593c0d386e57104b57a09bc7db97c2bbbcdb905bb68a397cfe612e07fb083"
+          },
+          "src/inferrsi/data/skill_library/skill-index.json": {
+            "bytes": 7514,
+            "sha256": "8ce35e18f7629260d67a95617b9918eecff73ed8d9370dd3d4595c725ada9aa1"
+          },
+          "src/inferrsi/data/skill_library/workflows/cite-check-downstream-pr/SKILL.md": {
+            "bytes": 1093,
+            "sha256": "485932cdf3b1fba5dc69be73e75d38f86a7b79442d26e82a296c963a7598d226"
+          },
+          "src/inferrsi/data/skill_library/workflows/claim-contract/SKILL.md": {
+            "bytes": 1075,
+            "sha256": "3af1dbbe81d91cb05fc2ef15a2ae25d7592a08558e1a5535f49c52f33a488c55"
+          },
+          "src/inferrsi/data/skill_library/workflows/competitor-scan/SKILL.md": {
+            "bytes": 1088,
+            "sha256": "0ede58ace65a9324a538b4816add1684f230ba536d4364fe5a5dd2c0bbed3776"
+          },
+          "src/inferrsi/data/skill_library/workflows/draft-acceptance-issue/SKILL.md": {
+            "bytes": 1230,
+            "sha256": "ce983acfb7d7d3272775e7cf00c92e4ddadd97511b602575a76e5ab7ade01b56"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/SKILL.md": {
+            "bytes": 5681,
+            "sha256": "364beed2ca617e71f4f955bbd024740d411b94eeb9557d7e055d4463cc22caba"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/cross-repository-mechanisms.md": {
+            "bytes": 17050,
+            "sha256": "28efa1b25f548763180b70d8d0ddebb7d4e7227df189f871fd233fe1d4881303"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/import.md": {
+            "bytes": 3463,
+            "sha256": "c8a57db94768cc6cd4193932def75ba67ec453e56ee0e54f176fb82f1af53091"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/model-aware-preparation.md": {
+            "bytes": 9056,
+            "sha256": "787c017aade68ee1133b979d636fa24062302cef6174c001171c963947ee8da4"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/overview.md": {
+            "bytes": 8099,
+            "sha256": "7ed0a966239b9481b82dd349245cfbfe966c030fc916771c91fd0692464ffc56"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/repository-source-scope.md": {
+            "bytes": 11967,
+            "sha256": "85202ca209171c50573c2fc209acb1e937d8211e1652e4ca9c6cc4bce04d03eb"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/source-manifest.json": {
+            "bytes": 1613,
+            "sha256": "0ce1678803d1ede6b485fb6b4b1c88aba909056ee6e3b5af691899ceb2056240"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/SKILL.md": {
+            "bytes": 6502,
+            "sha256": "74cd863df2e80dc2bda1713d24207809173a3116f30325e3487e1e3517cfec1e"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/metadata.json": {
+            "bytes": 3236,
+            "sha256": "12fc9272bd5fbf02dd9b897d8ca40984b72ae54f20492ace5304711bbc4d71c3"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/COMPARISON.md": {
+            "bytes": 4267,
+            "sha256": "38587678a96581f5c2dc697d982127343565504bffa191352811dffcc165aa23"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 5557,
+            "sha256": "fbc4a96b6f6c477706eed83eaa4fcc29a14f881637629d6c004601ed049f2523"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/QA.md": {
+            "bytes": 1853,
+            "sha256": "3f3785ec39a6938a5d1f2e874e99153fe92c0e666dd3e78a499c8b47d891346a"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/README.md": {
+            "bytes": 2151,
+            "sha256": "221785e91c45588f82b3caab6d52f1ea1d102b16139c20a0802921e2789e6d28"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/UNREAD_LEADS.md": {
+            "bytes": 1038,
+            "sha256": "6c1d87ebe491d27666703149eb05ad582c7e5aff77f0ac5c11a7b89c946f96ee"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/cases/methodology-counterexamples.md": {
+            "bytes": 5213,
+            "sha256": "f9fc21753ea41eef77ca020d78d031e945e5dd835b5d09c5ff0e53749ceb2ca4"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/knowledge/evidence-routing.md": {
+            "bytes": 5282,
+            "sha256": "406d5d74c3fef4c1a23d495492e21a084c3c46495a1a84b385703ed104358874"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/learning-to-commit.md": {
+            "bytes": 2847,
+            "sha256": "9c2d1d738e896d05c00f53e6adf14c9df2501ecc3f0020e71b5a1c2192547086"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/protocol-details.md": {
+            "bytes": 6520,
+            "sha256": "41298c6bde3a28f0712fa06a18bbc4ac59aea6d17cc9eb1318744f4121e3bab6"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/readset.md": {
+            "bytes": 7086,
+            "sha256": "6f953d20d8958c68339c9ceeae348604d0fea3e9c6ccb9b64a63d3b210426a22"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/repo-to-skill.md": {
+            "bytes": 3151,
+            "sha256": "5ebdac2a2bbbe6614f9d4096bc9895bd685acdf76414e2c3c1fb7b50d57c738d"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/skillopt.md": {
+            "bytes": 3040,
+            "sha256": "148075daa67e19f4b4351710498438d17c279a9bce56b2d01cab26ff3be33e99"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/source-manifest.json": {
+            "bytes": 2431,
+            "sha256": "716cd49ed01824ee9f6b6fc55ebf2b24fdd5d5a3e677f6e28e34a472616bbe25"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/trace2skill.md": {
+            "bytes": 2917,
+            "sha256": "230d13b77ba819b5e3edee147523e8cad6bfce95afccc140fb4946d6c8747a2e"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/updates/bounded-updates.md": {
+            "bytes": 7630,
+            "sha256": "d02c5bd0109846d1e1987de074810321cb4af277cd3f66ab2ac268aadd5ee142"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/skill-factory-design-20261014.md": {
+            "bytes": 9465,
+            "sha256": "fb8ce1ad83b7a6ddc657e89ee2eebb710873b213cfd85bdd6d823c8691b3cb9f"
+          },
+          "src/inferrsi/data/skill_library/workflows/rsi-evidence/SKILL.md": {
+            "bytes": 3680,
+            "sha256": "a55610874b658b86a39a1a049efb4e4eab0140f6579ed74dd7718a4ca8cedaca"
+          }
+        },
+        "sha256": "6e19365a0ad8e2951af0f3106f93aa545b4ad00b6e0f0317a20f32065b6a05fb"
+      },
+      "environment": {
+        "filesystem_cache": "not controlled; process cold is not OS-cache cold",
+        "implementation": "CPython",
+        "logical_cpu_count": 8,
+        "machine": "AMD64",
+        "perf_counter_resolution_seconds": 1e-07,
+        "python_executable_sha256": "737a7e3b71e3578f8432acc7dd88c452e593622c544bc13da4789d69c63da5ae",
+        "python_version": "3.12.6",
+        "release": "10",
+        "site_initializers_disabled": true,
+        "system": "Windows"
+      },
+      "git_head": "4817f92121df73fd9cd5fb4a319e3c56a68b4a5b",
+      "knowledge_record_count": 108,
+      "library": {
+        "bytes": 2294958,
+        "file_count": 342,
+        "sha256": "be8726296e6186601692dc2d95d99bfbba76f0b4c758de3a88a9b2a523eaa37e"
+      },
+      "modules": {
+        "inferrsi": {
+          "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+          "source": "src/inferrsi/__init__.py"
+        },
+        "inferrsi.catalog": {
+          "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+          "source": "src/inferrsi/catalog.py"
+        },
+        "inferrsi.cli": {
+          "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+          "source": "src/inferrsi/cli.py"
+        },
+        "inferrsi.knowledge_operations": {
+          "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+          "source": "src/inferrsi/knowledge_operations.py"
+        }
+      },
+      "packaged_skill_count": 60,
+      "python_sources": {
+        "bytes": 274444,
+        "file_count": 53,
+        "files": {
+          "src/inferrsi/__init__.py": {
+            "bytes": 102,
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47"
+          },
+          "src/inferrsi/__main__.py": {
+            "bytes": 48,
+            "sha256": "935a1c1166b0c1ea35a82256345000bf2c73ded718d77773bc27a71ecce28f7d"
+          },
+          "src/inferrsi/_skill_selection.py": {
+            "bytes": 5790,
+            "sha256": "4640399f4013425589f6702e2db844820619ba35967d1c3519978eb307168440"
+          },
+          "src/inferrsi/arguments.py": {
+            "bytes": 12200,
+            "sha256": "4456600cf4a362cf069594fe36db46932cf5cb14ddd37683aaf3538f5cf76b4d"
+          },
+          "src/inferrsi/catalog.py": {
+            "bytes": 6855,
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5"
+          },
+          "src/inferrsi/claim.py": {
+            "bytes": 805,
+            "sha256": "5f24c4fa91b48f0bb4d4554531607763fc8fc108e90f332816f06cfb520b5cc1"
+          },
+          "src/inferrsi/cli.py": {
+            "bytes": 14205,
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88"
+          },
+          "src/inferrsi/contracts.py": {
+            "bytes": 119,
+            "sha256": "03000cb9246ce8c0967147533f7ad25df918ba2674839801453ffb870f496490"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/scripts/roofline.py": {
+            "bytes": 9004,
+            "sha256": "524de48a6764401695fe1444f1fa3d38f363069b0cd2fba4715a86ca4b28a5d4"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/scripts/summarize_trials.py": {
+            "bytes": 5170,
+            "sha256": "6a243a08da15659ced6519f49f52a49a675bd9e952740c3ab03dd21964eefbe2"
+          },
+          "src/inferrsi/evaluation.py": {
+            "bytes": 183,
+            "sha256": "482c40710d428206c59fe3696703248cd86c4b3c7d19ad41017c1369bd6d780e"
+          },
+          "src/inferrsi/evidence_paths.py": {
+            "bytes": 859,
+            "sha256": "cc46231e0b8334ef5742defdda20e0093e22b30c86af59df1fb3ae4b5caa70fa"
+          },
+          "src/inferrsi/execution.py": {
+            "bytes": 755,
+            "sha256": "f43d8cbb4d43a72a7ff8a10b29a5fb9cf4cb22af54dd35943283c843fba0da92"
+          },
+          "src/inferrsi/harness/__init__.py": {
+            "bytes": 177,
+            "sha256": "1006e4713003f44bfb32b2e385a918e1652b9e9833801e4555674a71e57fb0d5"
+          },
+          "src/inferrsi/harness/evaluation.py": {
+            "bytes": 4160,
+            "sha256": "9b00b2e93959f16b7b76bb4b257df7dcf48520a06fd6adae96e9a165a05e6e80"
+          },
+          "src/inferrsi/harness/improvement.py": {
+            "bytes": 5777,
+            "sha256": "3faa0f5928a707a98f35f45d1ca18e226c898941e6a07b3829eebeb53e70aeb8"
+          },
+          "src/inferrsi/harness/reporting.py": {
+            "bytes": 3566,
+            "sha256": "38c58f02c44e8c8aeacf46527e142be0e3142a83e4b2693246d0ca76c0ddd7c3"
+          },
+          "src/inferrsi/harness/runner.py": {
+            "bytes": 6567,
+            "sha256": "a566307b9b586655941b627f428b5121ed38d7ac80bccd1e1799e443fb0de911"
+          },
+          "src/inferrsi/harness/scenarios/__init__.py": {
+            "bytes": 79,
+            "sha256": "4e97913c806077f0cadfeb29d986d2fa75b55ab8766cb86e6e3e9a195a9a1f5a"
+          },
+          "src/inferrsi/harness/scenarios/external.py": {
+            "bytes": 1016,
+            "sha256": "84d5f6920501a9d882cc4fffbc721d2051fdbf7c90694bde9cbfba3c2b866304"
+          },
+          "src/inferrsi/harness/scenarios/media.py": {
+            "bytes": 2914,
+            "sha256": "67d979a3ed6f13a8e3e1e40a304f78e5a3e6c2556b3aa8a11c41156745d7050a"
+          },
+          "src/inferrsi/harness/skill_evaluation.py": {
+            "bytes": 11093,
+            "sha256": "b710b8e53758757efd68c703449eb4bab7eb7ec50ae3dfc52a010cdebd8c3bf5"
+          },
+          "src/inferrsi/infra/__init__.py": {
+            "bytes": 85,
+            "sha256": "8779897733779266841b0f79b84b1dc43f0fa57c013855c906c809829ecfc77c"
+          },
+          "src/inferrsi/infra/artifacts.py": {
+            "bytes": 1040,
+            "sha256": "e72e1bebdbb88e42de430238489252386851f9e6645ed7780f2aad2e90338120"
+          },
+          "src/inferrsi/infra/campaign.py": {
+            "bytes": 1028,
+            "sha256": "be505fba1592611bd84ceaca55afbac027205413a2639baefa0c3bb343230209"
+          },
+          "src/inferrsi/infra/commands.py": {
+            "bytes": 811,
+            "sha256": "1890f115b3a2c5c6832173bf60432ad595bd6bf1c69783ecd460c6f8961194f1"
+          },
+          "src/inferrsi/infra/environment.py": {
+            "bytes": 2067,
+            "sha256": "60e46c2aefc216db202b41f44542e7b893e08e0721c667b9ccc50411d1ac84ee"
+          },
+          "src/inferrsi/infra/executors/__init__.py": {
+            "bytes": 64,
+            "sha256": "327c3cd5c821c832ff260a41dba0ee1e439cdabefc9edfb9cad1cd38d14bc87a"
+          },
+          "src/inferrsi/infra/executors/local.py": {
+            "bytes": 3654,
+            "sha256": "37bcbca9f3261c8c58a8bf806c9da40720b18d0676c82e8ed24c3901ac9bbe77"
+          },
+          "src/inferrsi/infra/http.py": {
+            "bytes": 634,
+            "sha256": "f65d6c2b55b6a89694caaf6dcdf7d74f417e9ced103eca1b3f218e305c6b6cb2"
+          },
+          "src/inferrsi/infra/inspection.py": {
+            "bytes": 2415,
+            "sha256": "d1c66deb0cced529bca996d2f88b60f51fa3417ff1cae365c650fec808946ecd"
+          },
+          "src/inferrsi/infra/platforms/__init__.py": {
+            "bytes": 1494,
+            "sha256": "9e88e8b9a396ed90c221e1585135c0a1b58da41c1bb73ed9237c53b6684abfed"
+          },
+          "src/inferrsi/infra/platforms/cuda.py": {
+            "bytes": 3309,
+            "sha256": "9e0e4276e1b2984605493b14b82fba3e86f6635204719c318302e624f66f8b6f"
+          },
+          "src/inferrsi/infra/platforms/mlx.py": {
+            "bytes": 1971,
+            "sha256": "f3b81470db28bd6183ccca2e763c5e079686bba6cdfe6c08ebe2cdf64216f2e3"
+          },
+          "src/inferrsi/infra/processes.py": {
+            "bytes": 8151,
+            "sha256": "3b96ebc7a50913e78e86ff3dcba74c8ee8d03d547d9591ca6488f390ac1e6189"
+          },
+          "src/inferrsi/infra/targets/__init__.py": {
+            "bytes": 186,
+            "sha256": "1f3ea75dded7c6cc967754c7b6d4f30ef00da8191a749938aabc3c6bb8f86eb3"
+          },
+          "src/inferrsi/infra/targets/git.py": {
+            "bytes": 2910,
+            "sha256": "6327c4f126086f310b6de309fc7659c9759da28ed76efaeeb94a3d2fba29a0c2"
+          },
+          "src/inferrsi/inspection.py": {
+            "bytes": 692,
+            "sha256": "afa44bf4cdfbb58a318c21f9b292ca142d1bbb22750e35a4cc8244da5350ebb3"
+          },
+          "src/inferrsi/knowledge_operations.py": {
+            "bytes": 27444,
+            "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b"
+          },
+          "src/inferrsi/knowledge_store.py": {
+            "bytes": 13700,
+            "sha256": "5e794e28906ce4687aff7f14a507e267184640fc8a4fac19e0ced279ce6a6cb6"
+          },
+          "src/inferrsi/model/__init__.py": {
+            "bytes": 77,
+            "sha256": "3c1dba4f5159fd9fbc8d4d96e2a83c8441491c1b1dc7b42031a3fc1782ed9d9f"
+          },
+          "src/inferrsi/model/anthropic_agent.py": {
+            "bytes": 24696,
+            "sha256": "b0ef9cd542544292eeb90ab05ef410be9ed6e5d1a40d729890f49562c28c12f0"
+          },
+          "src/inferrsi/model/protocols/__init__.py": {
+            "bytes": 85,
+            "sha256": "5a2f54483b164de04e309dbebb87a700b213682b0f44a50d89a752dd71a4ce47"
+          },
+          "src/inferrsi/model/protocols/image.py": {
+            "bytes": 2047,
+            "sha256": "251a1f8f885da2cd8e9d3baba766c525afe293091606b4e8668c43ba2cb7cb5f"
+          },
+          "src/inferrsi/model/protocols/metrics.py": {
+            "bytes": 610,
+            "sha256": "17db16d6ef15b41fb60d8f09507893accc43f9cb3ad686b8dc4e9839713ebc56"
+          },
+          "src/inferrsi/model/validators/__init__.py": {
+            "bytes": 65,
+            "sha256": "1d4b6ffa968e283219ddcc8a09de9280238a8d65ae2dc8cd2b60889877a8eb0b"
+          },
+          "src/inferrsi/model/validators/flare.py": {
+            "bytes": 19975,
+            "sha256": "3696b0609c772f2222b448833608565a033123d7e92b11fbc545dd356fec590f"
+          },
+          "src/inferrsi/model/validators/image.py": {
+            "bytes": 3204,
+            "sha256": "79470b6b06a22d53ec6c40636a0b3d821a2e6ee2f7bb7aed3dac7241504add1a"
+          },
+          "src/inferrsi/reporting.py": {
+            "bytes": 127,
+            "sha256": "1f771890ab3e69b1aa9246b548a7825596bdf904f6c395dd218208084266853d"
+          },
+          "src/inferrsi/research.py": {
+            "bytes": 5213,
+            "sha256": "9bfe59dd17b5476c18ce50d671711053c62800521c50d6b710e8140a2b4dca04"
+          },
+          "src/inferrsi/runspec.py": {
+            "bytes": 4398,
+            "sha256": "dca0952425caf42095ef0c7e2feb8ef4740cc422ecb61e777f203962d9b87709"
+          },
+          "src/inferrsi/skill_catalog.py": {
+            "bytes": 9938,
+            "sha256": "65f5a1f3f6894009abb98ec2fec30a63cc1bb2f6ce6aeb01f9f7a3dce0439f2d"
+          },
+          "src/inferrsi/summary_chart.py": {
+            "bytes": 40910,
+            "sha256": "d29023d60f6be804c1b01c9eb73765d4824d6c3bb870fe22c99a24544bc1d8bc"
+          }
+        },
+        "sha256": "4faa529dc47f23c24f8c5b8e0110d8e413aab20df5ad61db591fb751564a6448"
+      },
+      "tracked_source_dirty": false
+    },
+    "B": {
+      "corpus": {
+        "bytes": 2883003,
+        "file_count": 458,
+        "files": {
+          "src/inferrsi/data/knowledge_packs/README.md": {
+            "bytes": 14172,
+            "sha256": "b877e67d9cccf0dcbe57419eeb3ff9374e5bd03ff8eec63adef1025c446140e9"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/diffusers/cuda/diffusion/group-offload-transfer-overlap.json": {
+            "bytes": 3277,
+            "sha256": "9bec3d912192efa2f2c853d3fb70bdae21a75875630a02294fcd084b18c30414"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/flashattention/cuda/attention/generation-specific-dispatch.json": {
+            "bytes": 3085,
+            "sha256": "90290d7bc80b1626863e0520c91bb196d93ebfe08190a450e9de59c6505ae8ad"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/flashinfer/cuda/llm/paged-decode-dispatch.json": {
+            "bytes": 3745,
+            "sha256": "0f2986494fcc9b751df71e598fd61caa68e28bc4b0ac401e4cc763bc3f75283b"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/cross-stream-dependency.json": {
+            "bytes": 3809,
+            "sha256": "b829315e55adf8e11f0bd581d5f45dd130d7eb3f81ce7de532f63c7b9c27cbb5"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/evaluation-completion.json": {
+            "bytes": 3841,
+            "sha256": "19d2e506ac4f50895776a83a062059aa7a052175c8e4b21dd0d46e1a5a6716a1"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/post-detachment-buffer-lifetime.json": {
+            "bytes": 4663,
+            "sha256": "18d78c0e542dacb42ca33e5264d92aef80fa06d163fd16d4feafd5ed0827ab99"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/pytorch/cuda/compiler/cudagraph-shapes-and-lifetimes.json": {
+            "bytes": 3623,
+            "sha256": "7fef4122ff794967e61f55d02a4a6b558d7e498d9513607f5ca86db6386928a8"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/transformers/common/llm/kv-cache-strategy-contracts.json": {
+            "bytes": 3072,
+            "sha256": "a2df26052112a9f3fc6242b2bd636ab13991442db57e56646168d6c52032d38d"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/triton/common/kernel/bandwidth-fusion-resource-limits.json": {
+            "bytes": 2941,
+            "sha256": "6e9673d1a600a6ad34820e4325f2a6417f44f0f96d8ed0bcf21466d998f19ee4"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/cache_dit/common/diffusion/residual-cache-calibration.json": {
+            "bytes": 3841,
+            "sha256": "bb9509ad0225a91b52bc6787477581d05f0ecd5e44fa59f08f718eea36a2ac45"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/bounded-windowed-replay.json": {
+            "bytes": 10959,
+            "sha256": "43613225e9ccdbd34c0f7e3e0767887b9bbfd0f38b2c21684677d99e31a432f0"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/speculative-verification-budget.json": {
+            "bytes": 10177,
+            "sha256": "a729e1fc00be44e1480a5dc65bbcdd58df8599681df74dbc651a5a49f27e9028"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/zero-overhead-deferred-scheduling.json": {
+            "bytes": 8191,
+            "sha256": "ed5426af145f066f6183bd316c4073c46f324f5ee83aa9bd616387c0d0e7149c"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/lmdeploy/cuda/llm/kv-quantization-capacity-cost.json": {
+            "bytes": 3366,
+            "sha256": "873997b38981ccc03ff73e6fb84e96022f293a019843efd8c4aa50f27878ac9a"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/codec-constants-and-state.json": {
+            "bytes": 4250,
+            "sha256": "1cecd7905fef5af3ef8952a4d803900a7df91d662ebae34db47cc0dbf352d4e7"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/fixed-customvoice-prompt.json": {
+            "bytes": 3660,
+            "sha256": "ea5e8fdf54cb731b255b73aa0cf47ef53fac103baa2bd5246be45f95d54cb593"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/graph-batching-state-ownership.json": {
+            "bytes": 3950,
+            "sha256": "130741a9f7ac9d366784d6c4dd42860262235caff0c277de1e8a8f8ec933b37f"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/sglang/cuda/llm/hierarchical-prefix-cache.json": {
+            "bytes": 3641,
+            "sha256": "b50ca58df28d4343545d07adf0da23f8f181297b30aba7870e5690630e7d7315"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/tensorrt_llm/cuda/llm/prefill-decode-disaggregation.json": {
+            "bytes": 2967,
+            "sha256": "349856dd98628f94987e014889cd1b15d0e3ae6ab6c576cdde917f28fbafb102"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/xdit/cuda/diffusion/parallelism-freshness-and-topology.json": {
+            "bytes": 3236,
+            "sha256": "3674b5916cb7b384951bf3e449b6796b161878daabcc1d7e7f6d6f8ff861cc17"
+          },
+          "src/inferrsi/data/repo_packs/afd_plugin/pack.json": {
+            "bytes": 502,
+            "sha256": "00042abeabdc9c0e24f1a4edaba47e233df398aef0ad4e181fc70191d383da5f"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-aclgraph-capture-replay.json": {
+            "bytes": 4311,
+            "sha256": "327414443ea9537ae320aa405912c338351bc06cefdc0b99352e1eea36f61dc3"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-device-health.json": {
+            "bytes": 4115,
+            "sha256": "633fa5fdcea1e1e9edda8a218d6a3fde0d12637848348e1e2fe7d1b0bd744a16"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-distributed-parallel.json": {
+            "bytes": 5841,
+            "sha256": "709ca0d061e4d6917fe8fa049fe93165e71dfecf96c7bf264ced25346381f603"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-graph-mode.json": {
+            "bytes": 4135,
+            "sha256": "bb82ab3d628b91be957e557545810ef5c20790be7d43296096f20a21f45ec391"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-model-adaptation.json": {
+            "bytes": 4137,
+            "sha256": "7ea1ab0f0044afff935a91071b36ac09bc46b5ef044fe2a68ab0608256ddd853"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-profiling.json": {
+            "bytes": 4186,
+            "sha256": "0d27058fd8d0ef8b524b7a60d1cb76950622746c94b5a03915e64e6edb4ab3d4"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-tuning.json": {
+            "bytes": 4439,
+            "sha256": "fcb49d71f59d8b39c1eaa03d24e95e2cbfe5f9d7f6e0f22df03ee0d60f6a568a"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-optimization-casebook.json": {
+            "bytes": 4371,
+            "sha256": "05783630bfaf3c7ef99b848865f4bbab50d0b52e80ad04fcd028c624d7638e9b"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-performance-analysis.json": {
+            "bytes": 7200,
+            "sha256": "b6eea897c12d0c6ac9d172089535c82c25eb71c09d7a4972e64812cc4682b5c9"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascendc-tiling-design.json": {
+            "bytes": 4505,
+            "sha256": "e69b6e89dd342ba55ae54b5fac14aa1b91ee2a361da40cb669c287ce519f4517"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/cann-attribution-overlap.json": {
+            "bytes": 4915,
+            "sha256": "8ac01b60f5cd37fd594d6cb5119e4e6525199d017296fc8779dc68a576f5525c"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/torchnpu-device-memory-lifecycle.json": {
+            "bytes": 4349,
+            "sha256": "7a79d0eb2c50381d40086e6db72d4bdd2cec5d2261bfb00cf73a60701a94da2a"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/torchnpu-pinned-host-memory-lifecycle.json": {
+            "bytes": 4431,
+            "sha256": "8bb34dcfbebb48667a94434b036a1a408b970af9890b047a3de325f364ed5f7c"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/assess-serving-preemption.json": {
+            "bytes": 4198,
+            "sha256": "e3d2a1a72d0d980f31fc85deee27db26bc0daa0c709e8ae99e42deb814a4a7f4"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/assess-split-backward-schedule.json": {
+            "bytes": 3915,
+            "sha256": "a8561794786eb0a897d831f00c92b4fdd3b0492316c1b1d8ad13c1f8f75ac815"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-h3-runtime-residency.json": {
+            "bytes": 12169,
+            "sha256": "602dd625874bd4a89861d5bb25881cc63e4b5a7fdedc2867089f0952943f6eff"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-llama-runtime-state.json": {
+            "bytes": 5816,
+            "sha256": "64f6ab8a95172378350323cead08bc6c201c624b711e6242938a6a21a8e83ad7"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-sd-cpp-residency.json": {
+            "bytes": 8349,
+            "sha256": "95ed97b717255be3b817e5c69debd6d41197559dbf21771507d5da79134a72e1"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-sglang-diffusion-lifetimes.json": {
+            "bytes": 7276,
+            "sha256": "5e5889223878c6cf42a49c6f09c60e9d9add88cb7e7d0df050c4e75d88a05aa6"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-vllm-kv-lifetimes.json": {
+            "bytes": 4275,
+            "sha256": "00424005a5eeb8b15f239edd2cb2c71047787e3ecf55924860cb7a78d137b586"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/bound-windowed-attention-replay.json": {
+            "bytes": 7877,
+            "sha256": "98b2291147a796ef59cf20a075ff662998ee6a529f85b49e9c6fb62fe41fad9f"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/budget-speculative-verification.json": {
+            "bytes": 7490,
+            "sha256": "66448f59460c9dffac46e74160388fab650cf8bbbd5bdc95c1465b70daf0c2dc"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/calibrate-serving-simulation.json": {
+            "bytes": 3992,
+            "sha256": "055b70e0b86f65156775b519e4b8cbf06a956a8993b06a66175167cedb78f2cb"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/cgroup-resource-controls.json": {
+            "bytes": 4801,
+            "sha256": "ad406b6dd5d9415bd28d448e64269a388b50e11f9eb7f5a9adb6b0a6246fd303"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/checkpoint-recovery-consistency.json": {
+            "bytes": 4740,
+            "sha256": "ccb85314a831bde58f8cc515580d001b174c11dca0bbcbc6221303acf629046d"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/choose-parallel-training-plan.json": {
+            "bytes": 3901,
+            "sha256": "40afcb41c59988fc1af8fdd00d2340800c54e96847dd0de16ecddf7ac221d1b4"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/cite-check-downstream-pr.json": {
+            "bytes": 2533,
+            "sha256": "dabff14684d5ae9a2b05a9743ff5b76d00a90fdb29319588aba67f5087aad706"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/claim-contract.json": {
+            "bytes": 2469,
+            "sha256": "78f15f6533434fc29adf1df651f7a15eb8ba81733f8b02c710d90102541d707b"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/comfy-ui.json": {
+            "bytes": 6919,
+            "sha256": "1496807d958df055f720ca93e6dcf7f7e8f506f3f12c94c9a2166f789b00058d"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/competitor-scan.json": {
+            "bytes": 2449,
+            "sha256": "c2e97e49e5185710cdfaa08ac1c2b1c19b958e68ebfca885d5aee0ed8c9f8c8b"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dcp-async-save-performance.json": {
+            "bytes": 4538,
+            "sha256": "359a5acf8304a78fc074688676f9a0be62f56f463d0960a544e409f49843bdaf"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dense-performance-evidence.json": {
+            "bytes": 4227,
+            "sha256": "d87f93a9dc9ddc583e3929baaac85f7a425666dd6dd71665933f18344ccec834"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/diagnose-vllm-admission-pressure.json": {
+            "bytes": 9789,
+            "sha256": "a1de40def775b2e840c27330b67e6f37c3c244db1731fab2bd3e8a787143210e"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/diffusers-invocation-state.json": {
+            "bytes": 6109,
+            "sha256": "914d2ffe9776ba931bed780a2a270c1525e3c6cbaf0735b8531064d3180f3233"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/draft-acceptance-issue.json": {
+            "bytes": 2549,
+            "sha256": "6aaa63dbfe6cda982ccd3a4e368527b80fd2873672c000c66d06d28e40a2bf14"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/evaluate-elastic-serving-placement.json": {
+            "bytes": 4334,
+            "sha256": "fab1fa814e789ec2d86587fcf7de7c38174464b9e9ba0cf53ef3fb470053e7c4"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/generation-cache-lifecycle.json": {
+            "bytes": 5972,
+            "sha256": "e5679cdecb722bbd17280913913da8473052ad09e9e7a13b3a9b95463a139ef6"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/host-memory-management.json": {
+            "bytes": 6530,
+            "sha256": "a36aea21169281866b2468860d3ffeb912e61edd571308b02f82939f17f694a8"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/host-overhead-optimization.json": {
+            "bytes": 4243,
+            "sha256": "05fd7f054966b96892980d585352bc1c3fcd47f439cb34268fdf07feee58a998"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/inference-benchmark-design.json": {
+            "bytes": 10735,
+            "sha256": "091bb086c6f0123dbb5db78968c13f846cad135e8969982efe557644b7c7613b"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/inference-optimization-casebook.json": {
+            "bytes": 6531,
+            "sha256": "c512025d7a59934c91665e24752d9393ca9b6a9dfb624c72789294defae93c0e"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/kernel-writing-guide.json": {
+            "bytes": 4545,
+            "sha256": "f1f0ca1d11d3550c9b8c5fdbed8f6cd09b877ac14341f973d58dca02b5d8995d"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/kv-tier-management.json": {
+            "bytes": 4609,
+            "sha256": "6141f61fb6b0d147b704ab7a2627e9babdb53a90e15b659152a3d565967afddf"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/lazy-evaluation-stream-lifetimes.json": {
+            "bytes": 5934,
+            "sha256": "34f47009a8a33f8907bd78f110b95fcd62409bef1b38e1055acb46b2eab285ff"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/npu-skill-factory.json": {
+            "bytes": 11868,
+            "sha256": "24c1d3bff288e86b99ea674e3252ec6f54d9ba944274aef6a7666197afd6788a"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/optimization-search-policy.json": {
+            "bytes": 7101,
+            "sha256": "d7bf71d5491765f62eec02fe5013dc1a5ef56e1516892da48593c6165a86366d"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/os-process-placement.json": {
+            "bytes": 6304,
+            "sha256": "e6e8fe5d7d66f3e98f05762d94441726dfa757ce7fbffd2b35402b3f327f8fb3"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/performance-ceiling-analysis.json": {
+            "bytes": 13695,
+            "sha256": "47c034d96026e1c2a02848d9c9c3caf4024a5033664bfd089ad30d9f23a80deb"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/plan-activation-residency.json": {
+            "bytes": 3871,
+            "sha256": "436e364020cc854fd52a10ee76e4593c7519bb0546c8aa5238fcfd89158e8ab0"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/repository-evolution-analysis.json": {
+            "bytes": 5865,
+            "sha256": "e0af402c3320ebd9bd36c774cb52c0708721a86ca446c05f6e94306888838e30"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/rsi-evidence.json": {
+            "bytes": 2475,
+            "sha256": "c8b4bb5a8c88009ac5b1526e82639715381c940d3c3eb295f0333cae6c2d3ca9"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/skill-evolution-playbook.json": {
+            "bytes": 4091,
+            "sha256": "0e063c2e7686de0c27c061e1aac0eb2d8c4fa83246e32ddeeb0e28126c176425"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/storage-io-measurement.json": {
+            "bytes": 4783,
+            "sha256": "80e589e3a0b67d3a045e03feeadc45e720ceb277e6124c917eec04863eb67419"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/training-input-pipeline.json": {
+            "bytes": 4296,
+            "sha256": "ae696bea8d2b7bfdaa29b6e5bc0bd861abcc0e42e7d9bec6ae20c6e7d6f27669"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/training-resource-budget.json": {
+            "bytes": 6944,
+            "sha256": "ea5e5e864e3aa34bd00be67c8e741877ae01480f039652073b65b71dfcd54c97"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/weight-loading-optimization.json": {
+            "bytes": 8139,
+            "sha256": "7f2d1fbc6b1cfd80c3d6ed8df107d4ca9b97655626ee078c345445a303b1b367"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/audit-afd-p2p-exchanges.json": {
+            "bytes": 6640,
+            "sha256": "9ac45aa784b430ad1ca9ffddecce8dbae9740cfe9c98c10b330e01d350015a38"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/diagnose-attention-bottlenecks.json": {
+            "bytes": 7602,
+            "sha256": "e525d75578de1cc164e6553d01efbd0a014e74646200580f59b1161c030413ad"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/host-overhead-profiling.json": {
+            "bytes": 3965,
+            "sha256": "e9e6ae8c5b9986e83cade3bd029baae112859e01aa6c96f741a264f3bee44adc"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/reason-hopper-kernels.json": {
+            "bytes": 15086,
+            "sha256": "758a649e651d8f9635ac2df45d18043e156c8d54b8f8dbc95f8e68ebcb866a28"
+          },
+          "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/parallelism-and-extension.json": {
+            "bytes": 3399,
+            "sha256": "1253f5f01bcf1498703c7f34960f0c50580df5ba8f140af557a903e296ad81d3"
+          },
+          "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/runtime-radix-hicache.json": {
+            "bytes": 3317,
+            "sha256": "5a0b2aed51649af214cdc1034f57650b7ce195c934a17452ee67a35131f80cdd"
+          },
+          "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/tuning-and-validation.json": {
+            "bytes": 3385,
+            "sha256": "ef4d4f7182da4433d59d72dda0d21b4aeafadd73260be11ab8d4cb36c20fafc6"
+          },
+          "src/inferrsi/data/repo_packs/sglang/pack.json": {
+            "bytes": 620,
+            "sha256": "bce2e0c42796a9ef499d459c37b56d1d56f4b19344841359c534fca29c0359ce"
+          },
+          "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/audio/serving-and-streaming-quality.json": {
+            "bytes": 3440,
+            "sha256": "97fc1d5a11a39a0634dae9e9602a4ec47a4d2348e121076c2d5c836cf0692114"
+          },
+          "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/omni/extension-lifecycle-validation.json": {
+            "bytes": 3482,
+            "sha256": "7da8dc5d2655099de2bcd5f0fa922d9e1bcb0bfd2f571f026a004a7f402a75b6"
+          },
+          "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/omni/pipeline-state-and-transport.json": {
+            "bytes": 3404,
+            "sha256": "0c57692772019504264e39bf0444a69e20cff86f6e112b072c53cf7a4299cc82"
+          },
+          "src/inferrsi/data/repo_packs/sglang_omni/pack.json": {
+            "bytes": 630,
+            "sha256": "8e70336bc2aac0c74732a026337813652b5d5c6e1fc60f5571e9442345077b96"
+          },
+          "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/architecture-search-space.json": {
+            "bytes": 3294,
+            "sha256": "2c5ad72d5bab4f7082fc3be1532c347444cde38f5a84639fe16b22134b68685e"
+          },
+          "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/controlled-execution-tuning.json": {
+            "bytes": 3255,
+            "sha256": "6a2bf611a432213919e1c237687507ceb0ab753cccad0e934dc77f86604805e9"
+          },
+          "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/engine-validation-contract.json": {
+            "bytes": 3295,
+            "sha256": "ef89775bc82965fe87a832c7ec8a5cf55cb9eba4ecfc6279caea9ce3b0410c41"
+          },
+          "src/inferrsi/data/repo_packs/strata/pack.json": {
+            "bytes": 617,
+            "sha256": "643b32defb8db9aa230f624856f4fb4b595b5c23db58d9453091f5420cff71c3"
+          },
+          "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/bounded-serving-measurement.json": {
+            "bytes": 3239,
+            "sha256": "fb00ddac14ef6b06f48da9ba335529bd55d5d83c2a2fe331f0aa1ca16afceb37"
+          },
+          "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/extension-correctness-gates.json": {
+            "bytes": 3343,
+            "sha256": "4ea7083d277df205ea4c361bee87a8207951596e9ef6cfa136cc7d2aeab02f46"
+          },
+          "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/runtime-and-kv-ownership.json": {
+            "bytes": 3280,
+            "sha256": "e6142c1065754716a5f26f42119722908dc8fff63e667c986d57f1227b8610bd"
+          },
+          "src/inferrsi/data/repo_packs/vllm/pack.json": {
+            "bytes": 508,
+            "sha256": "1705d96710249c4514c15d8d011b43c1a1ae7c15fbc9ddf11ef2d0e6b2ab7d07"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/minimax-h3-reference-kv-reuse.json": {
+            "bytes": 4047,
+            "sha256": "80b853fc594d387ab48bed04dca245e0a22b8cdaf7c9b853551ab3bafa302e84"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/minimax-h3-sp-head-buckets.json": {
+            "bytes": 3742,
+            "sha256": "c4d29d1bceb5e2cf86151bf53c9dafb3d83428fb1e9d2a848d852bb8824bf857"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/qwen-image-21-eager-t2i.json": {
+            "bytes": 5268,
+            "sha256": "ef107d43cc448219eda4d64ab961b61cda6350e20d30d94de1ff7c4ffff7e775"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/audio/streaming-serving-measurement.json": {
+            "bytes": 3389,
+            "sha256": "2680f32f28907f6abc7a7e1e3ecab4f46ee412f40f4898ed2f371ffadf47ae4d"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/measurement-and-profiling.json": {
+            "bytes": 4472,
+            "sha256": "c8b25583cd160024c7472a20debea34cb6d4bc79714b4bad2953555adc9f794e"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/production-integration-gates.json": {
+            "bytes": 6112,
+            "sha256": "649b49ff64d3e9e88bd620027dabf9d126ca69bbe91b448f28e845ecbe1f8e2f"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/step-state-and-output-validation.json": {
+            "bytes": 5090,
+            "sha256": "e1ffe28c4fa54af739cdef2fedf4755bba18e078f8380d4fde8138ea712e5778"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/upstream-review-gates.json": {
+            "bytes": 7614,
+            "sha256": "5127709bb85ce31008d19468e1b4fcb5454a71776f69f991841cd9ef7558f9c4"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/omni/stage-ownership-and-search.json": {
+            "bytes": 3427,
+            "sha256": "e75d6c2829ae22881893f5da2ad7edba099fd64ccfb0fc93d30473e06e9d4d80"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/audio/concurrency-scaling-state-and-overlap.json": {
+            "bytes": 25324,
+            "sha256": "cb47fabb232db81ed525cbb54371becc7e4a28b2bae9d8492f60dfb6f31563e8"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-adaln-schedule-cache.json": {
+            "bytes": 4253,
+            "sha256": "ddbba3bb72b5789938595a82d3280e861f75f6ed54f05a1d2b7dcb1089d30f63"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-cuda-mxfp8.json": {
+            "bytes": 4468,
+            "sha256": "e4c538f03f4ab90a4acc8e3d8721bbd226aec1ee9bde45757a7c27d3d970821d"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-fp32-modulation.json": {
+            "bytes": 3625,
+            "sha256": "abb8d94a1b025f0e2c8e085275e96ee85d68aeb27cb12d0918e4647f8255b693"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-qk-norm-rope.json": {
+            "bytes": 3706,
+            "sha256": "4c1c4acfb9cc0baa90bc28fed1903020ea1ea38755b84315711795db45e48bb6"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-ref2va-teacache.json": {
+            "bytes": 3835,
+            "sha256": "8aeb3f99ac4227cf8b3a5cece13d938b19082e265a557eddae27b7e495bcc22e"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-vae-stacked-tiling.json": {
+            "bytes": 3508,
+            "sha256": "8cc18eacbbfd8fc4d3a6b4152c4b584d559fc9b15377da56ef072e542d807c93"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/pack.json": {
+            "bytes": 566,
+            "sha256": "58c971e9dcf6ac15752629ed4d9426aeb5b5d857fc37b5bcf921aa2e81c0cdd5"
+          },
+          "src/inferrsi/data/repo_packs/vllm_rlt/pack.json": {
+            "bytes": 649,
+            "sha256": "6097041b8d026740060e527f49b1b4e288315b1c3227f7583805b86d8fa415f7"
+          },
+          "src/inferrsi/data/skill_library/README.md": {
+            "bytes": 18819,
+            "sha256": "ae583f3578425df1dc91ba32b8b6543c0469c9a817ebea873a77051d6b13e1ce"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/SKILL.md": {
+            "bytes": 5494,
+            "sha256": "7581b56336d07213a87d01477793ae1a8acf0b7cf876fce36c9a761ee32c1ff4"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/metadata.json": {
+            "bytes": 1429,
+            "sha256": "68ac7f64b7ab947fc2db664bc3990f102c1cc5539b42350aa5818d0bbf17d86c"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/ownership.md": {
+            "bytes": 2081,
+            "sha256": "40688cc44183027efbf2fca8d3548c299cdf2e33fef54f76ab3549353eb9bd1b"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/source-map.md": {
+            "bytes": 4635,
+            "sha256": "10562a75683bd9bd07ec2e37a1cbd0f93f2534d6f85b55a566adb2d97485215f"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/validation.md": {
+            "bytes": 1477,
+            "sha256": "5a0b854b87d091d7d505b0bdd5c03330052507a851ab6edf35b93a6675f2d4f7"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/SKILL.md": {
+            "bytes": 4010,
+            "sha256": "a95569bcc0da6e715fd74783a6a1e0489e92b680c582f15dafbf98df589d505a"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/metadata.json": {
+            "bytes": 1519,
+            "sha256": "9c2293c8956508de987c886c7224d673d22912e7e915b026b05abdd287e97401"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/execution-lifetime.md": {
+            "bytes": 2147,
+            "sha256": "c4bc41e08aca68f661325f61e6f0db1760a079dc362f295c847bdc374fee3bf6"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/source-map.md": {
+            "bytes": 5918,
+            "sha256": "798b9cb89ed6b885eb47464ca9f137d2802a54571b9327e631d7742d3e5abd5d"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/validation.md": {
+            "bytes": 1608,
+            "sha256": "ebfc00bf7d6ab0a1dc415953ae84d56ccdc713d5614aff85e433c4a843793bff"
+          },
+          "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md": {
+            "bytes": 7322,
+            "sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845"
+          },
+          "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json": {
+            "bytes": 587,
+            "sha256": "db7ad368394ccbfdeff6a16e3423c300d51305b1f95ff5c3936df91738a8cc35"
+          },
+          "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md": {
+            "bytes": 5314,
+            "sha256": "22afe4f4d77f441096cc657d55d763473bfcbea8fb531c8e493fe49ab84955f7"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/SKILL.md": {
+            "bytes": 6745,
+            "sha256": "d494fc7e539cbb7388189bd7b3f04bb5c4983fdd7a82484348aad1cee7f741f2"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/metadata.json": {
+            "bytes": 558,
+            "sha256": "71c68431663ed2336cccc985a354adab6b13b4cfd6123a18a5d8d53c28268509"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/references/source-notes.md": {
+            "bytes": 13004,
+            "sha256": "6ffbfb68cc7e04923b4fbca2a73a7555b7c3b45a6f40819bb5f3213ef05cf88d"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/SKILL.md": {
+            "bytes": 5977,
+            "sha256": "03829039860ece35519f2e4c339a48b5014ec25e07a2c150a3efb6a3a4e35b4a"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/metadata.json": {
+            "bytes": 559,
+            "sha256": "80d80c15514919826fa72d6f6cf26b76db8accd85d21187fc239232ca454e7bf"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/SKILL.md": {
+            "bytes": 14703,
+            "sha256": "4ce9dcadd1ae3c7a8cc633336760d0a3c382f354f1b7028e3a10d28fa0d0c9fa"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/metadata.json": {
+            "bytes": 1403,
+            "sha256": "b6cb25879b1bca5f5af08b48776f2e664cf5dcd11a0efb1c4f56879e2afa9006"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/references/awesome-ascend-communication-tests.md": {
+            "bytes": 9388,
+            "sha256": "0cb27a3d24409171bb414d7ebf0c49ad62be5f46482f29624695496ba9eb3e63"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/SKILL.md": {
+            "bytes": 5839,
+            "sha256": "9203cc5a14a1d65df5fbf275734b93be023e06527459b2f79864f28a338cb44a"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/metadata.json": {
+            "bytes": 536,
+            "sha256": "72b41e7845915d34b09219624e82976386f3125816293540c2c38143fe6b477f"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/SKILL.md": {
+            "bytes": 6440,
+            "sha256": "8fdf46fb2743b67a66bc069d22a45305a222debfb606777bc451b4b8f31d9cc6"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/metadata.json": {
+            "bytes": 536,
+            "sha256": "be681a3206e5e183b4c9670c9d7784bcffce18c891449e5b15f94ca5f1324b9a"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/SKILL.md": {
+            "bytes": 12988,
+            "sha256": "401981cf7828edcab0c9ea882f7ecd83badfa3c2a95006e912cf99d67b3d6a12"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/metadata.json": {
+            "bytes": 760,
+            "sha256": "efd37bb23acb500dad4ec8ebe7ddfe5c3f196374052825f923ac85da8445f06b"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/SKILL.md": {
+            "bytes": 12236,
+            "sha256": "38b151bc6b4e9734c6fa41efb7a30343559400c0c75bbdd26bd8a968b8ea4592"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/metadata.json": {
+            "bytes": 557,
+            "sha256": "04d2002fc8a0a7becf86b0b73daaf2da8716ab466d7f988da43d41a10f26680e"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/SKILL.md": {
+            "bytes": 9172,
+            "sha256": "647f5e9a288f2c4e1617ba53df59d9d827df749eb4fe86182d21a9c41175caf8"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/metadata.json": {
+            "bytes": 536,
+            "sha256": "f425f7c1abbd616387273e8932c6bd02f4b1da012f099286ad4171e0f6037b22"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/SKILL.md": {
+            "bytes": 9880,
+            "sha256": "a057b7f86a24f34a8f84abbbb4910d58d4e50ff47060e1c404600017e3febe89"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/metadata.json": {
+            "bytes": 2297,
+            "sha256": "9e705625c8308013c287576b0247da6f6116d817a4a535cbecff80abd6ea35ee"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-CANN-2.0-LICENSE.md": {
+            "bytes": 5262,
+            "sha256": "89b1c114df62cfd183004f38d18d07f186436c3b37203dd69c3e84becf9d2282"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-stage-evidence.md": {
+            "bytes": 11789,
+            "sha256": "f56e4125325c9c1b9235a1d64fd732861c7686444492422daff44ba8dd648be4"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md": {
+            "bytes": 11860,
+            "sha256": "93aa1b775ced304cebfe6c4dec0503ddb2260ecb3cb01edc17ce329446c5ce2b"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/metadata.json": {
+            "bytes": 552,
+            "sha256": "f6d94f104cb70969024b2f4542078a240e200712c66279eca8b94f60bee1dd71"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md": {
+            "bytes": 8972,
+            "sha256": "0e6f094dc554abea12e07def50f20b233b47127779a8aefad595b4ab1efd24c2"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/metadata.json": {
+            "bytes": 569,
+            "sha256": "acccc8eb86de0df6caf78a9b3c6db76a2a686adc4a5ba0029a4d62d50ff9a8f8"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md": {
+            "bytes": 5989,
+            "sha256": "8428df7157af024fa17cfe0b3818ed54bd9f468306dda46eb8d4f0edf4f3f230"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/metadata.json": {
+            "bytes": 579,
+            "sha256": "c650dce88c07b636d68e063dbda4d319f6f7f3ecbcdd5e47f393cbfc3af7c71d"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/references/source-notes.md": {
+            "bytes": 11470,
+            "sha256": "80cba4d3a884e7ee2618326cf74b9f9c3a34e4d9cf443707f6bbe673f2d5ba65"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/SKILL.md": {
+            "bytes": 6159,
+            "sha256": "23e93ea7c98f99743509fdc04139d78ed2e7123c63e80b366c6beaba7eb70d0d"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/metadata.json": {
+            "bytes": 581,
+            "sha256": "97836a38465106d816a9ace894a7b277d6d330f9accade8f8493ce6f0fb57ef9"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/references/source-notes.md": {
+            "bytes": 13493,
+            "sha256": "435193590b4a2ba32e2e4c6fa9d7e8f889528d412cbbb48e35ec5433f7dab384"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/SKILL.md": {
+            "bytes": 7137,
+            "sha256": "a25738dbb3b9daa15c5827cb536a20fe13f7c8b9dd14aa3fb7cfa7e132879431"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/metadata.json": {
+            "bytes": 1007,
+            "sha256": "3abe626a6e4c73c943429f609ce2c62aa77aae7e1726475ebbbdfbe1794f6f7d"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/mechanisms.md": {
+            "bytes": 17376,
+            "sha256": "fe08b2292fa587be3874774ddf41063b84c579a0085df422ec6710d8aa90b8b7"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 4024,
+            "sha256": "371b01ee201ef51fecbfd6da892d517aaba295930b9395025f03f34e21c030a1"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/NEXT_GATE_LEADS.md": {
+            "bytes": 1623,
+            "sha256": "35b8b58ae5db02b2df51ccbe5cdf870a7383d17f345f9ee5eb7b0a4146dfc74b"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/README.md": {
+            "bytes": 2888,
+            "sha256": "e0b1ec1b39c10067f5ddce58320851ad089568c4b2296a24777389097cb5b388"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/io-reduction-can-plateau.md": {
+            "bytes": 2671,
+            "sha256": "5cc35f514224ac6e6efd2da6db221bba468b2f4db66807295e67b240bcdfdba5"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/overlap-and-fp8-counter-signals.md": {
+            "bytes": 3692,
+            "sha256": "80c8a60a6df55f3d84fab23e03567d8b38e23bcdbd63cc09881e46db1e2888ab"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/partitioning-after-fusion.md": {
+            "bytes": 3132,
+            "sha256": "4d4cabe0991eab554dcfbd4fa9c8199e61fa4767709b0c4cd646e2588bb398a2"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/knowledge/attention-cost-model-boundaries.md": {
+            "bytes": 5223,
+            "sha256": "5ed8f888443e38ccb72545b6025447aab7b92936f78c151fed680e582126239b"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/sources.md": {
+            "bytes": 9069,
+            "sha256": "53436e0de4e6dda2d29426591bf61f8526378cdc6f9ec7967b88b1bcd4c61e6f"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/host-overhead-profiling/SKILL.md": {
+            "bytes": 11199,
+            "sha256": "326e566fbe30d3ba04e4b50c3e87aad9df037f0ac4ad6fe129165ae97dc5f91d"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/host-overhead-profiling/metadata.json": {
+            "bytes": 527,
+            "sha256": "76e34c1c64aba473a9e126b032a776417caa0f913f6ac3cf4dba3a5d123b1ee0"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/SKILL.md": {
+            "bytes": 4905,
+            "sha256": "cec4e3aa1eb64c31c54b5f5ca40af2b61b7919d94be4174968581d5f752c0538"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/metadata.json": {
+            "bytes": 7453,
+            "sha256": "7520da1cc97412631a600957ecb18aa5a7a4461ea21e76c237330180193ed898"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/fp8.md": {
+            "bytes": 6287,
+            "sha256": "31d9c211c11c2224007abbdcd4059913c4a54833fe2d2f926411c5238fcd3501"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/cutlass-bsd.md": {
+            "bytes": 1547,
+            "sha256": "42fec630f410aa308f70a51a89fadcd19586fa620f9831a32bee528a9a10000e"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/deepgemm-mit.md": {
+            "bytes": 1065,
+            "sha256": "6c73a30c14a2938abaa6afe98ebb19e6d81dac5bd4ace0cb05c61c20afa37b85"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/kernelwiki-kda-mit.md": {
+            "bytes": 1063,
+            "sha256": "84aab57158a914f448b38d8be29dd842b634a08c192d844f416a256beed4ae94"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/notices.md": {
+            "bytes": 3240,
+            "sha256": "876ec016644bf5331b3f257985e126c0e61980e3bfd82569b7b0c3f6b2c6c64c"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/resources-and-scheduling.md": {
+            "bytes": 8919,
+            "sha256": "1ce25b426736072435a1ead9a2a8e1f56df5d0abb8198e8c4fdcb34321db4964"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/source-inventory.json": {
+            "bytes": 120506,
+            "sha256": "1dc09c895c970a0c9708158bcc6dd65cdb736274fd2c35d44542c70407d3849d"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/sources.md": {
+            "bytes": 9004,
+            "sha256": "6b74b9e774f2a80db0860510d9e6940d36284430f29c1e124f115a9eaf25833d"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/synchronization.md": {
+            "bytes": 6843,
+            "sha256": "4322f6b3c0e4368bf8c192b4ec752529767f50d19c0c38dad8fdf9cd781e41d9"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/SKILL.md": {
+            "bytes": 7348,
+            "sha256": "fd1f5a7e6a55714730044155f5b38c1d700383d4e16c3e91963bbfbe1c773af5"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/metadata.json": {
+            "bytes": 572,
+            "sha256": "19e40763fedc20990651bb989c32a7693306c5c54af46c3d4be77fd4f87b6c03"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/SKILL.md": {
+            "bytes": 23213,
+            "sha256": "2c807f3df2da2f533d25183de5771f3d412058d16410994292ef8f48c49f9b3a"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/metadata.json": {
+            "bytes": 1684,
+            "sha256": "9ac6af135f0b6d76a76bc201b6bbd8ac9afe5201dd1e2ea82c0ac4826de401da"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-LICENSE.md": {
+            "bytes": 11357,
+            "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-native-offload-readset.json": {
+            "bytes": 6267,
+            "sha256": "cc6c2011be746002ba3ff96ccefab1809863748ddd17eb102b75e7b0e134fe55"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-native-offload.md": {
+            "bytes": 8966,
+            "sha256": "6c2a72b5ba3708a01d229e0576570837a617e6a42dbb0b5312a26a65a6c31542"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/SKILL.md": {
+            "bytes": 16072,
+            "sha256": "36ef88826a32ac3147b869001004ded17a8d5a491e25bfe105005ba39ea9d1eb"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/metadata.json": {
+            "bytes": 539,
+            "sha256": "ae71c389dfdb1a19d267feecb4c6f1fbb4c9afad501ff6d8fab7e38b29bcc37a"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/references/deepinfer-zero-overhead-scheduling.md": {
+            "bytes": 10894,
+            "sha256": "f452bf40132c9195d739f1c86b3aa18857138c7ed0f1861af5853d93b9955924"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/os-process-placement/SKILL.md": {
+            "bytes": 10327,
+            "sha256": "2003492e4d89fde0443b180d74933e764bd1979435e109c8d26e20eb8f6e5053"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/os-process-placement/metadata.json": {
+            "bytes": 1399,
+            "sha256": "ec02995c075f0a540921f57b4846f2cafe334eb90f44389f17c3c50457c3e9c5"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/os-process-placement/references/llama-numa-thread-readset.md": {
+            "bytes": 6617,
+            "sha256": "748b254b3b043a5ddb0f480eb4a566854d27eedd7b111499863e9bf0a5a6951e"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md": {
+            "bytes": 8813,
+            "sha256": "63b883c10a8d3e7d918abfc7175b72e2b401ff1e02697f94b8382ec1c8f369b7"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/metadata.json": {
+            "bytes": 558,
+            "sha256": "0a41c7b00978332b54ee1e4a7c1ade88ecde3e961128f24d1b4f7708caff5ff5"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md": {
+            "bytes": 6402,
+            "sha256": "0a5f48d9352c5bf460775ef47f9136829978811f4a16e84983e299782cae305d"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-serving-preemption/metadata.json": {
+            "bytes": 537,
+            "sha256": "77ce5ba16245cbd80d46d0abdf80aaf26fc0924fa2d2a5624ff430de68a1daf0"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-serving-preemption/references/source-notes.md": {
+            "bytes": 30405,
+            "sha256": "f29fcecfba0490008a354c11b42403983fe165e56ff907ad61cc6040d2b6407f"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/SKILL.md": {
+            "bytes": 3181,
+            "sha256": "1e64f28ea71672fb53c56f6108b87864738c6d03d6dcd32eedaef9e0625e96e3"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/metadata.json": {
+            "bytes": 478,
+            "sha256": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/references/source-notes.md": {
+            "bytes": 10548,
+            "sha256": "61b1609f71c37b61d136b8c4d39f793de87cffb1cf31c8ec737a1c3fa6fb1b4e"
+          },
+          "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/SKILL.md": {
+            "bytes": 3884,
+            "sha256": "ac773c6f1c110a664dfca28a3ae9129b10c07614e60ca3640dea776220e7b352"
+          },
+          "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/metadata.json": {
+            "bytes": 648,
+            "sha256": "261ce9f644ea6eebb90c830f41fb8b320d2b401456e5bd8d494095a44b8620d0"
+          },
+          "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/references/source-notes.md": {
+            "bytes": 8415,
+            "sha256": "9696d3c0ef96b6611787446bd6f96d4d6255ca8d107105de8c1d9af83edb9fa5"
+          },
+          "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/SKILL.md": {
+            "bytes": 6416,
+            "sha256": "6b7db1eb2780dec27858cb7aa0a3c093c5b5dfa4c30261aebfeaa255213ff688"
+          },
+          "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/metadata.json": {
+            "bytes": 582,
+            "sha256": "4dc8659806539b4784ea9348a01d4194ac70d0bb297a916f90ff6419efa5c328"
+          },
+          "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/references/dlrover-persistence-replicas.md": {
+            "bytes": 8398,
+            "sha256": "055f4357f79026fcbcc766bbe77db610376ec353623f8898f5d14a30ee69e7bc"
+          },
+          "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/references/sources.md": {
+            "bytes": 7984,
+            "sha256": "9425b1aab963a2c8d35f3db897bd15ee69699bf5b0d0c8425c3c8bba3e9a0233"
+          },
+          "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/SKILL.md": {
+            "bytes": 3068,
+            "sha256": "25d04aed9782b2a8ff1186828becdc7fce93c4418039ceaef9b8dcc5ec9bcd24"
+          },
+          "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/metadata.json": {
+            "bytes": 478,
+            "sha256": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7"
+          },
+          "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/references/source-notes.md": {
+            "bytes": 8521,
+            "sha256": "9f9184d46b19312138a6e805e439e43dc0594a0cc6b46675255a582dc946ca5b"
+          },
+          "src/inferrsi/data/skill_library/methods/dense-performance-evidence/SKILL.md": {
+            "bytes": 8305,
+            "sha256": "53cc1867a13f54d1006d50517dea8848869ca50df3f0cd610b2424cc8d18c933"
+          },
+          "src/inferrsi/data/skill_library/methods/dense-performance-evidence/metadata.json": {
+            "bytes": 707,
+            "sha256": "a0d1e3b17140295510a764045d1b222e19be3a3406e6d1e7dddaad49ff474490"
+          },
+          "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/SKILL.md": {
+            "bytes": 6667,
+            "sha256": "82d0099d3c33706f46c657d4fa7f76d3d486e019b36760d1de001921328a60cc"
+          },
+          "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/metadata.json": {
+            "bytes": 547,
+            "sha256": "37379a4d4a058ca46da2b20a920dc01be9ec11e557c32b2332702ec76027f4b5"
+          },
+          "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/references/source-notes.md": {
+            "bytes": 44796,
+            "sha256": "ea416831fd7d0874e06d0a76fb0c7196df53072b717673a0cdf635db6bfd917b"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-benchmark-design/SKILL.md": {
+            "bytes": 8974,
+            "sha256": "19ebc1f567aadf49f7a534a18fdb629c72c0e4ce043d360983136491cb72bfab"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-benchmark-design/metadata.json": {
+            "bytes": 5923,
+            "sha256": "18310402a8ade03fd48ac687f79ba9d447871e90337de84c0e1bc15c0b827306"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-benchmark-design/references/benchmark-commands.md": {
+            "bytes": 8466,
+            "sha256": "0898424e0bff3a6246e1eeeda6abd6c5bd974c047fe32f3bbd5ab9d31ce17cd3"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-benchmark-design/references/task-benchmark-selection.md": {
+            "bytes": 25086,
+            "sha256": "0d4bb080a754e26846e81680023db0dff5e6a4c96be4327c411c3b72ac4ef8e2"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/SKILL.md": {
+            "bytes": 6644,
+            "sha256": "89a6a82c692f7f5192f4d7f4c3a5ff08a4df3b1c4bdaa0aafb7e04292d8aa5b4"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/admission-drain-occupancy.json": {
+            "bytes": 1354,
+            "sha256": "3319e23bbe2bd40e2d45fb89dc87bdc92d4df2b5ddd8086fcf84813b43885b7c"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/admission-drain-occupancy.md": {
+            "bytes": 3134,
+            "sha256": "e88bbac5d253a8e7e3a3d5e21047f23f2970c11ac23887d4303ab94fca52dd27"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/drama3-rejected-directions.json": {
+            "bytes": 1459,
+            "sha256": "2cc0a526fe0509551607a03e7626dd3d7844e643c4afb194837f75cddead8b12"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/drama3-rejected-directions.md": {
+            "bytes": 3557,
+            "sha256": "6f143bc363a3290fc008c08e57cb075f1776d1c8cca8a6a59297d5b1fbbf61bf"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-ascend-rejected-directions.json": {
+            "bytes": 1293,
+            "sha256": "b3d538f79c56724ccb2895cbf50cf2e6b0fec3979be62a684a8230cd932dff55"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-ascend-rejected-directions.md": {
+            "bytes": 2564,
+            "sha256": "be396983812262515899a3e2ba4f01b23ff92e1b836881882c7580372b8ae552"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-b300-rejected-directions.json": {
+            "bytes": 1196,
+            "sha256": "146ea1564a46ed03092c7dc5796b318fc05b04b1821105ee00eb7fa7fa1333e2"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-b300-rejected-directions.md": {
+            "bytes": 2656,
+            "sha256": "cce2930f544a0f78f21797fba4d8ded298e4a81f76c34d7977b0e31d361bc738"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/metadata.json": {
+            "bytes": 535,
+            "sha256": "fb55067fb45970319787068a3206110641cceaf0ccbabd706be6d0b3d2afc2ed"
+          },
+          "src/inferrsi/data/skill_library/methods/kernel-writing-guide/SKILL.md": {
+            "bytes": 10029,
+            "sha256": "01fb1712d7eb988c5e412f02340d2e48b4f8564fcf4ab232dc5897869d0d50a0"
+          },
+          "src/inferrsi/data/skill_library/methods/kernel-writing-guide/metadata.json": {
+            "bytes": 653,
+            "sha256": "0189d184337662056fc6952c7beffc1bf6154c4fa173ae3e3e463bd22d70f36c"
+          },
+          "src/inferrsi/data/skill_library/methods/kernel-writing-guide/references/akg-affine-tail-guards.md": {
+            "bytes": 11941,
+            "sha256": "113deb2db10b171db3425bb3f1792bee763706b794c8cc2e222b4cc0d0114118"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/SKILL.md": {
+            "bytes": 12914,
+            "sha256": "882e7850b9b49ad692a2d11618701afe0f4e83727ea78df114be2029923c725a"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/metadata.json": {
+            "bytes": 1403,
+            "sha256": "9f1fd8a2af52bc964021e21f16ecbdb891e08f2057370e700d67f86605d333db"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 3597,
+            "sha256": "157e74ba180426b279b548a729769f152bf18f9a78ffca4ebfcaf26b7d0c5292"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/MECHANISMS.md": {
+            "bytes": 5875,
+            "sha256": "18bb2f5d7e52773c516420a142edfb531d0b9c822a2934d60f3234ac1cb9d40e"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/NEXT_GATE_LEADS.md": {
+            "bytes": 941,
+            "sha256": "16a264c7159f9c2c700ce8ba46e63f3e636b277c987da4970f6d2fa085fcb053"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/knowledge/adaptation-interfaces.md": {
+            "bytes": 3893,
+            "sha256": "5e761957cb5ecd93926e6ffc13deee7c830c6e8f0ee3c3846e9dd3b50dc74446"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/references/source-notes.md": {
+            "bytes": 13370,
+            "sha256": "54d83c3713045d59a48821bcf93598374ba4a630e3891ad39498f919317c4fb8"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/updates/optimization-search-policy.md": {
+            "bytes": 4238,
+            "sha256": "d3db44fa9a6831b86a02c638b323ed943e69fe256866213f10d68c72d762879e"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/SKILL.md": {
+            "bytes": 15324,
+            "sha256": "c5f78961c23e3c8d659f5675b522518ab06e1f2a8ea8bc74c5d428ffb41693d5"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/metadata.json": {
+            "bytes": 3669,
+            "sha256": "60b5b24ead590b0fb222398fba47822bf6723690da683e256a450bb5c0eeb806"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 7349,
+            "sha256": "083e462e0a204dce599362d46712fcfb392d053c5cda56216e22e28d4b8fa0fa"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/MECHANISMS.md": {
+            "bytes": 4003,
+            "sha256": "742c249082ceca230326f4407a40130fe624553b796ae5a157a8a4c15a7e1398"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/README.md": {
+            "bytes": 1204,
+            "sha256": "17b2d1c7c14124d59ac8a0d5ad4200ad2803d991e784a1a98cd97188cc364999"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/READSET.md": {
+            "bytes": 3937,
+            "sha256": "66459ce393858b2b2160f1e2973373bfc4b9902a12644631934b4ebe3140296a"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/STATIC_CHECKS.md": {
+            "bytes": 1111,
+            "sha256": "cb0aa7d148352ebb7b1678fc42a557a099575054716a341dfa25b6a998029064"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/cases/decision-cases.md": {
+            "bytes": 5491,
+            "sha256": "781a675a6533e0a18c044aaaee0d010da7013b6504654b8b8292a35176db801e"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/govern-optimization-experiments/procedure.md": {
+            "bytes": 3823,
+            "sha256": "ce036a4d8a0908da204515c47e2ff9623548d09de264fa65bd58fedce52a78a0"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/govern-optimization-experiments/references/source-notes.md": {
+            "bytes": 21659,
+            "sha256": "f9a8254e45473ddb2b73e576957b6d3db83c62b99dff1922f4d1e606e569559e"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/knowledge/bounds-predictions-and-decisions.md": {
+            "bytes": 5297,
+            "sha256": "5780d6eb5257612ec45deb94f85f334a6ecabf7e466d1678706041f6549bda79"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/redirect-cross-layer-optimization/procedure.md": {
+            "bytes": 3648,
+            "sha256": "6593a63fce9c474b515663a560c94b7df048dc5f6e77a95b8e54e4a64cd10a09"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/redirect-cross-layer-optimization/references/source-notes.md": {
+            "bytes": 26469,
+            "sha256": "6ff3a7bc31f1172db92aedaf7df1f1aac688bf769227b583a4cd82bfad52da67"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/agentic-kernel.md": {
+            "bytes": 41353,
+            "sha256": "7190573b85f94a4391bb0d887ccbd37a31b3f08bb0531114ad7afd72fe1567a7"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/flashvector.md": {
+            "bytes": 9830,
+            "sha256": "3926369acd88ff3c370bee5718e74dd15f3219c3e4f6311095e30e917bffa4cf"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/inferencebench.md": {
+            "bytes": 9294,
+            "sha256": "21f47111197a0b7f8d6384ea918d139dc8318a62ca3313ce060048271f50b639"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/prior-coverage.md": {
+            "bytes": 3254,
+            "sha256": "35d47ac078591221fcb666219bcd824f9400fc17aafecdc0499d8336f7178238"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/vidur.md": {
+            "bytes": 7736,
+            "sha256": "d38602af5219f5feeb49e6cd6bf53d66098c77fad141dae676f1d8a6743cf25d"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/roofline-contract.md": {
+            "bytes": 7091,
+            "sha256": "419ec4bcaec45ac579a133bce083d4f48779f26e664d7a86f720fa65f343b1d1"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/roofline-example.json": {
+            "bytes": 2015,
+            "sha256": "597cd3205df043cff3ae9d15924c77934e8d8eb300c00c4840f6bb2921f90faf"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/scripts/roofline.py": {
+            "bytes": 9004,
+            "sha256": "524de48a6764401695fe1444f1fa3d38f363069b0cd2fba4715a86ca4b28a5d4"
+          },
+          "src/inferrsi/data/skill_library/methods/plan-activation-residency/SKILL.md": {
+            "bytes": 3507,
+            "sha256": "542d8a1e6963f5019986e1fbd79d3a247a045a51487c09313026087ac457c15e"
+          },
+          "src/inferrsi/data/skill_library/methods/plan-activation-residency/metadata.json": {
+            "bytes": 478,
+            "sha256": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7"
+          },
+          "src/inferrsi/data/skill_library/methods/plan-activation-residency/references/source-notes.md": {
+            "bytes": 13907,
+            "sha256": "5ff52655cfd06fcf970ed2a7d4d4fe1c31115b102af2d41497de0c74848ed2b3"
+          },
+          "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/SKILL.md": {
+            "bytes": 13800,
+            "sha256": "7e20d3390f2501263e5e298e026c47cac16c234f2397833131e15ae5d34c0b97"
+          },
+          "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/metadata.json": {
+            "bytes": 1903,
+            "sha256": "d0e4d6950250e4cec43d82e23324c9301a925d4c22be30b4d2713d790cb6504b"
+          },
+          "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/references/capability-presence-audit.md": {
+            "bytes": 11461,
+            "sha256": "c21eb4a6aabe811a49a6540a34c54dbcfdcf7efae3052a79b54034c7838026a5"
+          },
+          "src/inferrsi/data/skill_library/methods/skill-evolution-playbook/SKILL.md": {
+            "bytes": 9448,
+            "sha256": "9cd37f21fe0d862069e4798edca8a04d59f8b840e55d521901594a63dddd9ba3"
+          },
+          "src/inferrsi/data/skill_library/methods/skill-evolution-playbook/metadata.json": {
+            "bytes": 527,
+            "sha256": "92b36338b9257e2e41b95e4370694d7de52ed84bc7971056b3bfced07b0993d8"
+          },
+          "src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md": {
+            "bytes": 6865,
+            "sha256": "64bd2c6b795d966045d1d3cca55976d3d49b08b06df7c7e06a70c2824cf55b7e"
+          },
+          "src/inferrsi/data/skill_library/methods/training-input-pipeline/metadata.json": {
+            "bytes": 587,
+            "sha256": "80659099fce6a6ec05e9084b75452c877a5812e30f60a78f31ef8c7efcdc1bac"
+          },
+          "src/inferrsi/data/skill_library/methods/training-input-pipeline/references/sources.md": {
+            "bytes": 2637,
+            "sha256": "60e0f3e88655d8e1d5eb4797f39ca4370deda01614f265b8d2ea4559d8ac020f"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/SKILL.md": {
+            "bytes": 7200,
+            "sha256": "bdfa5a4ef37d9f303c3a23cd6f98c21137b97ac5ed58e0ed5fac6d71d33e52ef"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/metadata.json": {
+            "bytes": 976,
+            "sha256": "d42b614af9cee6ee0d4a25dc5d57c7b98438feed1e807a7c9f1755109a86e9cf"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/mindspeed-historical.md": {
+            "bytes": 3384,
+            "sha256": "eb97f935ed30517a8fc9edec5e9cddc04f5d679d156c16d75bd29c534593360a"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 7368,
+            "sha256": "c55c6ad2131ef788866411dcc7056f221420ebff30c95c6bc919b09a4818cb73"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/cases/paper-grounded-cases.md": {
+            "bytes": 5649,
+            "sha256": "19bd25b1a944adb2351ebc6b1d795150ea033041a44f6cc261e881677812a101"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/knowledge/offload-delta.md": {
+            "bytes": 2026,
+            "sha256": "f8740fc5c8bbbe5ac13ff1cfec7ca499569d0b46e8cd56cf1f5a44c3a5e72242"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/notes/training-memory-scheduling.md": {
+            "bytes": 18660,
+            "sha256": "9aa6d3e01452dff95c8a8940ff62ead8c146ce215c9175b78c9f2ef508a7b2de"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/references/source-readset.md": {
+            "bytes": 22946,
+            "sha256": "2a5c166767f56fa47e7eba7386f1f36da8cb73702e5b5bbe14d6a456ca9ebfae"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/sources.md": {
+            "bytes": 2752,
+            "sha256": "6d248f58af8b350fc66c22898988d6f451ac60d03bcd805cd6e84df5e55174ac"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/stas-decision-notes.md": {
+            "bytes": 4086,
+            "sha256": "f1a993e6c4ca9a035a2effcaf5500081f227b0cc56cc885a6b22fda0041fc744"
+          },
+          "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md": {
+            "bytes": 25969,
+            "sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213"
+          },
+          "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json": {
+            "bytes": 2418,
+            "sha256": "798cbff6f10b7c24b188080ebf233b188b0e98ef2c6fdb5e00e1fe3eff727c0f"
+          },
+          "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md": {
+            "bytes": 8656,
+            "sha256": "5a39f14b102b0cee304541e011aa31e44e6b13527d1bee9987c3f3b66c064d3d"
+          },
+          "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md": {
+            "bytes": 6392,
+            "sha256": "6c640f28ee3f0253a17b81461c865cb9f947367002cb75bf72550c4572d02e97"
+          },
+          "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/SKILL.md": {
+            "bytes": 7181,
+            "sha256": "a46110891eca860915e314d6921a22856d631453b22b17837fe30262d6780535"
+          },
+          "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/metadata.json": {
+            "bytes": 2175,
+            "sha256": "4e6100dfb9fc7bc00b19915f4fd38ad509704f2a609c658adf4d632d4e05b6c1"
+          },
+          "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/references/afd-plugin-LICENSE.md": {
+            "bytes": 12747,
+            "sha256": "0e8566ea362276878e1f38f97da49964dbea301c15d4aa9d727f97a4c39fd943"
+          },
+          "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/references/source-notes.md": {
+            "bytes": 17240,
+            "sha256": "5073d87450cb60d1b245308dff8807e73749c07de976d50f2fa26694fad81a72"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/LICENSE": {
+            "bytes": 35149,
+            "sha256": "3972dc9744f6499f0f9b2dbf76696f2ae7ad8af9b23dde66d6af86c9dfb36986"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/SKILL.md": {
+            "bytes": 4507,
+            "sha256": "c993bd6ff8e9b5f82f39f175ec8242341136e4ee0808af414e1d94ad3e196c5e"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/metadata.json": {
+            "bytes": 2032,
+            "sha256": "74742851a002a51c208362cdcd2a59fdab5d1734d8492b1f93fbec3cf21e0841"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/references/graph-cache-invalidation.md": {
+            "bytes": 7679,
+            "sha256": "614c9a4e6943d24d7d7be7d632682ce8b1dcc239124a37d9dc1d487113772eac"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/references/model-residency.md": {
+            "bytes": 7191,
+            "sha256": "014c68386e7cdfdb7b4f566836e8b2ff61bd5a5a506e7b4723f83fe80c5d4de7"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/references/repo-provenance.md": {
+            "bytes": 5648,
+            "sha256": "ca996ef515af42a3ce304747fa3f12a31c00533deaace5ba2ce21aefa0a62e2e"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/references/validation-boundaries.md": {
+            "bytes": 3523,
+            "sha256": "1913dedbeb558d872d6a0e015cc11d822910c3cb259a236db29fb60ad8a64263"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md": {
+            "bytes": 12294,
+            "sha256": "4703beb6e5194ef2ff2c0b5ec69f64e35417a4bc0dfb50a10c80a920b77ef3b5"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/metadata.json": {
+            "bytes": 2333,
+            "sha256": "014be24d0e669f4acbc0717dc323eec784026cf185186619d5a19d6d84496131"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/source-notes.md": {
+            "bytes": 12036,
+            "sha256": "fd917bd5ac80bee9c9f4525f8e0300f6d67dc144e2a16632af42a67f3d53d098"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/validation.md": {
+            "bytes": 5402,
+            "sha256": "60e71e70c5519bcdf6c171eec2be2f342e8801bedd746d6005f11adc646dfaff"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/SKILL.md": {
+            "bytes": 11202,
+            "sha256": "f157562d1edc905cb4fc07246567e3c5b844efb9fbf5fe39f0d40553a1b1a307"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/metadata.json": {
+            "bytes": 2156,
+            "sha256": "4d68bf4c1767eaee12ade963bbd551d0e61d1bcd81a984bb02fd3aae46965919"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/references/source-notes.md": {
+            "bytes": 11633,
+            "sha256": "6fa955022f16cd598f44fd6521901d1cb868b06cec545cabe2b82769a641b5d6"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/references/validation.md": {
+            "bytes": 7196,
+            "sha256": "673d93f6700052974b340fab25b3438e0e6344960569ac9abec7d7720f662288"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/LICENSE": {
+            "bytes": 1077,
+            "sha256": "663e51edcd9a3a8f858018ce6924ad2895a24c269c26c39ad75325bcd0f7e5bc"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/SKILL.md": {
+            "bytes": 5432,
+            "sha256": "fdb4d9483c6efa836f9e59617da605fa9c2186e11c2041b05d5c81850ba0475d"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/metadata.json": {
+            "bytes": 4923,
+            "sha256": "fe626a77af3f084f36fd73ffdb6f67976e7dceeafb94ba22078ef03c3a96cfae"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/THIRD_PARTY_NOTICES.md": {
+            "bytes": 1774,
+            "sha256": "a0d77db7d2aca0eda811ab3c386cc32b3143f2c2609a042cf4642954f0f6fb09"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/minimax-h3-apple.md": {
+            "bytes": 12059,
+            "sha256": "57ce01754ba96c378f13a91305ef3d418b926f99b2753f0af99315572bce063a"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/minimax-h3-weight-inventory.json": {
+            "bytes": 10664,
+            "sha256": "fbaacaeec340d3ce893556ee864405a0746af4d13530a38cb3d5aa9d60929271"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/phase-residency.md": {
+            "bytes": 9408,
+            "sha256": "2836a550eb2be7f5a862f51ec7b135ebd9715bf3c4faa84a7ecbc9d9c11f5d8d"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/precision-and-streaming.md": {
+            "bytes": 9440,
+            "sha256": "7e166c89193f4c7aed852ec95346a2de2b5008dac002e2e1862be1b3e57f6093"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/source-identity.json": {
+            "bytes": 7747,
+            "sha256": "730f8dd23648a3f8f47b9726cc613d893a79d1817d515eec609a9b06d8a9ff90"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/source-notes.md": {
+            "bytes": 9448,
+            "sha256": "7304bf9369efc157dc7aa40686dd828cc14db45c49442e908b3d3cc4b329eb86"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/validation.md": {
+            "bytes": 7880,
+            "sha256": "0754873a758e048480938314894e25b504ae0c269549312e7f8e6020d83fe294"
+          },
+          "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/SKILL.md": {
+            "bytes": 9338,
+            "sha256": "95e81b6bc096a954fa73139ea5bc001d1efa6c99747161da62f012b494e5c013"
+          },
+          "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/metadata.json": {
+            "bytes": 1718,
+            "sha256": "1d30b31dc8352fc6a8d7b8abbef8118c840e47e5b90795dae0c758bf9717278c"
+          },
+          "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/references/source-notes.md": {
+            "bytes": 15234,
+            "sha256": "3d12a401e9decb55a2f4f3b2bfa0d360ef5110f63a0469dfe285654e78054190"
+          },
+          "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md": {
+            "bytes": 10812,
+            "sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27"
+          },
+          "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/metadata.json": {
+            "bytes": 572,
+            "sha256": "c221189b8d85fac3bd41ef1c7cd560f1aebc8cfa0381c182d1d571a7169be829"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/SKILL.md": {
+            "bytes": 3788,
+            "sha256": "bf1177cb9985c12a9bf11edb64771018e1e314c4713920de8fd4a989ff70115f"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/metadata.json": {
+            "bytes": 1299,
+            "sha256": "d86cecb615e93ba0f6be09d1ead102cd5a17068c07e1be2549ac5cd8f7519279"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/cache-stop.md": {
+            "bytes": 2026,
+            "sha256": "bba6b5c73261065309cbcf11b1e591eef48916f0c79ceaaf74d98a9fa96b0992"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/generation-batch.md": {
+            "bytes": 1733,
+            "sha256": "f16d3c57d45e63bcaffd3871aa41c76972ee11f7dbf1e12b45d9b170303cad9b"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/source-map.md": {
+            "bytes": 6358,
+            "sha256": "91d26181f246307f2113f94b12ac9587d689b654dcd216e9f14b11cf5ebb650c"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/validation.md": {
+            "bytes": 1665,
+            "sha256": "73b5dd347017baba519ecc85415aead9993c1751d116b9ba4f07689f58c5f783"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/LICENSE": {
+            "bytes": 11352,
+            "sha256": "a54c8553df5d4313d293d1aef3bf2f13952bda8b7c7a1fdd729b2bbe27ee6015"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/SKILL.md": {
+            "bytes": 2304,
+            "sha256": "d75f5b2881b240ec2d9d4f5cc8e3f7d53e4378cb01732a2fb9699da2901ba38c"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/references/repo-provenance.md": {
+            "bytes": 1778,
+            "sha256": "5e8dbe168a8885e209d4b4532944ffa7fead658b23fd6588054d57b22476f63b"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/references/repo-routing-metadata.json": {
+            "bytes": 511,
+            "sha256": "6820707cb9bd877d1635928cf5a0efe38e4fe37dc04036dc70d53e479ed9bf7c"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/references/snapshot.md": {
+            "bytes": 3488,
+            "sha256": "0fb5d776349de845dc314f16ce73a06c9caf5db84dc1c293327d3b3ca1337c80"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-architecture/SKILL.md": {
+            "bytes": 2355,
+            "sha256": "766b2d66d2056dd779abd9e492cb3af6e25899298f44964f565509f130a7cb24"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-architecture/references/runtime.md": {
+            "bytes": 7734,
+            "sha256": "36bba9874fe054acc5217f9c265f5ae309c83943e0d46228370ae1bcdc69421e"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-extension/SKILL.md": {
+            "bytes": 2193,
+            "sha256": "9af38bca840d0004fc2ed049141476fc40fd40dcfa843f38a404a7a314c19425"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-extension/references/integration.md": {
+            "bytes": 5775,
+            "sha256": "6ce9383ca17bacff60a180f8fb8c43ea47eaa37117fb8da6051ad86135d228ef"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-serving/SKILL.md": {
+            "bytes": 2335,
+            "sha256": "990b6bdebfe22612de49dbebcb16810e8abef94d70691aa883b7937c85210eb6"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-serving/references/recipes-and-tuning.md": {
+            "bytes": 8918,
+            "sha256": "94afb762bbe185f58606154b3f0a5af56e7089cb84ea94e26cfcb9f93c95439f"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/LICENSE": {
+            "bytes": 11346,
+            "sha256": "1495e1e757ef4d0925a2350563cf5754bb23c51701a8ec4fb3c5cdcbedae6747"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/SKILL.md": {
+            "bytes": 5466,
+            "sha256": "ae72d1f0cc750699bf79d7a138a97219209bca7b7e5f19cd427d306245eaf841"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/LICENSE": {
+            "bytes": 11346,
+            "sha256": "1495e1e757ef4d0925a2350563cf5754bb23c51701a8ec4fb3c5cdcbedae6747"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/SKILL.md": {
+            "bytes": 5738,
+            "sha256": "b0a4ba4751087126143ef86983798deb0ed00e24b6ff06bf6ef76a9df6e274f2"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/metadata.json": {
+            "bytes": 2844,
+            "sha256": "6ed57dfb4f25ecb34b81f2188b055e05d172ba2581e6a3b605b814b2f733a4d6"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/references/source-notes.md": {
+            "bytes": 19512,
+            "sha256": "d0fcfd1177ae2f351041c5b2abba7f5121cec133006d9e4dca8f74ddb567332d"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/architecture-and-memory.md": {
+            "bytes": 7735,
+            "sha256": "798e747022b7ec135cc4861c2c9ac44fb14b503bb77a0068ee7fbea494b764a7"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/native-candidates.md": {
+            "bytes": 3961,
+            "sha256": "abafcc9a34df196a6365145554887b8d37afb36c6e236c92e50a4a06c8a4ec93"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/parallelism-and-extension.md": {
+            "bytes": 6324,
+            "sha256": "e3be7f4b51264ac68682a64d264e4d96857d5bb9f9d99cae5f327c56984a9e6b"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/repo-provenance.md": {
+            "bytes": 2274,
+            "sha256": "534d95ae8d0b29bcbc8f6707b3d38fb964e31f5b66ba13517a9a6b8f309bbb38"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/repo-routing-metadata.json": {
+            "bytes": 625,
+            "sha256": "b2b879c1cc42ddc2a5b95ae24115eb891562fa9a6e96d151f4ddf6f7719fe73c"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/source-map.md": {
+            "bytes": 8945,
+            "sha256": "c2adad64ceb0c305cdc4859c03012b3e0a4ab26cd0cb289bbd76e192b0cc3794"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/tuning-and-validation.md": {
+            "bytes": 10832,
+            "sha256": "49b3343b3292cc098a787de9d24d968b444d0ef4f73fedb5d8183390e8f39c9c"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-parallel-extension/SKILL.md": {
+            "bytes": 2316,
+            "sha256": "18bd54ebdd629e1ba6dfa589f391c7cc3c5355785c2ae82f741b64ee8022464f"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-runtime-memory/SKILL.md": {
+            "bytes": 2154,
+            "sha256": "33b1a65aba00fca819d2b6cc42202e7708e9155e4d8f4373a2387f4ee2f2fa3e"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-tuning-validation/SKILL.md": {
+            "bytes": 2018,
+            "sha256": "744b2ef9122101d3f50cbbb40fe020e2b3d3888adbe2d86041041e7c7ce424fc"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE": {
+            "bytes": 1062,
+            "sha256": "b53fa08f515cb5a6fff7b9fd8fcd0961a4b80df29d74372d25e1f9171aa042ee"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md": {
+            "bytes": 5504,
+            "sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json": {
+            "bytes": 2202,
+            "sha256": "f7b7581a3f44ba9774480c559408fc66d6a4d6d1dbbeabe4529f1eb7d4dd5ad6"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md": {
+            "bytes": 8136,
+            "sha256": "7b537cb972347dd68e303960f142892346d8b19b884e4d08016ea36a23bdaa7a"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json": {
+            "bytes": 3617,
+            "sha256": "af4830370b6eddc2d176a9ae14da4ab56a90b0fe048a01a89ee051df4365d8e9"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md": {
+            "bytes": 6589,
+            "sha256": "b67affbd7d237d82b7f51be49f1537e9820fe56ec70a71ee76809f9e599ca693"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md": {
+            "bytes": 3394,
+            "sha256": "5dcf5fa7bf734fbe13214f57d2dfb40ff8215d9cd34e072a9e615bf103b82fc2"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/LICENSE": {
+            "bytes": 1093,
+            "sha256": "949b9b0aada7476006469e58ee4ffb0e0f68c2d9137e90a356757b61a09d37f0"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/SKILL.md": {
+            "bytes": 2813,
+            "sha256": "0e5daeb6409c916be45ecd7ce2531ea5f1c60a4a53e2b0ed25dc1e2cdc5c17ed"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/design-space-analysis.md": {
+            "bytes": 27736,
+            "sha256": "11ecafe71530ddd4839268ea1b24b3361a3895012a705859ca06dae29fa72630"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/repo-provenance.md": {
+            "bytes": 2913,
+            "sha256": "05f17d0ae996ad6c12314bdd29f3427574f6aa611ea60effd18598a0f386638b"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/repo-routing-metadata.json": {
+            "bytes": 332,
+            "sha256": "a10ba49b7e6565109b566c105e17551d042075a1ff649c6f371134e3b7402c74"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/trial-format.md": {
+            "bytes": 2291,
+            "sha256": "92d8cf38549268ae4eef382890662d03e18a7dff858ad45ee5eb6f47eb39c8c0"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/troubleshooting.md": {
+            "bytes": 2079,
+            "sha256": "7301cfa5a5f6cfbf9576c3845a96c6b4de8f8a6509f2330970a8e2e4669b3d5a"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/scripts/summarize_trials.py": {
+            "bytes": 5170,
+            "sha256": "6a243a08da15659ced6519f49f52a49a675bd9e952740c3ab03dd21964eefbe2"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/design-space/SKILL.md": {
+            "bytes": 1826,
+            "sha256": "857bea3c64db45dc035728a41523142ac65778b41f63092148b8ab936f971d76"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/design-space/references/design-space.md": {
+            "bytes": 24231,
+            "sha256": "0a428dcd12d5290b554e9da6795a110f94d98f128771602e1ea5f608048b135e"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/tune-execution/SKILL.md": {
+            "bytes": 1145,
+            "sha256": "cdf51c537e7237296d52ef6cda9f2e4c9d9ce3fc87265a369cd9eb7497c53512"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/tune-execution/references/tuning.md": {
+            "bytes": 11310,
+            "sha256": "7b69aa7872d60dd3fb91e41b77a632572070dcf80dd5b5f6dc1722d8540177b1"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/validate-engine/SKILL.md": {
+            "bytes": 4613,
+            "sha256": "90bfd975e25f18895ce0e85f6f35d24149ce026305a4d479aea4924510053460"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/validate-engine/references/validation.md": {
+            "bytes": 16895,
+            "sha256": "dae846f4c47c68731b96ec1d50cb008c03871c72078f1c5b0f4a9423be9ba397"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/LICENSE": {
+            "bytes": 11357,
+            "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/SKILL.md": {
+            "bytes": 4945,
+            "sha256": "df3e63dc28de2dac2aa92824fd88719159138650d96487332e0effca3f0e2a80"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/references/evidence.md": {
+            "bytes": 4689,
+            "sha256": "ae0d51ba56b34ed81fba34de8fd3685bebf2f160b188fa53719936002881d220"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/references/repo-provenance.md": {
+            "bytes": 2221,
+            "sha256": "43bbfe431434ca4ad6157ba35cbb9ce2d89e4065f427abfdd5f372f51ba0c9c9"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/references/repo-routing-metadata.json": {
+            "bytes": 342,
+            "sha256": "7784b4d98d096718c4f4cc42ce86d5ec113c469844e12de4690b0fccca59d329"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-architecture/SKILL.md": {
+            "bytes": 1336,
+            "sha256": "fb3ff50a7b9122a2545a97ef5179011237c4c4ca214dde7236ad7bf79a7c8623"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-architecture/references/architecture.md": {
+            "bytes": 5600,
+            "sha256": "db845eb4207ddeedb8ae381f1b91704a42cb53ba0ee6c7ee581510ba1a412190"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-serving/SKILL.md": {
+            "bytes": 1343,
+            "sha256": "8ab3209b838635395b22deca6aa879172016531d45c2374eb8e498e7f71bb40f"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-serving/references/serving.md": {
+            "bytes": 5049,
+            "sha256": "9bd5eff300d9755137a7600c8f7a7753bb22d09c8853e08062d7e535662ba8a3"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/SKILL.md": {
+            "bytes": 2227,
+            "sha256": "93c83bc4b799b32f50ff54b870cbbde050a790af16d9b7b062039c4987c80148"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/references/abort-acknowledgment.md": {
+            "bytes": 8547,
+            "sha256": "e95ae5b2cc410917ddf5e9066890d0bef9264104529bdca5ac12bdf56fe80ae3"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/references/validation.md": {
+            "bytes": 4632,
+            "sha256": "1542061f9c76fff2acb6529a58238301484b0e23ec0c697d2fdf129e200ea973"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/LICENSE": {
+            "bytes": 11324,
+            "sha256": "50e6751797c50dedd75ef1b8a0d9e42f5f8472e9fbce91f34718e9f97b0c780a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/SKILL.md": {
+            "bytes": 2863,
+            "sha256": "38e522b8abeb0dc140a2194340432ed41dc8b569d33b63d5d4c59aa1092cd2d1"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/SKILL.md": {
+            "bytes": 6664,
+            "sha256": "313733a65cc48442b18dda1c4665139a689963a90e1032a2235bc6f40e3a8e06"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/metadata.json": {
+            "bytes": 781,
+            "sha256": "b43fc3a7ccae59c664fcdf9ff77962edcc07c2a3df8b59324db551642bd23e23"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/references/source-notes.md": {
+            "bytes": 13853,
+            "sha256": "3665b034594fc8cc767c408002f4e16f8803369242405027c7f37ce6c9ecebe9"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/SKILL.md": {
+            "bytes": 6565,
+            "sha256": "5fb9ea1d1373e110d9b506f69963c4857338b7dbf5e38ec9a49b59a0ff683ed0"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/metadata.json": {
+            "bytes": 1372,
+            "sha256": "78459416734f61a8faa819d9a1a3ad7b9a985e8bf8e149813c604d15be6bb92f"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 9006,
+            "sha256": "82b24be82451ee121892f61575d0cf4ecb9920813a25ccb26cce2b45dda6d8b3"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/migration-is-not-free-memory.md": {
+            "bytes": 2145,
+            "sha256": "2ac168f2725cba7fc92dfcc67c01bf2258dff6a2dbe1ddb063bc60c175d82d0a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/parallelism-change-with-state-cost.md": {
+            "bytes": 2069,
+            "sha256": "5accc81086e4500d66f894ef6af9951e1cbb76941f6c631aab5c5561a786d8d8"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/preemption-under-kv-pressure.md": {
+            "bytes": 2630,
+            "sha256": "462200a6d0b193ef01f32374a463e0f2838ca8edf33d9ab898def65ffd0b490a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/knowledge/scheduling-state-boundaries.md": {
+            "bytes": 7488,
+            "sha256": "d4ef29869959b835a5dcb06ddb2fd8e6cff49bb176994dc1d46eb9a852f6228d"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/fastserve.md": {
+            "bytes": 9289,
+            "sha256": "9b4e275843000ef453d508d9c3c9c3f8ea1de5545fbd01d7abe74baa8a4931a7"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/llumnix.md": {
+            "bytes": 10181,
+            "sha256": "f252ce33ac2db017701ca1d49813c5ec1996c4c92900cf35bb6bda7fc59c0f18"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/loongserve.md": {
+            "bytes": 12422,
+            "sha256": "7bc6d1e256d461b04acc8ed856cf99108faceda9cde729ab95bcca596fe019ca"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/orca.md": {
+            "bytes": 8826,
+            "sha256": "f5a3b5870d64226408a8fcda5d69483769d823f8a5efd7f11698c8da483ab195"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/fastserve-readset.md": {
+            "bytes": 7173,
+            "sha256": "9866bd0c78b2b638998b3b632f6bae26aef032462c81abee336d309af5b07831"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/llumnix-readset.md": {
+            "bytes": 9386,
+            "sha256": "3a2819d65ad24087c08530492ef39d8ea046ed1f780d53512f1d23e2b3578487"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/loongserve-readset.md": {
+            "bytes": 12490,
+            "sha256": "f1a383e7bdbbd5a5a472d7f6e1292c8a17d4910ff736ca5a6aa422e1bee74f8c"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/orca-readset.md": {
+            "bytes": 4796,
+            "sha256": "a497ef23b7e60c0fbc15d95b7e07cd20547cfe56f37c24dfe6ed70d79ef7a3cf"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/source-notes.md": {
+            "bytes": 9575,
+            "sha256": "84033fd514773f10c0bddd9914fabb980497ba2b308f226a922c3e1888413827"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/references/evidence-map.json": {
+            "bytes": 14384,
+            "sha256": "dd8bbab5d285ed3a1a19d2f06ad84e8c444251690d8f673def3e2681765ef6c1"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/references/repo-provenance.json": {
+            "bytes": 1313,
+            "sha256": "4e705a964697712e950e02bedc41320806a41311b7f4dcf32f923d60b1d2e919"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/references/repo-provenance.md": {
+            "bytes": 3261,
+            "sha256": "6e841c38c35ff2cd6ff8139cb1a6c8f75701647897f1089a94a5c62928609941"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/references/repo-routing-metadata.json": {
+            "bytes": 432,
+            "sha256": "a20b54f682eb52311e445ac62b95d1935c001296f5cf23b7764cb04477d535c9"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/SKILL.md": {
+            "bytes": 1835,
+            "sha256": "2a945b5fddacfb90d1cd7cb2b2f767067d641061423a65e1c19afecec7f3c06a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/references/capability-boundaries.md": {
+            "bytes": 6560,
+            "sha256": "c8081d03f04f7481080c46c7dcf728962729bc122f49f9502d50cd1e3d04bcf3"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/references/runtime-and-invariants.md": {
+            "bytes": 8265,
+            "sha256": "afa1f967545021f36bc1b59dbf6aa2c367f4d5ece4f08dce755e2017cf50dbe5"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/extend-validate/SKILL.md": {
+            "bytes": 1731,
+            "sha256": "495679e5b99341bf80daeaacd2d4c2699aba16bc16c447cbd8d2fe17f7559798"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/extend-validate/references/change-boundaries-and-validation.md": {
+            "bytes": 7624,
+            "sha256": "f7fdd2d699f09d96fbf5ca9382f9b56113c870db2688d8e1118eb8163519fc4a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/SKILL.md": {
+            "bytes": 1812,
+            "sha256": "0ef4d2e3a96cbb277557e169a3fc8edf0f0504aad03259771b432a33ee008f78"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/references/commands-and-measurements.md": {
+            "bytes": 5667,
+            "sha256": "76ab51083b14595c49fa1c255cb7e5ebb996f14930a51ab3cbe864cc334b5f80"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/references/experiment-decisions.md": {
+            "bytes": 7644,
+            "sha256": "242593c0d386e57104b57a09bc7db97c2bbbcdb905bb68a397cfe612e07fb083"
+          },
+          "src/inferrsi/data/skill_library/skill-index.json": {
+            "bytes": 7514,
+            "sha256": "8ce35e18f7629260d67a95617b9918eecff73ed8d9370dd3d4595c725ada9aa1"
+          },
+          "src/inferrsi/data/skill_library/workflows/cite-check-downstream-pr/SKILL.md": {
+            "bytes": 1093,
+            "sha256": "485932cdf3b1fba5dc69be73e75d38f86a7b79442d26e82a296c963a7598d226"
+          },
+          "src/inferrsi/data/skill_library/workflows/claim-contract/SKILL.md": {
+            "bytes": 1075,
+            "sha256": "3af1dbbe81d91cb05fc2ef15a2ae25d7592a08558e1a5535f49c52f33a488c55"
+          },
+          "src/inferrsi/data/skill_library/workflows/competitor-scan/SKILL.md": {
+            "bytes": 1088,
+            "sha256": "0ede58ace65a9324a538b4816add1684f230ba536d4364fe5a5dd2c0bbed3776"
+          },
+          "src/inferrsi/data/skill_library/workflows/draft-acceptance-issue/SKILL.md": {
+            "bytes": 1230,
+            "sha256": "ce983acfb7d7d3272775e7cf00c92e4ddadd97511b602575a76e5ab7ade01b56"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/SKILL.md": {
+            "bytes": 5681,
+            "sha256": "364beed2ca617e71f4f955bbd024740d411b94eeb9557d7e055d4463cc22caba"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/cross-repository-mechanisms.md": {
+            "bytes": 17050,
+            "sha256": "28efa1b25f548763180b70d8d0ddebb7d4e7227df189f871fd233fe1d4881303"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/import.md": {
+            "bytes": 3463,
+            "sha256": "c8a57db94768cc6cd4193932def75ba67ec453e56ee0e54f176fb82f1af53091"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/model-aware-preparation.md": {
+            "bytes": 9056,
+            "sha256": "787c017aade68ee1133b979d636fa24062302cef6174c001171c963947ee8da4"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/overview.md": {
+            "bytes": 8099,
+            "sha256": "7ed0a966239b9481b82dd349245cfbfe966c030fc916771c91fd0692464ffc56"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/repository-source-scope.md": {
+            "bytes": 11967,
+            "sha256": "85202ca209171c50573c2fc209acb1e937d8211e1652e4ca9c6cc4bce04d03eb"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/source-manifest.json": {
+            "bytes": 1613,
+            "sha256": "0ce1678803d1ede6b485fb6b4b1c88aba909056ee6e3b5af691899ceb2056240"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/SKILL.md": {
+            "bytes": 6502,
+            "sha256": "74cd863df2e80dc2bda1713d24207809173a3116f30325e3487e1e3517cfec1e"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/metadata.json": {
+            "bytes": 3236,
+            "sha256": "12fc9272bd5fbf02dd9b897d8ca40984b72ae54f20492ace5304711bbc4d71c3"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/COMPARISON.md": {
+            "bytes": 4267,
+            "sha256": "38587678a96581f5c2dc697d982127343565504bffa191352811dffcc165aa23"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 5557,
+            "sha256": "fbc4a96b6f6c477706eed83eaa4fcc29a14f881637629d6c004601ed049f2523"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/QA.md": {
+            "bytes": 1853,
+            "sha256": "3f3785ec39a6938a5d1f2e874e99153fe92c0e666dd3e78a499c8b47d891346a"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/README.md": {
+            "bytes": 2151,
+            "sha256": "221785e91c45588f82b3caab6d52f1ea1d102b16139c20a0802921e2789e6d28"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/UNREAD_LEADS.md": {
+            "bytes": 1038,
+            "sha256": "6c1d87ebe491d27666703149eb05ad582c7e5aff77f0ac5c11a7b89c946f96ee"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/cases/methodology-counterexamples.md": {
+            "bytes": 5213,
+            "sha256": "f9fc21753ea41eef77ca020d78d031e945e5dd835b5d09c5ff0e53749ceb2ca4"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/knowledge/evidence-routing.md": {
+            "bytes": 5282,
+            "sha256": "406d5d74c3fef4c1a23d495492e21a084c3c46495a1a84b385703ed104358874"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/learning-to-commit.md": {
+            "bytes": 2847,
+            "sha256": "9c2d1d738e896d05c00f53e6adf14c9df2501ecc3f0020e71b5a1c2192547086"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/protocol-details.md": {
+            "bytes": 6520,
+            "sha256": "41298c6bde3a28f0712fa06a18bbc4ac59aea6d17cc9eb1318744f4121e3bab6"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/readset.md": {
+            "bytes": 7086,
+            "sha256": "6f953d20d8958c68339c9ceeae348604d0fea3e9c6ccb9b64a63d3b210426a22"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/repo-to-skill.md": {
+            "bytes": 3151,
+            "sha256": "5ebdac2a2bbbe6614f9d4096bc9895bd685acdf76414e2c3c1fb7b50d57c738d"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/skillopt.md": {
+            "bytes": 3040,
+            "sha256": "148075daa67e19f4b4351710498438d17c279a9bce56b2d01cab26ff3be33e99"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/source-manifest.json": {
+            "bytes": 2431,
+            "sha256": "716cd49ed01824ee9f6b6fc55ebf2b24fdd5d5a3e677f6e28e34a472616bbe25"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/trace2skill.md": {
+            "bytes": 2917,
+            "sha256": "230d13b77ba819b5e3edee147523e8cad6bfce95afccc140fb4946d6c8747a2e"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/updates/bounded-updates.md": {
+            "bytes": 7630,
+            "sha256": "d02c5bd0109846d1e1987de074810321cb4af277cd3f66ab2ac268aadd5ee142"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/skill-factory-design-20261014.md": {
+            "bytes": 9465,
+            "sha256": "fb8ce1ad83b7a6ddc657e89ee2eebb710873b213cfd85bdd6d823c8691b3cb9f"
+          },
+          "src/inferrsi/data/skill_library/workflows/rsi-evidence/SKILL.md": {
+            "bytes": 3680,
+            "sha256": "a55610874b658b86a39a1a049efb4e4eab0140f6579ed74dd7718a4ca8cedaca"
+          }
+        },
+        "sha256": "6e19365a0ad8e2951af0f3106f93aa545b4ad00b6e0f0317a20f32065b6a05fb"
+      },
+      "environment": {
+        "filesystem_cache": "not controlled; process cold is not OS-cache cold",
+        "implementation": "CPython",
+        "logical_cpu_count": 8,
+        "machine": "AMD64",
+        "perf_counter_resolution_seconds": 1e-07,
+        "python_executable_sha256": "737a7e3b71e3578f8432acc7dd88c452e593622c544bc13da4789d69c63da5ae",
+        "python_version": "3.12.6",
+        "release": "10",
+        "site_initializers_disabled": true,
+        "system": "Windows"
+      },
+      "git_head": "4817f92121df73fd9cd5fb4a319e3c56a68b4a5b",
+      "knowledge_record_count": 108,
+      "library": {
+        "bytes": 2294958,
+        "file_count": 342,
+        "sha256": "be8726296e6186601692dc2d95d99bfbba76f0b4c758de3a88a9b2a523eaa37e"
+      },
+      "modules": {
+        "inferrsi": {
+          "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+          "source": "src/inferrsi/__init__.py"
+        },
+        "inferrsi.catalog": {
+          "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+          "source": "src/inferrsi/catalog.py"
+        },
+        "inferrsi.cli": {
+          "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+          "source": "src/inferrsi/cli.py"
+        },
+        "inferrsi.knowledge_operations": {
+          "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+          "source": "src/inferrsi/knowledge_operations.py"
+        }
+      },
+      "packaged_skill_count": 60,
+      "python_sources": {
+        "bytes": 274444,
+        "file_count": 53,
+        "files": {
+          "src/inferrsi/__init__.py": {
+            "bytes": 102,
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47"
+          },
+          "src/inferrsi/__main__.py": {
+            "bytes": 48,
+            "sha256": "935a1c1166b0c1ea35a82256345000bf2c73ded718d77773bc27a71ecce28f7d"
+          },
+          "src/inferrsi/_skill_selection.py": {
+            "bytes": 5790,
+            "sha256": "4640399f4013425589f6702e2db844820619ba35967d1c3519978eb307168440"
+          },
+          "src/inferrsi/arguments.py": {
+            "bytes": 12200,
+            "sha256": "4456600cf4a362cf069594fe36db46932cf5cb14ddd37683aaf3538f5cf76b4d"
+          },
+          "src/inferrsi/catalog.py": {
+            "bytes": 6855,
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5"
+          },
+          "src/inferrsi/claim.py": {
+            "bytes": 805,
+            "sha256": "5f24c4fa91b48f0bb4d4554531607763fc8fc108e90f332816f06cfb520b5cc1"
+          },
+          "src/inferrsi/cli.py": {
+            "bytes": 14205,
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88"
+          },
+          "src/inferrsi/contracts.py": {
+            "bytes": 119,
+            "sha256": "03000cb9246ce8c0967147533f7ad25df918ba2674839801453ffb870f496490"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/scripts/roofline.py": {
+            "bytes": 9004,
+            "sha256": "524de48a6764401695fe1444f1fa3d38f363069b0cd2fba4715a86ca4b28a5d4"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/scripts/summarize_trials.py": {
+            "bytes": 5170,
+            "sha256": "6a243a08da15659ced6519f49f52a49a675bd9e952740c3ab03dd21964eefbe2"
+          },
+          "src/inferrsi/evaluation.py": {
+            "bytes": 183,
+            "sha256": "482c40710d428206c59fe3696703248cd86c4b3c7d19ad41017c1369bd6d780e"
+          },
+          "src/inferrsi/evidence_paths.py": {
+            "bytes": 859,
+            "sha256": "cc46231e0b8334ef5742defdda20e0093e22b30c86af59df1fb3ae4b5caa70fa"
+          },
+          "src/inferrsi/execution.py": {
+            "bytes": 755,
+            "sha256": "f43d8cbb4d43a72a7ff8a10b29a5fb9cf4cb22af54dd35943283c843fba0da92"
+          },
+          "src/inferrsi/harness/__init__.py": {
+            "bytes": 177,
+            "sha256": "1006e4713003f44bfb32b2e385a918e1652b9e9833801e4555674a71e57fb0d5"
+          },
+          "src/inferrsi/harness/evaluation.py": {
+            "bytes": 4160,
+            "sha256": "9b00b2e93959f16b7b76bb4b257df7dcf48520a06fd6adae96e9a165a05e6e80"
+          },
+          "src/inferrsi/harness/improvement.py": {
+            "bytes": 5777,
+            "sha256": "3faa0f5928a707a98f35f45d1ca18e226c898941e6a07b3829eebeb53e70aeb8"
+          },
+          "src/inferrsi/harness/reporting.py": {
+            "bytes": 3566,
+            "sha256": "38c58f02c44e8c8aeacf46527e142be0e3142a83e4b2693246d0ca76c0ddd7c3"
+          },
+          "src/inferrsi/harness/runner.py": {
+            "bytes": 6567,
+            "sha256": "a566307b9b586655941b627f428b5121ed38d7ac80bccd1e1799e443fb0de911"
+          },
+          "src/inferrsi/harness/scenarios/__init__.py": {
+            "bytes": 79,
+            "sha256": "4e97913c806077f0cadfeb29d986d2fa75b55ab8766cb86e6e3e9a195a9a1f5a"
+          },
+          "src/inferrsi/harness/scenarios/external.py": {
+            "bytes": 1016,
+            "sha256": "84d5f6920501a9d882cc4fffbc721d2051fdbf7c90694bde9cbfba3c2b866304"
+          },
+          "src/inferrsi/harness/scenarios/media.py": {
+            "bytes": 2914,
+            "sha256": "67d979a3ed6f13a8e3e1e40a304f78e5a3e6c2556b3aa8a11c41156745d7050a"
+          },
+          "src/inferrsi/harness/skill_evaluation.py": {
+            "bytes": 11093,
+            "sha256": "b710b8e53758757efd68c703449eb4bab7eb7ec50ae3dfc52a010cdebd8c3bf5"
+          },
+          "src/inferrsi/infra/__init__.py": {
+            "bytes": 85,
+            "sha256": "8779897733779266841b0f79b84b1dc43f0fa57c013855c906c809829ecfc77c"
+          },
+          "src/inferrsi/infra/artifacts.py": {
+            "bytes": 1040,
+            "sha256": "e72e1bebdbb88e42de430238489252386851f9e6645ed7780f2aad2e90338120"
+          },
+          "src/inferrsi/infra/campaign.py": {
+            "bytes": 1028,
+            "sha256": "be505fba1592611bd84ceaca55afbac027205413a2639baefa0c3bb343230209"
+          },
+          "src/inferrsi/infra/commands.py": {
+            "bytes": 811,
+            "sha256": "1890f115b3a2c5c6832173bf60432ad595bd6bf1c69783ecd460c6f8961194f1"
+          },
+          "src/inferrsi/infra/environment.py": {
+            "bytes": 2067,
+            "sha256": "60e46c2aefc216db202b41f44542e7b893e08e0721c667b9ccc50411d1ac84ee"
+          },
+          "src/inferrsi/infra/executors/__init__.py": {
+            "bytes": 64,
+            "sha256": "327c3cd5c821c832ff260a41dba0ee1e439cdabefc9edfb9cad1cd38d14bc87a"
+          },
+          "src/inferrsi/infra/executors/local.py": {
+            "bytes": 3654,
+            "sha256": "37bcbca9f3261c8c58a8bf806c9da40720b18d0676c82e8ed24c3901ac9bbe77"
+          },
+          "src/inferrsi/infra/http.py": {
+            "bytes": 634,
+            "sha256": "f65d6c2b55b6a89694caaf6dcdf7d74f417e9ced103eca1b3f218e305c6b6cb2"
+          },
+          "src/inferrsi/infra/inspection.py": {
+            "bytes": 2415,
+            "sha256": "d1c66deb0cced529bca996d2f88b60f51fa3417ff1cae365c650fec808946ecd"
+          },
+          "src/inferrsi/infra/platforms/__init__.py": {
+            "bytes": 1494,
+            "sha256": "9e88e8b9a396ed90c221e1585135c0a1b58da41c1bb73ed9237c53b6684abfed"
+          },
+          "src/inferrsi/infra/platforms/cuda.py": {
+            "bytes": 3309,
+            "sha256": "9e0e4276e1b2984605493b14b82fba3e86f6635204719c318302e624f66f8b6f"
+          },
+          "src/inferrsi/infra/platforms/mlx.py": {
+            "bytes": 1971,
+            "sha256": "f3b81470db28bd6183ccca2e763c5e079686bba6cdfe6c08ebe2cdf64216f2e3"
+          },
+          "src/inferrsi/infra/processes.py": {
+            "bytes": 8151,
+            "sha256": "3b96ebc7a50913e78e86ff3dcba74c8ee8d03d547d9591ca6488f390ac1e6189"
+          },
+          "src/inferrsi/infra/targets/__init__.py": {
+            "bytes": 186,
+            "sha256": "1f3ea75dded7c6cc967754c7b6d4f30ef00da8191a749938aabc3c6bb8f86eb3"
+          },
+          "src/inferrsi/infra/targets/git.py": {
+            "bytes": 2910,
+            "sha256": "6327c4f126086f310b6de309fc7659c9759da28ed76efaeeb94a3d2fba29a0c2"
+          },
+          "src/inferrsi/inspection.py": {
+            "bytes": 692,
+            "sha256": "afa44bf4cdfbb58a318c21f9b292ca142d1bbb22750e35a4cc8244da5350ebb3"
+          },
+          "src/inferrsi/knowledge_operations.py": {
+            "bytes": 27444,
+            "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b"
+          },
+          "src/inferrsi/knowledge_store.py": {
+            "bytes": 13700,
+            "sha256": "5e794e28906ce4687aff7f14a507e267184640fc8a4fac19e0ced279ce6a6cb6"
+          },
+          "src/inferrsi/model/__init__.py": {
+            "bytes": 77,
+            "sha256": "3c1dba4f5159fd9fbc8d4d96e2a83c8441491c1b1dc7b42031a3fc1782ed9d9f"
+          },
+          "src/inferrsi/model/anthropic_agent.py": {
+            "bytes": 24696,
+            "sha256": "b0ef9cd542544292eeb90ab05ef410be9ed6e5d1a40d729890f49562c28c12f0"
+          },
+          "src/inferrsi/model/protocols/__init__.py": {
+            "bytes": 85,
+            "sha256": "5a2f54483b164de04e309dbebb87a700b213682b0f44a50d89a752dd71a4ce47"
+          },
+          "src/inferrsi/model/protocols/image.py": {
+            "bytes": 2047,
+            "sha256": "251a1f8f885da2cd8e9d3baba766c525afe293091606b4e8668c43ba2cb7cb5f"
+          },
+          "src/inferrsi/model/protocols/metrics.py": {
+            "bytes": 610,
+            "sha256": "17db16d6ef15b41fb60d8f09507893accc43f9cb3ad686b8dc4e9839713ebc56"
+          },
+          "src/inferrsi/model/validators/__init__.py": {
+            "bytes": 65,
+            "sha256": "1d4b6ffa968e283219ddcc8a09de9280238a8d65ae2dc8cd2b60889877a8eb0b"
+          },
+          "src/inferrsi/model/validators/flare.py": {
+            "bytes": 19975,
+            "sha256": "3696b0609c772f2222b448833608565a033123d7e92b11fbc545dd356fec590f"
+          },
+          "src/inferrsi/model/validators/image.py": {
+            "bytes": 3204,
+            "sha256": "79470b6b06a22d53ec6c40636a0b3d821a2e6ee2f7bb7aed3dac7241504add1a"
+          },
+          "src/inferrsi/reporting.py": {
+            "bytes": 127,
+            "sha256": "1f771890ab3e69b1aa9246b548a7825596bdf904f6c395dd218208084266853d"
+          },
+          "src/inferrsi/research.py": {
+            "bytes": 5213,
+            "sha256": "9bfe59dd17b5476c18ce50d671711053c62800521c50d6b710e8140a2b4dca04"
+          },
+          "src/inferrsi/runspec.py": {
+            "bytes": 4398,
+            "sha256": "dca0952425caf42095ef0c7e2feb8ef4740cc422ecb61e777f203962d9b87709"
+          },
+          "src/inferrsi/skill_catalog.py": {
+            "bytes": 9938,
+            "sha256": "65f5a1f3f6894009abb98ec2fec30a63cc1bb2f6ce6aeb01f9f7a3dce0439f2d"
+          },
+          "src/inferrsi/summary_chart.py": {
+            "bytes": 40910,
+            "sha256": "d29023d60f6be804c1b01c9eb73765d4824d6c3bb870fe22c99a24544bc1d8bc"
+          }
+        },
+        "sha256": "4faa529dc47f23c24f8c5b8e0110d8e413aab20df5ad61db591fb751564a6448"
+      },
+      "tracked_source_dirty": false
+    }
+  },
+  "fixture": {
+    "cases": 2,
+    "answerable": 2,
+    "negative": 0,
+    "fixture_sha256": "47909db5996a971c61953d8b996a43a8432b89835a0509cb0e909e11f94cb3ff"
+  },
+  "variants": {
+    "A": {
+      "modules": {
+        "inferrsi": {
+          "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+          "source": "src/inferrsi/__init__.py"
+        },
+        "inferrsi.catalog": {
+          "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+          "source": "src/inferrsi/catalog.py"
+        },
+        "inferrsi.cli": {
+          "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+          "source": "src/inferrsi/cli.py"
+        },
+        "inferrsi.knowledge_operations": {
+          "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+          "source": "src/inferrsi/knowledge_operations.py"
+        }
+      },
+      "outputs": [
+        {
+          "answerable": true,
+          "case": "dev01",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev01",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 428778,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_device_health",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.dcp_async_save_performance",
+                "ref": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "resource_count": 343,
+                "resource_hashes_sha256": "b59637d16a7076477f6011fcceeb3cb5b402aa95af17b841ac9b2ede396216e3",
+                "revision": "sha256:36243cf65b6eaa5f370fd4f7633b8fccb119ff75ed38ba652a6061469de0b5b3"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.checkpoint_recovery_consistency",
+                "ref": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "15c906f896eab2a681eb9bc9e42ee29d11b0594a1e1aeacbf6f414233209d1fa",
+                "revision": "sha256:73f77f89e298ce8724258cb3d74315438c036297bb38c63aac3d7dc95473ce53"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.inference_benchmark_design",
+                "ref": "package:inferrsi.common.skills.inference_benchmark_design",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb743a7fe0662feead981795dc60e2a5c9355ed876339f14ad7a47caab0bb670",
+                "revision": "sha256:6b0cbeaa5d409c04228883884e192f275cad4874c7f6b3039abef9fd443fafb5"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.kv_tier_management",
+                "ref": "package:inferrsi.common.skills.kv_tier_management",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ee42b55fa8a404149ecd1dccc32a892566ee8867710531bc0af553409758a019",
+                "revision": "sha256:5816c2e526a71152486af2230ace24d27b54baeb373bf5d27e353ccd44f640aa"
+              }
+            ],
+            "selection": {
+              "cue": "Why must checkpoint upload finish before cached staging buffers can be reused?",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 434223,
+            "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev01",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 64465,
+            "context_omissions": [
+              {
+                "available_bytes": 11883,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "required_bytes": 14400
+              },
+              {
+                "available_bytes": 11883,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.inference_benchmark_design",
+                "required_bytes": 47014
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md"
+                  }
+                ],
+                "content_bytes": 12636,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md",
+                    "role": "core",
+                    "sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md",
+                    "role": "reference",
+                    "sha256": "22afe4f4d77f441096cc657d55d763473bfcbea8fb531c8e493fe49ab84955f7"
+                  }
+                ],
+                "revision": "sha256:36243cf65b6eaa5f370fd4f7633b8fccb119ff75ed38ba652a6061469de0b5b3",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md"
+                  }
+                ],
+                "content_bytes": 41017,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Normalized provenance covers the existing llama.cpp loading slice at 3e390af18de1131dda27913c00511725b4274c8a and the fastsafetensors distributed-delivery slice at 219e2561a41ede2d5d8eba611c248bb0effbf580; other inherited claims remain unnormalized with original identities and limits retained.",
+                  "The readsets record static mechanisms and source recommendations, not validated startup, memory, placement or backend performance.",
+                  "The fastsafetensors delivery readset does not revalidate inherited I/O/save claims or qualify uneven scatter, accelerator/NPU compatibility, inter-node transport or agent efficacy.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/weight-loading-optimization/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                    "role": "core",
+                    "sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                    "role": "reference",
+                    "sha256": "6c640f28ee3f0253a17b81461c865cb9f947367002cb75bf72550c4572d02e97"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                    "role": "reference",
+                    "sha256": "5a39f14b102b0cee304541e011aa31e44e6b13527d1bee9987c3f3b66c064d3d"
+                  }
+                ],
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md"
+                  }
+                ],
+                "content_bytes": 10812,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/repositories/lmcache/kv-tier-management/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.kv_tier_management",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md",
+                    "role": "core",
+                    "sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27"
+                  }
+                ],
+                "revision": "sha256:5816c2e526a71152486af2230ace24d27b54baeb373bf5d27e353ccd44f640aa",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 458836,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_device_health",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.dcp_async_save_performance",
+                "ref": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "resource_count": 343,
+                "resource_hashes_sha256": "b59637d16a7076477f6011fcceeb3cb5b402aa95af17b841ac9b2ede396216e3",
+                "revision": "sha256:36243cf65b6eaa5f370fd4f7633b8fccb119ff75ed38ba652a6061469de0b5b3"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.checkpoint_recovery_consistency",
+                "ref": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "15c906f896eab2a681eb9bc9e42ee29d11b0594a1e1aeacbf6f414233209d1fa",
+                "revision": "sha256:73f77f89e298ce8724258cb3d74315438c036297bb38c63aac3d7dc95473ce53"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.inference_benchmark_design",
+                "ref": "package:inferrsi.common.skills.inference_benchmark_design",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb743a7fe0662feead981795dc60e2a5c9355ed876339f14ad7a47caab0bb670",
+                "revision": "sha256:6b0cbeaa5d409c04228883884e192f275cad4874c7f6b3039abef9fd443fafb5"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.kv_tier_management",
+                "ref": "package:inferrsi.common.skills.kv_tier_management",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ee42b55fa8a404149ecd1dccc32a892566ee8867710531bc0af553409758a019",
+                "revision": "sha256:5816c2e526a71152486af2230ace24d27b54baeb373bf5d27e353ccd44f640aa"
+              }
+            ],
+            "selection": {
+              "cue": "Why must checkpoint upload finish before cached staging buffers can be reused?",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 540055,
+            "sha256": "1afecd2e68d87dd1ef1e3c1c22ea846fd33c26bd751560219948aae98577f481"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev01",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev02",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev02",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 85545,
+            "no_match": false,
+            "omissions": [],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.dcp_async_save_performance",
+                "ref": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "resource_count": 343,
+                "resource_hashes_sha256": "b59637d16a7076477f6011fcceeb3cb5b402aa95af17b841ac9b2ede396216e3",
+                "revision": "sha256:36243cf65b6eaa5f370fd4f7633b8fccb119ff75ed38ba652a6061469de0b5b3"
+              }
+            ],
+            "selection": {
+              "cue": "DefaultStager",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 85990,
+            "sha256": "27b25a4ec6fa85219601eeb68b85fd54f3dc51c0d529a480f782ec7102a49684"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev02",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 12636,
+            "context_omissions": [],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md"
+                  }
+                ],
+                "content_bytes": 12636,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md",
+                    "role": "core",
+                    "sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md",
+                    "role": "reference",
+                    "sha256": "22afe4f4d77f441096cc657d55d763473bfcbea8fb531c8e493fe49ab84955f7"
+                  }
+                ],
+                "revision": "sha256:36243cf65b6eaa5f370fd4f7633b8fccb119ff75ed38ba652a6061469de0b5b3",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 89815,
+            "no_match": false,
+            "omissions": [],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.dcp_async_save_performance",
+                "ref": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "resource_count": 343,
+                "resource_hashes_sha256": "b59637d16a7076477f6011fcceeb3cb5b402aa95af17b841ac9b2ede396216e3",
+                "revision": "sha256:36243cf65b6eaa5f370fd4f7633b8fccb119ff75ed38ba652a6061469de0b5b3"
+              }
+            ],
+            "selection": {
+              "cue": "DefaultStager",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 106414,
+            "sha256": "672f7b868df0011c492de557341606d2b17761814ad5c5c5b8bfc5107699cd12"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev02",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        }
+      ]
+    },
+    "B": {
+      "modules": {
+        "inferrsi": {
+          "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+          "source": "src/inferrsi/__init__.py"
+        },
+        "inferrsi.catalog": {
+          "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+          "source": "src/inferrsi/catalog.py"
+        },
+        "inferrsi.cli": {
+          "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+          "source": "src/inferrsi/cli.py"
+        },
+        "inferrsi.knowledge_operations": {
+          "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+          "source": "src/inferrsi/knowledge_operations.py"
+        }
+      },
+      "outputs": [
+        {
+          "answerable": true,
+          "case": "dev01",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev01",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 428778,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_device_health",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.dcp_async_save_performance",
+                "ref": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "resource_count": 343,
+                "resource_hashes_sha256": "b59637d16a7076477f6011fcceeb3cb5b402aa95af17b841ac9b2ede396216e3",
+                "revision": "sha256:36243cf65b6eaa5f370fd4f7633b8fccb119ff75ed38ba652a6061469de0b5b3"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.checkpoint_recovery_consistency",
+                "ref": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "15c906f896eab2a681eb9bc9e42ee29d11b0594a1e1aeacbf6f414233209d1fa",
+                "revision": "sha256:73f77f89e298ce8724258cb3d74315438c036297bb38c63aac3d7dc95473ce53"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.inference_benchmark_design",
+                "ref": "package:inferrsi.common.skills.inference_benchmark_design",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb743a7fe0662feead981795dc60e2a5c9355ed876339f14ad7a47caab0bb670",
+                "revision": "sha256:6b0cbeaa5d409c04228883884e192f275cad4874c7f6b3039abef9fd443fafb5"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.kv_tier_management",
+                "ref": "package:inferrsi.common.skills.kv_tier_management",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ee42b55fa8a404149ecd1dccc32a892566ee8867710531bc0af553409758a019",
+                "revision": "sha256:5816c2e526a71152486af2230ace24d27b54baeb373bf5d27e353ccd44f640aa"
+              }
+            ],
+            "selection": {
+              "cue": "Why must checkpoint upload finish before cached staging buffers can be reused?",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 434223,
+            "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev01",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 64465,
+            "context_omissions": [
+              {
+                "available_bytes": 11883,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "required_bytes": 14400
+              },
+              {
+                "available_bytes": 11883,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.inference_benchmark_design",
+                "required_bytes": 47014
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md"
+                  }
+                ],
+                "content_bytes": 12636,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md",
+                    "role": "core",
+                    "sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md",
+                    "role": "reference",
+                    "sha256": "22afe4f4d77f441096cc657d55d763473bfcbea8fb531c8e493fe49ab84955f7"
+                  }
+                ],
+                "revision": "sha256:36243cf65b6eaa5f370fd4f7633b8fccb119ff75ed38ba652a6061469de0b5b3",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md"
+                  }
+                ],
+                "content_bytes": 41017,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Normalized provenance covers the existing llama.cpp loading slice at 3e390af18de1131dda27913c00511725b4274c8a and the fastsafetensors distributed-delivery slice at 219e2561a41ede2d5d8eba611c248bb0effbf580; other inherited claims remain unnormalized with original identities and limits retained.",
+                  "The readsets record static mechanisms and source recommendations, not validated startup, memory, placement or backend performance.",
+                  "The fastsafetensors delivery readset does not revalidate inherited I/O/save claims or qualify uneven scatter, accelerator/NPU compatibility, inter-node transport or agent efficacy.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/weight-loading-optimization/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                    "role": "core",
+                    "sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                    "role": "reference",
+                    "sha256": "6c640f28ee3f0253a17b81461c865cb9f947367002cb75bf72550c4572d02e97"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                    "role": "reference",
+                    "sha256": "5a39f14b102b0cee304541e011aa31e44e6b13527d1bee9987c3f3b66c064d3d"
+                  }
+                ],
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md"
+                  }
+                ],
+                "content_bytes": 10812,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/repositories/lmcache/kv-tier-management/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.kv_tier_management",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md",
+                    "role": "core",
+                    "sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27"
+                  }
+                ],
+                "revision": "sha256:5816c2e526a71152486af2230ace24d27b54baeb373bf5d27e353ccd44f640aa",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 458836,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_device_health",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.dcp_async_save_performance",
+                "ref": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "resource_count": 343,
+                "resource_hashes_sha256": "b59637d16a7076477f6011fcceeb3cb5b402aa95af17b841ac9b2ede396216e3",
+                "revision": "sha256:36243cf65b6eaa5f370fd4f7633b8fccb119ff75ed38ba652a6061469de0b5b3"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.checkpoint_recovery_consistency",
+                "ref": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "15c906f896eab2a681eb9bc9e42ee29d11b0594a1e1aeacbf6f414233209d1fa",
+                "revision": "sha256:73f77f89e298ce8724258cb3d74315438c036297bb38c63aac3d7dc95473ce53"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.inference_benchmark_design",
+                "ref": "package:inferrsi.common.skills.inference_benchmark_design",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb743a7fe0662feead981795dc60e2a5c9355ed876339f14ad7a47caab0bb670",
+                "revision": "sha256:6b0cbeaa5d409c04228883884e192f275cad4874c7f6b3039abef9fd443fafb5"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.kv_tier_management",
+                "ref": "package:inferrsi.common.skills.kv_tier_management",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ee42b55fa8a404149ecd1dccc32a892566ee8867710531bc0af553409758a019",
+                "revision": "sha256:5816c2e526a71152486af2230ace24d27b54baeb373bf5d27e353ccd44f640aa"
+              }
+            ],
+            "selection": {
+              "cue": "Why must checkpoint upload finish before cached staging buffers can be reused?",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 540055,
+            "sha256": "1afecd2e68d87dd1ef1e3c1c22ea846fd33c26bd751560219948aae98577f481"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev01",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev02",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev02",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 85545,
+            "no_match": false,
+            "omissions": [],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.dcp_async_save_performance",
+                "ref": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "resource_count": 343,
+                "resource_hashes_sha256": "b59637d16a7076477f6011fcceeb3cb5b402aa95af17b841ac9b2ede396216e3",
+                "revision": "sha256:36243cf65b6eaa5f370fd4f7633b8fccb119ff75ed38ba652a6061469de0b5b3"
+              }
+            ],
+            "selection": {
+              "cue": "DefaultStager",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 85990,
+            "sha256": "27b25a4ec6fa85219601eeb68b85fd54f3dc51c0d529a480f782ec7102a49684"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev02",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 12636,
+            "context_omissions": [],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md"
+                  }
+                ],
+                "content_bytes": 12636,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md",
+                    "role": "core",
+                    "sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md",
+                    "role": "reference",
+                    "sha256": "22afe4f4d77f441096cc657d55d763473bfcbea8fb531c8e493fe49ab84955f7"
+                  }
+                ],
+                "revision": "sha256:36243cf65b6eaa5f370fd4f7633b8fccb119ff75ed38ba652a6061469de0b5b3",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 89815,
+            "no_match": false,
+            "omissions": [],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.dcp_async_save_performance",
+                "ref": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "resource_count": 343,
+                "resource_hashes_sha256": "b59637d16a7076477f6011fcceeb3cb5b402aa95af17b841ac9b2ede396216e3",
+                "revision": "sha256:36243cf65b6eaa5f370fd4f7633b8fccb119ff75ed38ba652a6061469de0b5b3"
+              }
+            ],
+            "selection": {
+              "cue": "DefaultStager",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 106414,
+            "sha256": "672f7b868df0011c492de557341606d2b17761814ad5c5c5b8bfc5107699cd12"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev02",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        }
+      ]
+    }
+  },
+  "metrics": {
+    "A": {
+      "development/search": {
+        "answerable_count": 2,
+        "success_at_5": 0,
+        "mrr_at_5": 0,
+        "negative_count": 0,
+        "negative_empty_count": 0
+      },
+      "development/relevance": {
+        "answerable_count": 2,
+        "success_at_5": 1,
+        "mrr_at_5": 1.0,
+        "negative_count": 0,
+        "negative_empty_count": 0
+      },
+      "development/context_material": {
+        "answerable_count": 2,
+        "success_at_5": 1,
+        "mrr_at_5": 1.0,
+        "negative_count": 0,
+        "negative_empty_count": 0
+      },
+      "development/default_diagnostic": {
+        "answerable_count": 2,
+        "success_at_5": 0,
+        "mrr_at_5": 0,
+        "negative_count": 0,
+        "negative_empty_count": 0
+      },
+      "all/search": {
+        "answerable_count": 2,
+        "success_at_5": 0,
+        "mrr_at_5": 0,
+        "negative_count": 0,
+        "negative_empty_count": 0
+      },
+      "all/relevance": {
+        "answerable_count": 2,
+        "success_at_5": 1,
+        "mrr_at_5": 1.0,
+        "negative_count": 0,
+        "negative_empty_count": 0
+      },
+      "all/context_material": {
+        "answerable_count": 2,
+        "success_at_5": 1,
+        "mrr_at_5": 1.0,
+        "negative_count": 0,
+        "negative_empty_count": 0
+      },
+      "all/default_diagnostic": {
+        "answerable_count": 2,
+        "success_at_5": 0,
+        "mrr_at_5": 0,
+        "negative_count": 0,
+        "negative_empty_count": 0
+      }
+    },
+    "B": {
+      "development/search": {
+        "answerable_count": 2,
+        "success_at_5": 0,
+        "mrr_at_5": 0,
+        "negative_count": 0,
+        "negative_empty_count": 0
+      },
+      "development/relevance": {
+        "answerable_count": 2,
+        "success_at_5": 1,
+        "mrr_at_5": 1.0,
+        "negative_count": 0,
+        "negative_empty_count": 0
+      },
+      "development/context_material": {
+        "answerable_count": 2,
+        "success_at_5": 1,
+        "mrr_at_5": 1.0,
+        "negative_count": 0,
+        "negative_empty_count": 0
+      },
+      "development/default_diagnostic": {
+        "answerable_count": 2,
+        "success_at_5": 0,
+        "mrr_at_5": 0,
+        "negative_count": 0,
+        "negative_empty_count": 0
+      },
+      "all/search": {
+        "answerable_count": 2,
+        "success_at_5": 0,
+        "mrr_at_5": 0,
+        "negative_count": 0,
+        "negative_empty_count": 0
+      },
+      "all/relevance": {
+        "answerable_count": 2,
+        "success_at_5": 1,
+        "mrr_at_5": 1.0,
+        "negative_count": 0,
+        "negative_empty_count": 0
+      },
+      "all/context_material": {
+        "answerable_count": 2,
+        "success_at_5": 1,
+        "mrr_at_5": 1.0,
+        "negative_count": 0,
+        "negative_empty_count": 0
+      },
+      "all/default_diagnostic": {
+        "answerable_count": 2,
+        "success_at_5": 0,
+        "mrr_at_5": 0,
+        "negative_count": 0,
+        "negative_empty_count": 0
+      }
+    }
+  },
+  "parity_failures": [],
+  "completed_at_utc": "2026-10-27T11:12:19.079172+00:00"
+}
diff --git a/evidence/knowledge-retrieval-20261027/results/environment.md b/evidence/knowledge-retrieval-20261027/results/environment.md
new file mode 100644
index 00000000..a041f14c
--- /dev/null
+++ b/evidence/knowledge-retrieval-20261027/results/environment.md
@@ -0,0 +1,35 @@
+# Local retrieval experiment environment
+
+Hardware/OS inventory was read locally by the accountable orchestrator:
+
+| Field | Observed inventory |
+| --- | --- |
+| CPU | Intel Core i7-8565U @ 1.80 GHz |
+| Cores | 4 physical, 8 logical |
+| RAM | 15.8 GiB |
+| OS | Windows 10 Pro, 10.0.19045 |
+
+The processor string's frequency is inventory metadata, not proof of a fixed
+operating clock. Clock/power state and unrelated background load are not held
+constant by this evaluator. Twenty sequential paired groups alternate AB/BA;
+there are no concurrent baseline/candidate workers. The formal benchmark starts
+after software tests, quality comparison and diagram rendering finish. Their
+durations are excluded from the latency observations.
+
+Each result's identity section is authoritative for the actual Python version,
+implementation, executable-binary digest, source revision, imported modules,
+optimization flags and full source/corpus inventories. Both variants must use
+the same interpreter and complete corpus. No personal path, hostname or network
+address is needed in these shared records.
+
+The completed formal runs used CPython 3.12.6 with site initializers disabled
+and optimization flag zero on both sides. Their snapshots cover 53 Python
+source files and 458 corpus files / 2,883,003 bytes, including 108 knowledge
+records and the 342-file / 2,294,958-byte shared library.
+
+Fresh-reader API measurements construct a new KnowledgeReader; reused-reader
+measurements follow the frozen warmup. Every CLI measurement starts a fresh
+Python process. The OS filesystem cache is uncontrolled and is not claimed
+cold. API wall time, full CLI startup/capture and output decoding are distinct
+measurements. This CPU-only experiment runs no model or inference service and
+establishes no GPU, service, task-quality or production-distribution benefit.
diff --git a/evidence/knowledge-retrieval-20261027/results/independent-review.md b/evidence/knowledge-retrieval-20261027/results/independent-review.md
new file mode 100644
index 00000000..4bd18d0d
--- /dev/null
+++ b/evidence/knowledge-retrieval-20261027/results/independent-review.md
@@ -0,0 +1,22 @@
+# Independent read-only review record
+
+The accountable orchestrator coordinated independent read-only reviewers and
+relayed the findings recorded here. Their model/effort settings are **unknown**;
+they made no material repository edits and are not listed as material authors.
+This record is not human/maintainer goal acceptance or merge/closure authority.
+
+| Scope | Observed review result |
+| --- | --- |
+| Validator before formal measurement | Review found positional truncation could hide missing rows and Python assertions could be disabled. Authors corrected both before candidate measurement; v1 freeze and the prospective amendment remain preserved. The current v2 freeze is `a465367319f3058398eb1f8579ee5d4174b8cb90457793e90788fb5796151057`. |
+| Runtime/integrity boundary | Review retained per-call mutation/revision validation and the bounded digest/path reuse scope. Relevant regression checks passed; full Windows suite failures and their baseline reproduction remain qualified. |
+| Quality | Review checked 96 exact outputs per revision, unchanged source/corpus identity and item metrics. It separately found labelled-owner contexts present for 19/20 positives, including the 170,782-byte required set omitted against the 65,536-byte budget. |
+| Benchmark | Independent recomputation reproduced 20 pairs, ten AB/ten BA, 640 API calls, 160 CLI processes, all 400 cross-version output pairs and 80 exact stdout pairs, all 48 side-specific p50/p95 groups, and the exact 40.0068267% reduction / [38.3789227%, 41.0592071%] bootstrap interval. All 53 source and 458 corpus file identities matched. |
+| Regressions and interpretation | Review required retaining API/CLI search p50 regressions and CLI relevance decode p95 1.810→3.046 ms (+1.236 ms), with no inferred cause or blanket tail-latency claim. Each group has 20 repeats; four hot query types remain four query types. |
+| Diagrams | Final editable EN/ZH diagrams were reviewed. The English required-set budget wording was corrected; final previews were clear and the Chinese figure content was unchanged. |
+| Evidence registration | The minimal nested bundle hash selector retained safe-path/digest checks. Correct/tampered/missing/nonobject field regressions passed; inventory check reported 9 bundles/120 hashes and no errors. |
+
+The frozen software parity and local CPU performance gates passed. The local
+whole-suite run remains failed with its initial failures/errors reproduced on
+baseline. Remote CI is pending at PR creation, and human/maintainer acceptance
+has not been recorded. Goal acceptance, merge and Issue closure remain with the
+authorized requester/repository maintainers.
diff --git a/evidence/knowledge-retrieval-20261027/results/latency-summary.json b/evidence/knowledge-retrieval-20261027/results/latency-summary.json
new file mode 100644
index 00000000..22efe429
--- /dev/null
+++ b/evidence/knowledge-retrieval-20261027/results/latency-summary.json
@@ -0,0 +1,533 @@
+{
+  "schema_version": 1,
+  "source": "benchmark.json",
+  "source_sha256": "85cf49d3bc96df35763f3997abd5af72a5f580787fe952420a56e764fbb5bc89",
+  "analysis": "Post-measurement display of every frozen per-mode/query summary; no new acceptance gate.",
+  "positive_delta_means": "candidate slower",
+  "rows": [
+    {
+      "interface": "api",
+      "mode": "cold_context_material",
+      "case": "dev09",
+      "timer": "call_ns",
+      "n_baseline": 20,
+      "n_candidate": 20,
+      "baseline_ns": {
+        "p50": 1060605750.0,
+        "p95": 1570935585.0
+      },
+      "candidate_ns": {
+        "p50": 686351850.0,
+        "p95": 885685090.0000001
+      },
+      "delta_percent": {
+        "p50": -35.286806619707654,
+        "p95": -43.62053425634252
+      }
+    },
+    {
+      "interface": "api",
+      "mode": "cold_default_diagnostic",
+      "case": "dev05",
+      "timer": "call_ns",
+      "n_baseline": 20,
+      "n_candidate": 20,
+      "baseline_ns": {
+        "p50": 999688350.0,
+        "p95": 1657808180.0000002
+      },
+      "candidate_ns": {
+        "p50": 614150600.0,
+        "p95": 741096665.0
+      },
+      "delta_percent": {
+        "p50": -38.56579402970935,
+        "p95": -55.29659740248116
+      }
+    },
+    {
+      "interface": "api",
+      "mode": "cold_relevance",
+      "case": "dev01",
+      "timer": "call_ns",
+      "n_baseline": 20,
+      "n_candidate": 20,
+      "baseline_ns": {
+        "p50": 1141533250.0,
+        "p95": 1427484050.0000002
+      },
+      "candidate_ns": {
+        "p50": 788233350.0,
+        "p95": 1146733305.0000002
+      },
+      "delta_percent": {
+        "p50": -30.94959345249032,
+        "p95": -19.66752237967212
+      }
+    },
+    {
+      "interface": "api",
+      "mode": "cold_relevance",
+      "case": "dev05",
+      "timer": "call_ns",
+      "n_baseline": 20,
+      "n_candidate": 20,
+      "baseline_ns": {
+        "p50": 1061825450.0,
+        "p95": 1247372300.0000002
+      },
+      "candidate_ns": {
+        "p50": 683019350.0,
+        "p95": 817369650.0000001
+      },
+      "delta_percent": {
+        "p50": -35.674987824034545,
+        "p95": -34.47267908706968
+      }
+    },
+    {
+      "interface": "api",
+      "mode": "cold_relevance",
+      "case": "dev09",
+      "timer": "call_ns",
+      "n_baseline": 20,
+      "n_candidate": 20,
+      "baseline_ns": {
+        "p50": 1083633800.0,
+        "p95": 1901819180.0
+      },
+      "candidate_ns": {
+        "p50": 680550900.0,
+        "p95": 872077780.0000001
+      },
+      "delta_percent": {
+        "p50": -37.197335483629246,
+        "p95": -54.145073876055875
+      }
+    },
+    {
+      "interface": "api",
+      "mode": "cold_relevance",
+      "case": "dev13",
+      "timer": "call_ns",
+      "n_baseline": 20,
+      "n_candidate": 20,
+      "baseline_ns": {
+        "p50": 1057574450.0,
+        "p95": 1766673930.0
+      },
+      "candidate_ns": {
+        "p50": 677725500.0,
+        "p95": 844747565.0000001
+      },
+      "delta_percent": {
+        "p50": -35.91699383433479,
+        "p95": -52.184296680033086
+      }
+    },
+    {
+      "interface": "api",
+      "mode": "hot_context_material",
+      "case": "dev09",
+      "timer": "call_ns",
+      "n_baseline": 20,
+      "n_candidate": 20,
+      "baseline_ns": {
+        "p50": 944589000.0,
+        "p95": 1495627210.0000002
+      },
+      "candidate_ns": {
+        "p50": 577179800.0,
+        "p95": 780719880.0
+      },
+      "delta_percent": {
+        "p50": -38.896197182054834,
+        "p95": -47.7998344253178
+      }
+    },
+    {
+      "interface": "api",
+      "mode": "hot_default_diagnostic",
+      "case": "dev05",
+      "timer": "call_ns",
+      "n_baseline": 20,
+      "n_candidate": 20,
+      "baseline_ns": {
+        "p50": 882934200.0,
+        "p95": 1529218095.0
+      },
+      "candidate_ns": {
+        "p50": 504027700.0,
+        "p95": 640953050.0000001
+      },
+      "delta_percent": {
+        "p50": -42.91446633282525,
+        "p95": -58.086223796612856
+      }
+    },
+    {
+      "interface": "api",
+      "mode": "hot_relevance",
+      "case": "dev01",
+      "timer": "call_ns",
+      "n_baseline": 20,
+      "n_candidate": 20,
+      "baseline_ns": {
+        "p50": 1021204450.0,
+        "p95": 1317269300.0
+      },
+      "candidate_ns": {
+        "p50": 663072600.0,
+        "p95": 910342455.0000002
+      },
+      "delta_percent": {
+        "p50": -35.069554387468635,
+        "p95": -30.89169731656236
+      }
+    },
+    {
+      "interface": "api",
+      "mode": "hot_relevance",
+      "case": "dev05",
+      "timer": "call_ns",
+      "n_baseline": 20,
+      "n_candidate": 20,
+      "baseline_ns": {
+        "p50": 938520600.0,
+        "p95": 1226803705.0
+      },
+      "candidate_ns": {
+        "p50": 579671050.0,
+        "p95": 879715645.0
+      },
+      "delta_percent": {
+        "p50": -38.23566046392588,
+        "p95": -28.292061605731778
+      }
+    },
+    {
+      "interface": "api",
+      "mode": "hot_relevance",
+      "case": "dev09",
+      "timer": "call_ns",
+      "n_baseline": 20,
+      "n_candidate": 20,
+      "baseline_ns": {
+        "p50": 957841900.0,
+        "p95": 1702202280.000001
+      },
+      "candidate_ns": {
+        "p50": 577390550.0,
+        "p95": 725926025.0000001
+      },
+      "delta_percent": {
+        "p50": -39.71963953550163,
+        "p95": -57.35371562303396
+      }
+    },
+    {
+      "interface": "api",
+      "mode": "hot_relevance",
+      "case": "dev13",
+      "timer": "call_ns",
+      "n_baseline": 20,
+      "n_candidate": 20,
+      "baseline_ns": {
+        "p50": 949684850.0,
+        "p95": 1935473355.0
+      },
+      "candidate_ns": {
+        "p50": 566259750.0,
+        "p95": 806470585.0
+      },
+      "delta_percent": {
+        "p50": -40.37393036226702,
+        "p95": -58.33212671636081
+      }
+    },
+    {
+      "interface": "api",
+      "mode": "search",
+      "case": "dev01",
+      "timer": "call_ns",
+      "n_baseline": 20,
+      "n_candidate": 20,
+      "baseline_ns": {
+        "p50": 31409350.0,
+        "p95": 54011865.0
+      },
+      "candidate_ns": {
+        "p50": 31781600.0,
+        "p95": 53469575.0
+      },
+      "delta_percent": {
+        "p50": 1.185156649214325,
+        "p95": -1.0040201352054745
+      }
+    },
+    {
+      "interface": "api",
+      "mode": "search",
+      "case": "dev05",
+      "timer": "call_ns",
+      "n_baseline": 20,
+      "n_candidate": 20,
+      "baseline_ns": {
+        "p50": 31934550.0,
+        "p95": 39200650.00000001
+      },
+      "candidate_ns": {
+        "p50": 31307300.0,
+        "p95": 38680315.00000001
+      },
+      "delta_percent": {
+        "p50": -1.9641735988138231,
+        "p95": -1.3273631942327446
+      }
+    },
+    {
+      "interface": "api",
+      "mode": "search",
+      "case": "dev09",
+      "timer": "call_ns",
+      "n_baseline": 20,
+      "n_candidate": 20,
+      "baseline_ns": {
+        "p50": 31589500.0,
+        "p95": 54718180.0
+      },
+      "candidate_ns": {
+        "p50": 31505500.0,
+        "p95": 46758085.00000001
+      },
+      "delta_percent": {
+        "p50": -0.2659111413602644,
+        "p95": -14.547441088135592
+      }
+    },
+    {
+      "interface": "api",
+      "mode": "search",
+      "case": "dev13",
+      "timer": "call_ns",
+      "n_baseline": 20,
+      "n_candidate": 20,
+      "baseline_ns": {
+        "p50": 31384950.0,
+        "p95": 65095300.000000015
+      },
+      "candidate_ns": {
+        "p50": 31158650.0,
+        "p95": 40072955.0
+      },
+      "delta_percent": {
+        "p50": -0.7210462339433343,
+        "p95": -38.43955708015787
+      }
+    },
+    {
+      "interface": "cli",
+      "mode": "context_material",
+      "case": "dev09",
+      "timer": "decode_ns",
+      "n_baseline": 20,
+      "n_candidate": 20,
+      "baseline_ns": {
+        "p50": 3289450.0,
+        "p95": 5639460.000000002
+      },
+      "candidate_ns": {
+        "p50": 3295700.0,
+        "p95": 4242245.000000001
+      },
+      "delta_percent": {
+        "p50": 0.19000136800984002,
+        "p95": -24.77568774315272
+      }
+    },
+    {
+      "interface": "cli",
+      "mode": "context_material",
+      "case": "dev09",
+      "timer": "process_ns",
+      "n_baseline": 20,
+      "n_candidate": 20,
+      "baseline_ns": {
+        "p50": 1413646050.0,
+        "p95": 2644763040.0
+      },
+      "candidate_ns": {
+        "p50": 1017285750.0,
+        "p95": 1423809670.0000002
+      },
+      "delta_percent": {
+        "p50": -28.03815707616486,
+        "p95": -46.164943759952116
+      }
+    },
+    {
+      "interface": "cli",
+      "mode": "default_diagnostic",
+      "case": "dev05",
+      "timer": "decode_ns",
+      "n_baseline": 20,
+      "n_candidate": 20,
+      "baseline_ns": {
+        "p50": 22100.0,
+        "p95": 30240.000000000004
+      },
+      "candidate_ns": {
+        "p50": 22000.0,
+        "p95": 32300.0
+      },
+      "delta_percent": {
+        "p50": -0.4524886877828038,
+        "p95": 6.812169312169303
+      }
+    },
+    {
+      "interface": "cli",
+      "mode": "default_diagnostic",
+      "case": "dev05",
+      "timer": "process_ns",
+      "n_baseline": 20,
+      "n_candidate": 20,
+      "baseline_ns": {
+        "p50": 1613022250.0,
+        "p95": 2133127665.0
+      },
+      "candidate_ns": {
+        "p50": 1103380950.0,
+        "p95": 1382574130.0000002
+      },
+      "delta_percent": {
+        "p50": -31.595429015315812,
+        "p95": -35.18558909131207
+      }
+    },
+    {
+      "interface": "cli",
+      "mode": "relevance",
+      "case": "dev05",
+      "timer": "decode_ns",
+      "n_baseline": 20,
+      "n_candidate": 20,
+      "baseline_ns": {
+        "p50": 1531700.0,
+        "p95": 1809890.0000000005
+      },
+      "candidate_ns": {
+        "p50": 1536950.0,
+        "p95": 3046270.0000000005
+      },
+      "delta_percent": {
+        "p50": 0.34275641444148075,
+        "p95": 68.31243887750082
+      }
+    },
+    {
+      "interface": "cli",
+      "mode": "relevance",
+      "case": "dev05",
+      "timer": "process_ns",
+      "n_baseline": 20,
+      "n_candidate": 20,
+      "baseline_ns": {
+        "p50": 1413155950.0,
+        "p95": 1718849305.0000005
+      },
+      "candidate_ns": {
+        "p50": 998828450.0,
+        "p95": 1369453170.0000002
+      },
+      "delta_percent": {
+        "p50": -29.31930478019783,
+        "p95": -20.327327938733998
+      }
+    },
+    {
+      "interface": "cli",
+      "mode": "search",
+      "case": "dev05",
+      "timer": "decode_ns",
+      "n_baseline": 20,
+      "n_candidate": 20,
+      "baseline_ns": {
+        "p50": 12400.0,
+        "p95": 15865.000000000002
+      },
+      "candidate_ns": {
+        "p50": 12700.0,
+        "p95": 14095.000000000002
+      },
+      "delta_percent": {
+        "p50": 2.4193548387096753,
+        "p95": -11.156634100220607
+      }
+    },
+    {
+      "interface": "cli",
+      "mode": "search",
+      "case": "dev05",
+      "timer": "process_ns",
+      "n_baseline": 20,
+      "n_candidate": 20,
+      "baseline_ns": {
+        "p50": 340560150.0,
+        "p95": 487368400.0
+      },
+      "candidate_ns": {
+        "p50": 349066900.0,
+        "p95": 451238360.00000024
+      },
+      "delta_percent": {
+        "p50": 2.497870053204987,
+        "p95": -7.413291464936944
+      }
+    }
+  ],
+  "paired_groups": 20,
+  "api_measurements": 640,
+  "cli_processes": 160,
+  "measurement_loop_elapsed_seconds": 836.1620858,
+  "primary": {
+    "paired_median_reduction": 0.40006826681186614,
+    "confidence_95": [
+      0.3837892274877763,
+      0.41059207129958375
+    ],
+    "paired_reductions": [
+      0.4007778491130225,
+      0.40642607415007126,
+      0.4064291001593511,
+      0.42023606373908406,
+      0.32710519192428966,
+      0.40762419055158194,
+      0.5685129874771919,
+      0.310213196858093,
+      0.41355995204758556,
+      0.32041790673188464,
+      0.382473882382114,
+      0.4164316199492143,
+      0.42341487351923734,
+      0.38510457259343855,
+      0.39935868451070977,
+      0.4681914863937914,
+      0.39654425028446916,
+      0.3316604754684327,
+      0.363968338719332,
+      0.39880913457301126
+    ],
+    "bootstrap": {
+      "method": "percentile bootstrap of paired reductions; each draw samples 20 reductions with replacement, then computes their median",
+      "iterations": 10000,
+      "seed": 19020261027,
+      "confidence": 0.95,
+      "quantiles": [
+        0.025,
+        0.975
+      ],
+      "quantile_method": "linear interpolation at (n-1)*p"
+    }
+  }
+}
diff --git a/evidence/knowledge-retrieval-20261027/results/performance.md b/evidence/knowledge-retrieval-20261027/results/performance.md
new file mode 100644
index 00000000..a16364b1
--- /dev/null
+++ b/evidence/knowledge-retrieval-20261027/results/performance.md
@@ -0,0 +1,94 @@
+# CPU offline retrieval performance
+
+The [frozen protocol](../plan/protocol.json)
+used **20 paired groups**, ten AB and ten BA, over four fixed development cues:
+`dev01`, `dev05`, `dev09` and `dev13`. Repeats do not add query types. Each side
+ran in an isolated source-selected subprocess using CPython **3.12.6** with
+site initializers disabled; import origins and complete source/corpus digests
+were checked. The corpus contained **108 records**, including **60 generated
+packaged skills**, within **458 evaluated corpus files / 2,883,003 bytes**.
+The shared library accounted for **342 files / 2,294,958 bytes**. There were
+**53 Python source files** in the source identity snapshot.
+
+Hardware was an Intel Core i7-8565U, four physical/eight logical cores,
+15.8 GiB RAM, Windows 10 Pro 10.0.19045. Clock/power state and unrelated
+background load were uncontrolled; the CPU inventory's 1.80 GHz is not a
+clock-lock observation. Fresh-reader calls include constructing a new reader;
+hot calls reuse one reader after the fixed warmup. Neither establishes a cold
+OS filesystem cache. CLI process/capture time and UTF-8 decode/JSON parse time
+are measured separately. Collect CLI lacks the API's repo/backend filters, so
+its wider-scope costs are diagnostics, not part of the fair scoped API gate.
+
+For each pair, the estimator takes the median of its four hot-query latencies
+on each revision and computes `1 - candidate_median / baseline_median`; the
+primary estimate is the median of those 20 reductions. The frozen 10,000-draw
+percentile bootstrap resamples complete paired groups with seed `19020261027`.
+The observed reduction is **40.0068267%**, with 95% interval
+**[38.3789227%, 41.0592071%]**. This satisfies the prospective issue bar of at
+least 20% paired-median reduction and a lower 95% bound above zero. It is an
+offline CPU retrieval result for this corpus, interpreter, machine and workload.
+
+The measurement loop retained **640 API calls and 160 CLI processes** over
+**836.162 seconds**; preflight/final inventory work is outside that elapsed
+counter. All **400 cross-version output pairs**, including **80 exact CLI
+stdout pairs**, preserved their digests. Complete measurement sets and final
+source/corpus identities passed. Independent read-only review reproduced all
+48 side-specific p50/p95 groups and the exact primary/bootstrap numbers from
+the raw observations.
+
+Regressions remain visible: API search `dev01` p50 rose **31.409 → 31.782 ms
+(+1.19%)**, and CLI search `dev05` p50 rose **340.560 → 349.067 ms (+2.50%)**.
+CLI relevance decode p95 rose **1.810 → 3.046 ms**, an absolute increase of
+**1.236 ms (+68.31%)**. These observations do not establish a cause. Individual
+rows are descriptive; no per-row significance gate was prespecified. Each p95
+below is a sample quantile from **20 repeats per revision**, not a service tail
+guarantee or evidence for additional query populations.
+
+Every measured mode/query appears below. Values are milliseconds; positive
+deltas mean the candidate was slower. The complete exact-nanosecond summaries
+are in [latency-summary.json](latency-summary.json),
+with all raw samples and output hashes in
+[benchmark.json](benchmark.json).
+The baseline [profile](baseline-profile.json)
+and [#129 trace](baseline-trace-129.json)
+are separate diagnostics excluded from the timing samples. No model, GPU,
+inference-service or real-agent-task comparison ran.
+
+### API call wall time
+
+| Mode / case | Baseline p50 | Candidate p50 | Delta p50 | Baseline p95 | Candidate p95 | Delta p95 |
+| --- | --- | --- | --- | --- | --- | --- |
+| `cold_context_material` / `dev09` | 1060.606 | 686.352 | -35.29% | 1570.936 | 885.685 | -43.62% |
+| `cold_default_diagnostic` / `dev05` | 999.688 | 614.151 | -38.57% | 1657.808 | 741.097 | -55.30% |
+| `cold_relevance` / `dev01` | 1141.533 | 788.233 | -30.95% | 1427.484 | 1146.733 | -19.67% |
+| `cold_relevance` / `dev05` | 1061.825 | 683.019 | -35.67% | 1247.372 | 817.370 | -34.47% |
+| `cold_relevance` / `dev09` | 1083.634 | 680.551 | -37.20% | 1901.819 | 872.078 | -54.15% |
+| `cold_relevance` / `dev13` | 1057.574 | 677.726 | -35.92% | 1766.674 | 844.748 | -52.18% |
+| `hot_context_material` / `dev09` | 944.589 | 577.180 | -38.90% | 1495.627 | 780.720 | -47.80% |
+| `hot_default_diagnostic` / `dev05` | 882.934 | 504.028 | -42.91% | 1529.218 | 640.953 | -58.09% |
+| `hot_relevance` / `dev01` | 1021.204 | 663.073 | -35.07% | 1317.269 | 910.342 | -30.89% |
+| `hot_relevance` / `dev05` | 938.521 | 579.671 | -38.24% | 1226.804 | 879.716 | -28.29% |
+| `hot_relevance` / `dev09` | 957.842 | 577.391 | -39.72% | 1702.202 | 725.926 | -57.35% |
+| `hot_relevance` / `dev13` | 949.685 | 566.260 | -40.37% | 1935.473 | 806.471 | -58.33% |
+| `search` / `dev01` | 31.409 | 31.782 | +1.19% | 54.012 | 53.470 | -1.00% |
+| `search` / `dev05` | 31.935 | 31.307 | -1.96% | 39.201 | 38.680 | -1.33% |
+| `search` / `dev09` | 31.590 | 31.506 | -0.27% | 54.718 | 46.758 | -14.55% |
+| `search` / `dev13` | 31.385 | 31.159 | -0.72% | 65.095 | 40.073 | -38.44% |
+
+### Full CLI process plus stdout capture
+
+| Mode / case | Baseline p50 | Candidate p50 | Delta p50 | Baseline p95 | Candidate p95 | Delta p95 |
+| --- | --- | --- | --- | --- | --- | --- |
+| `context_material` / `dev09` | 1413.646 | 1017.286 | -28.04% | 2644.763 | 1423.810 | -46.16% |
+| `default_diagnostic` / `dev05` | 1613.022 | 1103.381 | -31.60% | 2133.128 | 1382.574 | -35.19% |
+| `relevance` / `dev05` | 1413.156 | 998.828 | -29.32% | 1718.849 | 1369.453 | -20.33% |
+| `search` / `dev05` | 340.560 | 349.067 | +2.50% | 487.368 | 451.238 | -7.41% |
+
+### CLI UTF-8 decode and JSON parse
+
+| Mode / case | Baseline p50 | Candidate p50 | Delta p50 | Baseline p95 | Candidate p95 | Delta p95 |
+| --- | --- | --- | --- | --- | --- | --- |
+| `context_material` / `dev09` | 3.289 | 3.296 | +0.19% | 5.639 | 4.242 | -24.78% |
+| `default_diagnostic` / `dev05` | 0.022 | 0.022 | -0.45% | 0.030 | 0.032 | +6.81% |
+| `relevance` / `dev05` | 1.532 | 1.537 | +0.34% | 1.810 | 3.046 | +68.31% |
+| `search` / `dev05` | 0.012 | 0.013 | +2.42% | 0.016 | 0.014 | -11.16% |
diff --git a/evidence/knowledge-retrieval-20261027/results/quality.json b/evidence/knowledge-retrieval-20261027/results/quality.json
new file mode 100644
index 00000000..c06f2743
--- /dev/null
+++ b/evidence/knowledge-retrieval-20261027/results/quality.json
@@ -0,0 +1,29464 @@
+{
+  "started_at_utc": "2026-10-27T11:26:16.116393+00:00",
+  "dry_run": false,
+  "entry_python_optimization": 0,
+  "verdict": "PASS",
+  "freeze_content_sha256": "a465367319f3058398eb1f8579ee5d4174b8cb90457793e90788fb5796151057",
+  "identities": {
+    "A": {
+      "corpus": {
+        "bytes": 2883003,
+        "file_count": 458,
+        "files": {
+          "src/inferrsi/data/knowledge_packs/README.md": {
+            "bytes": 14172,
+            "sha256": "b877e67d9cccf0dcbe57419eeb3ff9374e5bd03ff8eec63adef1025c446140e9"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/diffusers/cuda/diffusion/group-offload-transfer-overlap.json": {
+            "bytes": 3277,
+            "sha256": "9bec3d912192efa2f2c853d3fb70bdae21a75875630a02294fcd084b18c30414"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/flashattention/cuda/attention/generation-specific-dispatch.json": {
+            "bytes": 3085,
+            "sha256": "90290d7bc80b1626863e0520c91bb196d93ebfe08190a450e9de59c6505ae8ad"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/flashinfer/cuda/llm/paged-decode-dispatch.json": {
+            "bytes": 3745,
+            "sha256": "0f2986494fcc9b751df71e598fd61caa68e28bc4b0ac401e4cc763bc3f75283b"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/cross-stream-dependency.json": {
+            "bytes": 3809,
+            "sha256": "b829315e55adf8e11f0bd581d5f45dd130d7eb3f81ce7de532f63c7b9c27cbb5"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/evaluation-completion.json": {
+            "bytes": 3841,
+            "sha256": "19d2e506ac4f50895776a83a062059aa7a052175c8e4b21dd0d46e1a5a6716a1"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/post-detachment-buffer-lifetime.json": {
+            "bytes": 4663,
+            "sha256": "18d78c0e542dacb42ca33e5264d92aef80fa06d163fd16d4feafd5ed0827ab99"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/pytorch/cuda/compiler/cudagraph-shapes-and-lifetimes.json": {
+            "bytes": 3623,
+            "sha256": "7fef4122ff794967e61f55d02a4a6b558d7e498d9513607f5ca86db6386928a8"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/transformers/common/llm/kv-cache-strategy-contracts.json": {
+            "bytes": 3072,
+            "sha256": "a2df26052112a9f3fc6242b2bd636ab13991442db57e56646168d6c52032d38d"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/triton/common/kernel/bandwidth-fusion-resource-limits.json": {
+            "bytes": 2941,
+            "sha256": "6e9673d1a600a6ad34820e4325f2a6417f44f0f96d8ed0bcf21466d998f19ee4"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/cache_dit/common/diffusion/residual-cache-calibration.json": {
+            "bytes": 3841,
+            "sha256": "bb9509ad0225a91b52bc6787477581d05f0ecd5e44fa59f08f718eea36a2ac45"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/bounded-windowed-replay.json": {
+            "bytes": 10959,
+            "sha256": "43613225e9ccdbd34c0f7e3e0767887b9bbfd0f38b2c21684677d99e31a432f0"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/speculative-verification-budget.json": {
+            "bytes": 10177,
+            "sha256": "a729e1fc00be44e1480a5dc65bbcdd58df8599681df74dbc651a5a49f27e9028"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/zero-overhead-deferred-scheduling.json": {
+            "bytes": 8191,
+            "sha256": "ed5426af145f066f6183bd316c4073c46f324f5ee83aa9bd616387c0d0e7149c"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/lmdeploy/cuda/llm/kv-quantization-capacity-cost.json": {
+            "bytes": 3366,
+            "sha256": "873997b38981ccc03ff73e6fb84e96022f293a019843efd8c4aa50f27878ac9a"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/codec-constants-and-state.json": {
+            "bytes": 4250,
+            "sha256": "1cecd7905fef5af3ef8952a4d803900a7df91d662ebae34db47cc0dbf352d4e7"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/fixed-customvoice-prompt.json": {
+            "bytes": 3660,
+            "sha256": "ea5e8fdf54cb731b255b73aa0cf47ef53fac103baa2bd5246be45f95d54cb593"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/graph-batching-state-ownership.json": {
+            "bytes": 3950,
+            "sha256": "130741a9f7ac9d366784d6c4dd42860262235caff0c277de1e8a8f8ec933b37f"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/sglang/cuda/llm/hierarchical-prefix-cache.json": {
+            "bytes": 3641,
+            "sha256": "b50ca58df28d4343545d07adf0da23f8f181297b30aba7870e5690630e7d7315"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/tensorrt_llm/cuda/llm/prefill-decode-disaggregation.json": {
+            "bytes": 2967,
+            "sha256": "349856dd98628f94987e014889cd1b15d0e3ae6ab6c576cdde917f28fbafb102"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/xdit/cuda/diffusion/parallelism-freshness-and-topology.json": {
+            "bytes": 3236,
+            "sha256": "3674b5916cb7b384951bf3e449b6796b161878daabcc1d7e7f6d6f8ff861cc17"
+          },
+          "src/inferrsi/data/repo_packs/afd_plugin/pack.json": {
+            "bytes": 502,
+            "sha256": "00042abeabdc9c0e24f1a4edaba47e233df398aef0ad4e181fc70191d383da5f"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-aclgraph-capture-replay.json": {
+            "bytes": 4311,
+            "sha256": "327414443ea9537ae320aa405912c338351bc06cefdc0b99352e1eea36f61dc3"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-device-health.json": {
+            "bytes": 4115,
+            "sha256": "633fa5fdcea1e1e9edda8a218d6a3fde0d12637848348e1e2fe7d1b0bd744a16"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-distributed-parallel.json": {
+            "bytes": 5841,
+            "sha256": "709ca0d061e4d6917fe8fa049fe93165e71dfecf96c7bf264ced25346381f603"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-graph-mode.json": {
+            "bytes": 4135,
+            "sha256": "bb82ab3d628b91be957e557545810ef5c20790be7d43296096f20a21f45ec391"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-model-adaptation.json": {
+            "bytes": 4137,
+            "sha256": "7ea1ab0f0044afff935a91071b36ac09bc46b5ef044fe2a68ab0608256ddd853"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-profiling.json": {
+            "bytes": 4186,
+            "sha256": "0d27058fd8d0ef8b524b7a60d1cb76950622746c94b5a03915e64e6edb4ab3d4"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-tuning.json": {
+            "bytes": 4439,
+            "sha256": "fcb49d71f59d8b39c1eaa03d24e95e2cbfe5f9d7f6e0f22df03ee0d60f6a568a"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-optimization-casebook.json": {
+            "bytes": 4371,
+            "sha256": "05783630bfaf3c7ef99b848865f4bbab50d0b52e80ad04fcd028c624d7638e9b"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-performance-analysis.json": {
+            "bytes": 7200,
+            "sha256": "b6eea897c12d0c6ac9d172089535c82c25eb71c09d7a4972e64812cc4682b5c9"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascendc-tiling-design.json": {
+            "bytes": 4505,
+            "sha256": "e69b6e89dd342ba55ae54b5fac14aa1b91ee2a361da40cb669c287ce519f4517"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/cann-attribution-overlap.json": {
+            "bytes": 4915,
+            "sha256": "8ac01b60f5cd37fd594d6cb5119e4e6525199d017296fc8779dc68a576f5525c"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/torchnpu-device-memory-lifecycle.json": {
+            "bytes": 4349,
+            "sha256": "7a79d0eb2c50381d40086e6db72d4bdd2cec5d2261bfb00cf73a60701a94da2a"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/torchnpu-pinned-host-memory-lifecycle.json": {
+            "bytes": 4431,
+            "sha256": "8bb34dcfbebb48667a94434b036a1a408b970af9890b047a3de325f364ed5f7c"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/assess-serving-preemption.json": {
+            "bytes": 4198,
+            "sha256": "e3d2a1a72d0d980f31fc85deee27db26bc0daa0c709e8ae99e42deb814a4a7f4"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/assess-split-backward-schedule.json": {
+            "bytes": 3915,
+            "sha256": "a8561794786eb0a897d831f00c92b4fdd3b0492316c1b1d8ad13c1f8f75ac815"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-h3-runtime-residency.json": {
+            "bytes": 12169,
+            "sha256": "602dd625874bd4a89861d5bb25881cc63e4b5a7fdedc2867089f0952943f6eff"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-llama-runtime-state.json": {
+            "bytes": 5816,
+            "sha256": "64f6ab8a95172378350323cead08bc6c201c624b711e6242938a6a21a8e83ad7"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-sd-cpp-residency.json": {
+            "bytes": 8349,
+            "sha256": "95ed97b717255be3b817e5c69debd6d41197559dbf21771507d5da79134a72e1"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-sglang-diffusion-lifetimes.json": {
+            "bytes": 7276,
+            "sha256": "5e5889223878c6cf42a49c6f09c60e9d9add88cb7e7d0df050c4e75d88a05aa6"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-vllm-kv-lifetimes.json": {
+            "bytes": 4275,
+            "sha256": "00424005a5eeb8b15f239edd2cb2c71047787e3ecf55924860cb7a78d137b586"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/bound-windowed-attention-replay.json": {
+            "bytes": 7877,
+            "sha256": "98b2291147a796ef59cf20a075ff662998ee6a529f85b49e9c6fb62fe41fad9f"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/budget-speculative-verification.json": {
+            "bytes": 7490,
+            "sha256": "66448f59460c9dffac46e74160388fab650cf8bbbd5bdc95c1465b70daf0c2dc"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/calibrate-serving-simulation.json": {
+            "bytes": 3992,
+            "sha256": "055b70e0b86f65156775b519e4b8cbf06a956a8993b06a66175167cedb78f2cb"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/cgroup-resource-controls.json": {
+            "bytes": 4801,
+            "sha256": "ad406b6dd5d9415bd28d448e64269a388b50e11f9eb7f5a9adb6b0a6246fd303"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/checkpoint-recovery-consistency.json": {
+            "bytes": 4740,
+            "sha256": "ccb85314a831bde58f8cc515580d001b174c11dca0bbcbc6221303acf629046d"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/choose-parallel-training-plan.json": {
+            "bytes": 3901,
+            "sha256": "40afcb41c59988fc1af8fdd00d2340800c54e96847dd0de16ecddf7ac221d1b4"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/cite-check-downstream-pr.json": {
+            "bytes": 2533,
+            "sha256": "dabff14684d5ae9a2b05a9743ff5b76d00a90fdb29319588aba67f5087aad706"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/claim-contract.json": {
+            "bytes": 2469,
+            "sha256": "78f15f6533434fc29adf1df651f7a15eb8ba81733f8b02c710d90102541d707b"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/comfy-ui.json": {
+            "bytes": 6919,
+            "sha256": "1496807d958df055f720ca93e6dcf7f7e8f506f3f12c94c9a2166f789b00058d"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/competitor-scan.json": {
+            "bytes": 2449,
+            "sha256": "c2e97e49e5185710cdfaa08ac1c2b1c19b958e68ebfca885d5aee0ed8c9f8c8b"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dcp-async-save-performance.json": {
+            "bytes": 4538,
+            "sha256": "359a5acf8304a78fc074688676f9a0be62f56f463d0960a544e409f49843bdaf"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dense-performance-evidence.json": {
+            "bytes": 4227,
+            "sha256": "d87f93a9dc9ddc583e3929baaac85f7a425666dd6dd71665933f18344ccec834"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/diagnose-vllm-admission-pressure.json": {
+            "bytes": 9789,
+            "sha256": "a1de40def775b2e840c27330b67e6f37c3c244db1731fab2bd3e8a787143210e"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/diffusers-invocation-state.json": {
+            "bytes": 6109,
+            "sha256": "914d2ffe9776ba931bed780a2a270c1525e3c6cbaf0735b8531064d3180f3233"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/draft-acceptance-issue.json": {
+            "bytes": 2549,
+            "sha256": "6aaa63dbfe6cda982ccd3a4e368527b80fd2873672c000c66d06d28e40a2bf14"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/evaluate-elastic-serving-placement.json": {
+            "bytes": 4334,
+            "sha256": "fab1fa814e789ec2d86587fcf7de7c38174464b9e9ba0cf53ef3fb470053e7c4"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/generation-cache-lifecycle.json": {
+            "bytes": 5972,
+            "sha256": "e5679cdecb722bbd17280913913da8473052ad09e9e7a13b3a9b95463a139ef6"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/host-memory-management.json": {
+            "bytes": 6530,
+            "sha256": "a36aea21169281866b2468860d3ffeb912e61edd571308b02f82939f17f694a8"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/host-overhead-optimization.json": {
+            "bytes": 4243,
+            "sha256": "05fd7f054966b96892980d585352bc1c3fcd47f439cb34268fdf07feee58a998"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/inference-benchmark-design.json": {
+            "bytes": 10735,
+            "sha256": "091bb086c6f0123dbb5db78968c13f846cad135e8969982efe557644b7c7613b"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/inference-optimization-casebook.json": {
+            "bytes": 6531,
+            "sha256": "c512025d7a59934c91665e24752d9393ca9b6a9dfb624c72789294defae93c0e"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/kernel-writing-guide.json": {
+            "bytes": 4545,
+            "sha256": "f1f0ca1d11d3550c9b8c5fdbed8f6cd09b877ac14341f973d58dca02b5d8995d"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/kv-tier-management.json": {
+            "bytes": 4609,
+            "sha256": "6141f61fb6b0d147b704ab7a2627e9babdb53a90e15b659152a3d565967afddf"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/lazy-evaluation-stream-lifetimes.json": {
+            "bytes": 5934,
+            "sha256": "34f47009a8a33f8907bd78f110b95fcd62409bef1b38e1055acb46b2eab285ff"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/npu-skill-factory.json": {
+            "bytes": 11868,
+            "sha256": "24c1d3bff288e86b99ea674e3252ec6f54d9ba944274aef6a7666197afd6788a"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/optimization-search-policy.json": {
+            "bytes": 7101,
+            "sha256": "d7bf71d5491765f62eec02fe5013dc1a5ef56e1516892da48593c6165a86366d"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/os-process-placement.json": {
+            "bytes": 6304,
+            "sha256": "e6e8fe5d7d66f3e98f05762d94441726dfa757ce7fbffd2b35402b3f327f8fb3"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/performance-ceiling-analysis.json": {
+            "bytes": 13695,
+            "sha256": "47c034d96026e1c2a02848d9c9c3caf4024a5033664bfd089ad30d9f23a80deb"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/plan-activation-residency.json": {
+            "bytes": 3871,
+            "sha256": "436e364020cc854fd52a10ee76e4593c7519bb0546c8aa5238fcfd89158e8ab0"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/repository-evolution-analysis.json": {
+            "bytes": 5865,
+            "sha256": "e0af402c3320ebd9bd36c774cb52c0708721a86ca446c05f6e94306888838e30"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/rsi-evidence.json": {
+            "bytes": 2475,
+            "sha256": "c8b4bb5a8c88009ac5b1526e82639715381c940d3c3eb295f0333cae6c2d3ca9"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/skill-evolution-playbook.json": {
+            "bytes": 4091,
+            "sha256": "0e063c2e7686de0c27c061e1aac0eb2d8c4fa83246e32ddeeb0e28126c176425"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/storage-io-measurement.json": {
+            "bytes": 4783,
+            "sha256": "80e589e3a0b67d3a045e03feeadc45e720ceb277e6124c917eec04863eb67419"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/training-input-pipeline.json": {
+            "bytes": 4296,
+            "sha256": "ae696bea8d2b7bfdaa29b6e5bc0bd861abcc0e42e7d9bec6ae20c6e7d6f27669"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/training-resource-budget.json": {
+            "bytes": 6944,
+            "sha256": "ea5e5e864e3aa34bd00be67c8e741877ae01480f039652073b65b71dfcd54c97"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/weight-loading-optimization.json": {
+            "bytes": 8139,
+            "sha256": "7f2d1fbc6b1cfd80c3d6ed8df107d4ca9b97655626ee078c345445a303b1b367"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/audit-afd-p2p-exchanges.json": {
+            "bytes": 6640,
+            "sha256": "9ac45aa784b430ad1ca9ffddecce8dbae9740cfe9c98c10b330e01d350015a38"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/diagnose-attention-bottlenecks.json": {
+            "bytes": 7602,
+            "sha256": "e525d75578de1cc164e6553d01efbd0a014e74646200580f59b1161c030413ad"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/host-overhead-profiling.json": {
+            "bytes": 3965,
+            "sha256": "e9e6ae8c5b9986e83cade3bd029baae112859e01aa6c96f741a264f3bee44adc"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/reason-hopper-kernels.json": {
+            "bytes": 15086,
+            "sha256": "758a649e651d8f9635ac2df45d18043e156c8d54b8f8dbc95f8e68ebcb866a28"
+          },
+          "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/parallelism-and-extension.json": {
+            "bytes": 3399,
+            "sha256": "1253f5f01bcf1498703c7f34960f0c50580df5ba8f140af557a903e296ad81d3"
+          },
+          "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/runtime-radix-hicache.json": {
+            "bytes": 3317,
+            "sha256": "5a0b2aed51649af214cdc1034f57650b7ce195c934a17452ee67a35131f80cdd"
+          },
+          "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/tuning-and-validation.json": {
+            "bytes": 3385,
+            "sha256": "ef4d4f7182da4433d59d72dda0d21b4aeafadd73260be11ab8d4cb36c20fafc6"
+          },
+          "src/inferrsi/data/repo_packs/sglang/pack.json": {
+            "bytes": 620,
+            "sha256": "bce2e0c42796a9ef499d459c37b56d1d56f4b19344841359c534fca29c0359ce"
+          },
+          "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/audio/serving-and-streaming-quality.json": {
+            "bytes": 3440,
+            "sha256": "97fc1d5a11a39a0634dae9e9602a4ec47a4d2348e121076c2d5c836cf0692114"
+          },
+          "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/omni/extension-lifecycle-validation.json": {
+            "bytes": 3482,
+            "sha256": "7da8dc5d2655099de2bcd5f0fa922d9e1bcb0bfd2f571f026a004a7f402a75b6"
+          },
+          "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/omni/pipeline-state-and-transport.json": {
+            "bytes": 3404,
+            "sha256": "0c57692772019504264e39bf0444a69e20cff86f6e112b072c53cf7a4299cc82"
+          },
+          "src/inferrsi/data/repo_packs/sglang_omni/pack.json": {
+            "bytes": 630,
+            "sha256": "8e70336bc2aac0c74732a026337813652b5d5c6e1fc60f5571e9442345077b96"
+          },
+          "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/architecture-search-space.json": {
+            "bytes": 3294,
+            "sha256": "2c5ad72d5bab4f7082fc3be1532c347444cde38f5a84639fe16b22134b68685e"
+          },
+          "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/controlled-execution-tuning.json": {
+            "bytes": 3255,
+            "sha256": "6a2bf611a432213919e1c237687507ceb0ab753cccad0e934dc77f86604805e9"
+          },
+          "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/engine-validation-contract.json": {
+            "bytes": 3295,
+            "sha256": "ef89775bc82965fe87a832c7ec8a5cf55cb9eba4ecfc6279caea9ce3b0410c41"
+          },
+          "src/inferrsi/data/repo_packs/strata/pack.json": {
+            "bytes": 617,
+            "sha256": "643b32defb8db9aa230f624856f4fb4b595b5c23db58d9453091f5420cff71c3"
+          },
+          "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/bounded-serving-measurement.json": {
+            "bytes": 3239,
+            "sha256": "fb00ddac14ef6b06f48da9ba335529bd55d5d83c2a2fe331f0aa1ca16afceb37"
+          },
+          "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/extension-correctness-gates.json": {
+            "bytes": 3343,
+            "sha256": "4ea7083d277df205ea4c361bee87a8207951596e9ef6cfa136cc7d2aeab02f46"
+          },
+          "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/runtime-and-kv-ownership.json": {
+            "bytes": 3280,
+            "sha256": "e6142c1065754716a5f26f42119722908dc8fff63e667c986d57f1227b8610bd"
+          },
+          "src/inferrsi/data/repo_packs/vllm/pack.json": {
+            "bytes": 508,
+            "sha256": "1705d96710249c4514c15d8d011b43c1a1ae7c15fbc9ddf11ef2d0e6b2ab7d07"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/minimax-h3-reference-kv-reuse.json": {
+            "bytes": 4047,
+            "sha256": "80b853fc594d387ab48bed04dca245e0a22b8cdaf7c9b853551ab3bafa302e84"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/minimax-h3-sp-head-buckets.json": {
+            "bytes": 3742,
+            "sha256": "c4d29d1bceb5e2cf86151bf53c9dafb3d83428fb1e9d2a848d852bb8824bf857"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/qwen-image-21-eager-t2i.json": {
+            "bytes": 5268,
+            "sha256": "ef107d43cc448219eda4d64ab961b61cda6350e20d30d94de1ff7c4ffff7e775"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/audio/streaming-serving-measurement.json": {
+            "bytes": 3389,
+            "sha256": "2680f32f28907f6abc7a7e1e3ecab4f46ee412f40f4898ed2f371ffadf47ae4d"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/measurement-and-profiling.json": {
+            "bytes": 4472,
+            "sha256": "c8b25583cd160024c7472a20debea34cb6d4bc79714b4bad2953555adc9f794e"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/production-integration-gates.json": {
+            "bytes": 6112,
+            "sha256": "649b49ff64d3e9e88bd620027dabf9d126ca69bbe91b448f28e845ecbe1f8e2f"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/step-state-and-output-validation.json": {
+            "bytes": 5090,
+            "sha256": "e1ffe28c4fa54af739cdef2fedf4755bba18e078f8380d4fde8138ea712e5778"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/upstream-review-gates.json": {
+            "bytes": 7614,
+            "sha256": "5127709bb85ce31008d19468e1b4fcb5454a71776f69f991841cd9ef7558f9c4"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/omni/stage-ownership-and-search.json": {
+            "bytes": 3427,
+            "sha256": "e75d6c2829ae22881893f5da2ad7edba099fd64ccfb0fc93d30473e06e9d4d80"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/audio/concurrency-scaling-state-and-overlap.json": {
+            "bytes": 25324,
+            "sha256": "cb47fabb232db81ed525cbb54371becc7e4a28b2bae9d8492f60dfb6f31563e8"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-adaln-schedule-cache.json": {
+            "bytes": 4253,
+            "sha256": "ddbba3bb72b5789938595a82d3280e861f75f6ed54f05a1d2b7dcb1089d30f63"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-cuda-mxfp8.json": {
+            "bytes": 4468,
+            "sha256": "e4c538f03f4ab90a4acc8e3d8721bbd226aec1ee9bde45757a7c27d3d970821d"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-fp32-modulation.json": {
+            "bytes": 3625,
+            "sha256": "abb8d94a1b025f0e2c8e085275e96ee85d68aeb27cb12d0918e4647f8255b693"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-qk-norm-rope.json": {
+            "bytes": 3706,
+            "sha256": "4c1c4acfb9cc0baa90bc28fed1903020ea1ea38755b84315711795db45e48bb6"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-ref2va-teacache.json": {
+            "bytes": 3835,
+            "sha256": "8aeb3f99ac4227cf8b3a5cece13d938b19082e265a557eddae27b7e495bcc22e"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-vae-stacked-tiling.json": {
+            "bytes": 3508,
+            "sha256": "8cc18eacbbfd8fc4d3a6b4152c4b584d559fc9b15377da56ef072e542d807c93"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/pack.json": {
+            "bytes": 566,
+            "sha256": "58c971e9dcf6ac15752629ed4d9426aeb5b5d857fc37b5bcf921aa2e81c0cdd5"
+          },
+          "src/inferrsi/data/repo_packs/vllm_rlt/pack.json": {
+            "bytes": 649,
+            "sha256": "6097041b8d026740060e527f49b1b4e288315b1c3227f7583805b86d8fa415f7"
+          },
+          "src/inferrsi/data/skill_library/README.md": {
+            "bytes": 18819,
+            "sha256": "ae583f3578425df1dc91ba32b8b6543c0469c9a817ebea873a77051d6b13e1ce"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/SKILL.md": {
+            "bytes": 5494,
+            "sha256": "7581b56336d07213a87d01477793ae1a8acf0b7cf876fce36c9a761ee32c1ff4"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/metadata.json": {
+            "bytes": 1429,
+            "sha256": "68ac7f64b7ab947fc2db664bc3990f102c1cc5539b42350aa5818d0bbf17d86c"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/ownership.md": {
+            "bytes": 2081,
+            "sha256": "40688cc44183027efbf2fca8d3548c299cdf2e33fef54f76ab3549353eb9bd1b"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/source-map.md": {
+            "bytes": 4635,
+            "sha256": "10562a75683bd9bd07ec2e37a1cbd0f93f2534d6f85b55a566adb2d97485215f"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/validation.md": {
+            "bytes": 1477,
+            "sha256": "5a0b854b87d091d7d505b0bdd5c03330052507a851ab6edf35b93a6675f2d4f7"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/SKILL.md": {
+            "bytes": 4010,
+            "sha256": "a95569bcc0da6e715fd74783a6a1e0489e92b680c582f15dafbf98df589d505a"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/metadata.json": {
+            "bytes": 1519,
+            "sha256": "9c2293c8956508de987c886c7224d673d22912e7e915b026b05abdd287e97401"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/execution-lifetime.md": {
+            "bytes": 2147,
+            "sha256": "c4bc41e08aca68f661325f61e6f0db1760a079dc362f295c847bdc374fee3bf6"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/source-map.md": {
+            "bytes": 5918,
+            "sha256": "798b9cb89ed6b885eb47464ca9f137d2802a54571b9327e631d7742d3e5abd5d"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/validation.md": {
+            "bytes": 1608,
+            "sha256": "ebfc00bf7d6ab0a1dc415953ae84d56ccdc713d5614aff85e433c4a843793bff"
+          },
+          "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md": {
+            "bytes": 7322,
+            "sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845"
+          },
+          "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json": {
+            "bytes": 587,
+            "sha256": "db7ad368394ccbfdeff6a16e3423c300d51305b1f95ff5c3936df91738a8cc35"
+          },
+          "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md": {
+            "bytes": 5314,
+            "sha256": "22afe4f4d77f441096cc657d55d763473bfcbea8fb531c8e493fe49ab84955f7"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/SKILL.md": {
+            "bytes": 6745,
+            "sha256": "d494fc7e539cbb7388189bd7b3f04bb5c4983fdd7a82484348aad1cee7f741f2"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/metadata.json": {
+            "bytes": 558,
+            "sha256": "71c68431663ed2336cccc985a354adab6b13b4cfd6123a18a5d8d53c28268509"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/references/source-notes.md": {
+            "bytes": 13004,
+            "sha256": "6ffbfb68cc7e04923b4fbca2a73a7555b7c3b45a6f40819bb5f3213ef05cf88d"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/SKILL.md": {
+            "bytes": 5977,
+            "sha256": "03829039860ece35519f2e4c339a48b5014ec25e07a2c150a3efb6a3a4e35b4a"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/metadata.json": {
+            "bytes": 559,
+            "sha256": "80d80c15514919826fa72d6f6cf26b76db8accd85d21187fc239232ca454e7bf"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/SKILL.md": {
+            "bytes": 14703,
+            "sha256": "4ce9dcadd1ae3c7a8cc633336760d0a3c382f354f1b7028e3a10d28fa0d0c9fa"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/metadata.json": {
+            "bytes": 1403,
+            "sha256": "b6cb25879b1bca5f5af08b48776f2e664cf5dcd11a0efb1c4f56879e2afa9006"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/references/awesome-ascend-communication-tests.md": {
+            "bytes": 9388,
+            "sha256": "0cb27a3d24409171bb414d7ebf0c49ad62be5f46482f29624695496ba9eb3e63"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/SKILL.md": {
+            "bytes": 5839,
+            "sha256": "9203cc5a14a1d65df5fbf275734b93be023e06527459b2f79864f28a338cb44a"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/metadata.json": {
+            "bytes": 536,
+            "sha256": "72b41e7845915d34b09219624e82976386f3125816293540c2c38143fe6b477f"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/SKILL.md": {
+            "bytes": 6440,
+            "sha256": "8fdf46fb2743b67a66bc069d22a45305a222debfb606777bc451b4b8f31d9cc6"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/metadata.json": {
+            "bytes": 536,
+            "sha256": "be681a3206e5e183b4c9670c9d7784bcffce18c891449e5b15f94ca5f1324b9a"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/SKILL.md": {
+            "bytes": 12988,
+            "sha256": "401981cf7828edcab0c9ea882f7ecd83badfa3c2a95006e912cf99d67b3d6a12"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/metadata.json": {
+            "bytes": 760,
+            "sha256": "efd37bb23acb500dad4ec8ebe7ddfe5c3f196374052825f923ac85da8445f06b"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/SKILL.md": {
+            "bytes": 12236,
+            "sha256": "38b151bc6b4e9734c6fa41efb7a30343559400c0c75bbdd26bd8a968b8ea4592"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/metadata.json": {
+            "bytes": 557,
+            "sha256": "04d2002fc8a0a7becf86b0b73daaf2da8716ab466d7f988da43d41a10f26680e"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/SKILL.md": {
+            "bytes": 9172,
+            "sha256": "647f5e9a288f2c4e1617ba53df59d9d827df749eb4fe86182d21a9c41175caf8"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/metadata.json": {
+            "bytes": 536,
+            "sha256": "f425f7c1abbd616387273e8932c6bd02f4b1da012f099286ad4171e0f6037b22"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/SKILL.md": {
+            "bytes": 9880,
+            "sha256": "a057b7f86a24f34a8f84abbbb4910d58d4e50ff47060e1c404600017e3febe89"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/metadata.json": {
+            "bytes": 2297,
+            "sha256": "9e705625c8308013c287576b0247da6f6116d817a4a535cbecff80abd6ea35ee"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-CANN-2.0-LICENSE.md": {
+            "bytes": 5262,
+            "sha256": "89b1c114df62cfd183004f38d18d07f186436c3b37203dd69c3e84becf9d2282"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-stage-evidence.md": {
+            "bytes": 11789,
+            "sha256": "f56e4125325c9c1b9235a1d64fd732861c7686444492422daff44ba8dd648be4"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md": {
+            "bytes": 11860,
+            "sha256": "93aa1b775ced304cebfe6c4dec0503ddb2260ecb3cb01edc17ce329446c5ce2b"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/metadata.json": {
+            "bytes": 552,
+            "sha256": "f6d94f104cb70969024b2f4542078a240e200712c66279eca8b94f60bee1dd71"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md": {
+            "bytes": 8972,
+            "sha256": "0e6f094dc554abea12e07def50f20b233b47127779a8aefad595b4ab1efd24c2"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/metadata.json": {
+            "bytes": 569,
+            "sha256": "acccc8eb86de0df6caf78a9b3c6db76a2a686adc4a5ba0029a4d62d50ff9a8f8"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md": {
+            "bytes": 5989,
+            "sha256": "8428df7157af024fa17cfe0b3818ed54bd9f468306dda46eb8d4f0edf4f3f230"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/metadata.json": {
+            "bytes": 579,
+            "sha256": "c650dce88c07b636d68e063dbda4d319f6f7f3ecbcdd5e47f393cbfc3af7c71d"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/references/source-notes.md": {
+            "bytes": 11470,
+            "sha256": "80cba4d3a884e7ee2618326cf74b9f9c3a34e4d9cf443707f6bbe673f2d5ba65"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/SKILL.md": {
+            "bytes": 6159,
+            "sha256": "23e93ea7c98f99743509fdc04139d78ed2e7123c63e80b366c6beaba7eb70d0d"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/metadata.json": {
+            "bytes": 581,
+            "sha256": "97836a38465106d816a9ace894a7b277d6d330f9accade8f8493ce6f0fb57ef9"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/references/source-notes.md": {
+            "bytes": 13493,
+            "sha256": "435193590b4a2ba32e2e4c6fa9d7e8f889528d412cbbb48e35ec5433f7dab384"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/SKILL.md": {
+            "bytes": 7137,
+            "sha256": "a25738dbb3b9daa15c5827cb536a20fe13f7c8b9dd14aa3fb7cfa7e132879431"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/metadata.json": {
+            "bytes": 1007,
+            "sha256": "3abe626a6e4c73c943429f609ce2c62aa77aae7e1726475ebbbdfbe1794f6f7d"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/mechanisms.md": {
+            "bytes": 17376,
+            "sha256": "fe08b2292fa587be3874774ddf41063b84c579a0085df422ec6710d8aa90b8b7"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 4024,
+            "sha256": "371b01ee201ef51fecbfd6da892d517aaba295930b9395025f03f34e21c030a1"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/NEXT_GATE_LEADS.md": {
+            "bytes": 1623,
+            "sha256": "35b8b58ae5db02b2df51ccbe5cdf870a7383d17f345f9ee5eb7b0a4146dfc74b"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/README.md": {
+            "bytes": 2888,
+            "sha256": "e0b1ec1b39c10067f5ddce58320851ad089568c4b2296a24777389097cb5b388"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/io-reduction-can-plateau.md": {
+            "bytes": 2671,
+            "sha256": "5cc35f514224ac6e6efd2da6db221bba468b2f4db66807295e67b240bcdfdba5"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/overlap-and-fp8-counter-signals.md": {
+            "bytes": 3692,
+            "sha256": "80c8a60a6df55f3d84fab23e03567d8b38e23bcdbd63cc09881e46db1e2888ab"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/partitioning-after-fusion.md": {
+            "bytes": 3132,
+            "sha256": "4d4cabe0991eab554dcfbd4fa9c8199e61fa4767709b0c4cd646e2588bb398a2"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/knowledge/attention-cost-model-boundaries.md": {
+            "bytes": 5223,
+            "sha256": "5ed8f888443e38ccb72545b6025447aab7b92936f78c151fed680e582126239b"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/sources.md": {
+            "bytes": 9069,
+            "sha256": "53436e0de4e6dda2d29426591bf61f8526378cdc6f9ec7967b88b1bcd4c61e6f"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/host-overhead-profiling/SKILL.md": {
+            "bytes": 11199,
+            "sha256": "326e566fbe30d3ba04e4b50c3e87aad9df037f0ac4ad6fe129165ae97dc5f91d"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/host-overhead-profiling/metadata.json": {
+            "bytes": 527,
+            "sha256": "76e34c1c64aba473a9e126b032a776417caa0f913f6ac3cf4dba3a5d123b1ee0"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/SKILL.md": {
+            "bytes": 4905,
+            "sha256": "cec4e3aa1eb64c31c54b5f5ca40af2b61b7919d94be4174968581d5f752c0538"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/metadata.json": {
+            "bytes": 7453,
+            "sha256": "7520da1cc97412631a600957ecb18aa5a7a4461ea21e76c237330180193ed898"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/fp8.md": {
+            "bytes": 6287,
+            "sha256": "31d9c211c11c2224007abbdcd4059913c4a54833fe2d2f926411c5238fcd3501"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/cutlass-bsd.md": {
+            "bytes": 1547,
+            "sha256": "42fec630f410aa308f70a51a89fadcd19586fa620f9831a32bee528a9a10000e"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/deepgemm-mit.md": {
+            "bytes": 1065,
+            "sha256": "6c73a30c14a2938abaa6afe98ebb19e6d81dac5bd4ace0cb05c61c20afa37b85"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/kernelwiki-kda-mit.md": {
+            "bytes": 1063,
+            "sha256": "84aab57158a914f448b38d8be29dd842b634a08c192d844f416a256beed4ae94"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/notices.md": {
+            "bytes": 3240,
+            "sha256": "876ec016644bf5331b3f257985e126c0e61980e3bfd82569b7b0c3f6b2c6c64c"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/resources-and-scheduling.md": {
+            "bytes": 8919,
+            "sha256": "1ce25b426736072435a1ead9a2a8e1f56df5d0abb8198e8c4fdcb34321db4964"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/source-inventory.json": {
+            "bytes": 120506,
+            "sha256": "1dc09c895c970a0c9708158bcc6dd65cdb736274fd2c35d44542c70407d3849d"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/sources.md": {
+            "bytes": 9004,
+            "sha256": "6b74b9e774f2a80db0860510d9e6940d36284430f29c1e124f115a9eaf25833d"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/synchronization.md": {
+            "bytes": 6843,
+            "sha256": "4322f6b3c0e4368bf8c192b4ec752529767f50d19c0c38dad8fdf9cd781e41d9"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/SKILL.md": {
+            "bytes": 7348,
+            "sha256": "fd1f5a7e6a55714730044155f5b38c1d700383d4e16c3e91963bbfbe1c773af5"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/metadata.json": {
+            "bytes": 572,
+            "sha256": "19e40763fedc20990651bb989c32a7693306c5c54af46c3d4be77fd4f87b6c03"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/SKILL.md": {
+            "bytes": 23213,
+            "sha256": "2c807f3df2da2f533d25183de5771f3d412058d16410994292ef8f48c49f9b3a"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/metadata.json": {
+            "bytes": 1684,
+            "sha256": "9ac6af135f0b6d76a76bc201b6bbd8ac9afe5201dd1e2ea82c0ac4826de401da"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-LICENSE.md": {
+            "bytes": 11357,
+            "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-native-offload-readset.json": {
+            "bytes": 6267,
+            "sha256": "cc6c2011be746002ba3ff96ccefab1809863748ddd17eb102b75e7b0e134fe55"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-native-offload.md": {
+            "bytes": 8966,
+            "sha256": "6c2a72b5ba3708a01d229e0576570837a617e6a42dbb0b5312a26a65a6c31542"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/SKILL.md": {
+            "bytes": 16072,
+            "sha256": "36ef88826a32ac3147b869001004ded17a8d5a491e25bfe105005ba39ea9d1eb"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/metadata.json": {
+            "bytes": 539,
+            "sha256": "ae71c389dfdb1a19d267feecb4c6f1fbb4c9afad501ff6d8fab7e38b29bcc37a"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/references/deepinfer-zero-overhead-scheduling.md": {
+            "bytes": 10894,
+            "sha256": "f452bf40132c9195d739f1c86b3aa18857138c7ed0f1861af5853d93b9955924"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/os-process-placement/SKILL.md": {
+            "bytes": 10327,
+            "sha256": "2003492e4d89fde0443b180d74933e764bd1979435e109c8d26e20eb8f6e5053"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/os-process-placement/metadata.json": {
+            "bytes": 1399,
+            "sha256": "ec02995c075f0a540921f57b4846f2cafe334eb90f44389f17c3c50457c3e9c5"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/os-process-placement/references/llama-numa-thread-readset.md": {
+            "bytes": 6617,
+            "sha256": "748b254b3b043a5ddb0f480eb4a566854d27eedd7b111499863e9bf0a5a6951e"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md": {
+            "bytes": 8813,
+            "sha256": "63b883c10a8d3e7d918abfc7175b72e2b401ff1e02697f94b8382ec1c8f369b7"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/metadata.json": {
+            "bytes": 558,
+            "sha256": "0a41c7b00978332b54ee1e4a7c1ade88ecde3e961128f24d1b4f7708caff5ff5"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md": {
+            "bytes": 6402,
+            "sha256": "0a5f48d9352c5bf460775ef47f9136829978811f4a16e84983e299782cae305d"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-serving-preemption/metadata.json": {
+            "bytes": 537,
+            "sha256": "77ce5ba16245cbd80d46d0abdf80aaf26fc0924fa2d2a5624ff430de68a1daf0"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-serving-preemption/references/source-notes.md": {
+            "bytes": 30405,
+            "sha256": "f29fcecfba0490008a354c11b42403983fe165e56ff907ad61cc6040d2b6407f"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/SKILL.md": {
+            "bytes": 3181,
+            "sha256": "1e64f28ea71672fb53c56f6108b87864738c6d03d6dcd32eedaef9e0625e96e3"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/metadata.json": {
+            "bytes": 478,
+            "sha256": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/references/source-notes.md": {
+            "bytes": 10548,
+            "sha256": "61b1609f71c37b61d136b8c4d39f793de87cffb1cf31c8ec737a1c3fa6fb1b4e"
+          },
+          "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/SKILL.md": {
+            "bytes": 3884,
+            "sha256": "ac773c6f1c110a664dfca28a3ae9129b10c07614e60ca3640dea776220e7b352"
+          },
+          "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/metadata.json": {
+            "bytes": 648,
+            "sha256": "261ce9f644ea6eebb90c830f41fb8b320d2b401456e5bd8d494095a44b8620d0"
+          },
+          "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/references/source-notes.md": {
+            "bytes": 8415,
+            "sha256": "9696d3c0ef96b6611787446bd6f96d4d6255ca8d107105de8c1d9af83edb9fa5"
+          },
+          "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/SKILL.md": {
+            "bytes": 6416,
+            "sha256": "6b7db1eb2780dec27858cb7aa0a3c093c5b5dfa4c30261aebfeaa255213ff688"
+          },
+          "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/metadata.json": {
+            "bytes": 582,
+            "sha256": "4dc8659806539b4784ea9348a01d4194ac70d0bb297a916f90ff6419efa5c328"
+          },
+          "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/references/dlrover-persistence-replicas.md": {
+            "bytes": 8398,
+            "sha256": "055f4357f79026fcbcc766bbe77db610376ec353623f8898f5d14a30ee69e7bc"
+          },
+          "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/references/sources.md": {
+            "bytes": 7984,
+            "sha256": "9425b1aab963a2c8d35f3db897bd15ee69699bf5b0d0c8425c3c8bba3e9a0233"
+          },
+          "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/SKILL.md": {
+            "bytes": 3068,
+            "sha256": "25d04aed9782b2a8ff1186828becdc7fce93c4418039ceaef9b8dcc5ec9bcd24"
+          },
+          "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/metadata.json": {
+            "bytes": 478,
+            "sha256": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7"
+          },
+          "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/references/source-notes.md": {
+            "bytes": 8521,
+            "sha256": "9f9184d46b19312138a6e805e439e43dc0594a0cc6b46675255a582dc946ca5b"
+          },
+          "src/inferrsi/data/skill_library/methods/dense-performance-evidence/SKILL.md": {
+            "bytes": 8305,
+            "sha256": "53cc1867a13f54d1006d50517dea8848869ca50df3f0cd610b2424cc8d18c933"
+          },
+          "src/inferrsi/data/skill_library/methods/dense-performance-evidence/metadata.json": {
+            "bytes": 707,
+            "sha256": "a0d1e3b17140295510a764045d1b222e19be3a3406e6d1e7dddaad49ff474490"
+          },
+          "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/SKILL.md": {
+            "bytes": 6667,
+            "sha256": "82d0099d3c33706f46c657d4fa7f76d3d486e019b36760d1de001921328a60cc"
+          },
+          "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/metadata.json": {
+            "bytes": 547,
+            "sha256": "37379a4d4a058ca46da2b20a920dc01be9ec11e557c32b2332702ec76027f4b5"
+          },
+          "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/references/source-notes.md": {
+            "bytes": 44796,
+            "sha256": "ea416831fd7d0874e06d0a76fb0c7196df53072b717673a0cdf635db6bfd917b"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-benchmark-design/SKILL.md": {
+            "bytes": 8974,
+            "sha256": "19ebc1f567aadf49f7a534a18fdb629c72c0e4ce043d360983136491cb72bfab"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-benchmark-design/metadata.json": {
+            "bytes": 5923,
+            "sha256": "18310402a8ade03fd48ac687f79ba9d447871e90337de84c0e1bc15c0b827306"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-benchmark-design/references/benchmark-commands.md": {
+            "bytes": 8466,
+            "sha256": "0898424e0bff3a6246e1eeeda6abd6c5bd974c047fe32f3bbd5ab9d31ce17cd3"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-benchmark-design/references/task-benchmark-selection.md": {
+            "bytes": 25086,
+            "sha256": "0d4bb080a754e26846e81680023db0dff5e6a4c96be4327c411c3b72ac4ef8e2"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/SKILL.md": {
+            "bytes": 6644,
+            "sha256": "89a6a82c692f7f5192f4d7f4c3a5ff08a4df3b1c4bdaa0aafb7e04292d8aa5b4"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/admission-drain-occupancy.json": {
+            "bytes": 1354,
+            "sha256": "3319e23bbe2bd40e2d45fb89dc87bdc92d4df2b5ddd8086fcf84813b43885b7c"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/admission-drain-occupancy.md": {
+            "bytes": 3134,
+            "sha256": "e88bbac5d253a8e7e3a3d5e21047f23f2970c11ac23887d4303ab94fca52dd27"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/drama3-rejected-directions.json": {
+            "bytes": 1459,
+            "sha256": "2cc0a526fe0509551607a03e7626dd3d7844e643c4afb194837f75cddead8b12"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/drama3-rejected-directions.md": {
+            "bytes": 3557,
+            "sha256": "6f143bc363a3290fc008c08e57cb075f1776d1c8cca8a6a59297d5b1fbbf61bf"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-ascend-rejected-directions.json": {
+            "bytes": 1293,
+            "sha256": "b3d538f79c56724ccb2895cbf50cf2e6b0fec3979be62a684a8230cd932dff55"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-ascend-rejected-directions.md": {
+            "bytes": 2564,
+            "sha256": "be396983812262515899a3e2ba4f01b23ff92e1b836881882c7580372b8ae552"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-b300-rejected-directions.json": {
+            "bytes": 1196,
+            "sha256": "146ea1564a46ed03092c7dc5796b318fc05b04b1821105ee00eb7fa7fa1333e2"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-b300-rejected-directions.md": {
+            "bytes": 2656,
+            "sha256": "cce2930f544a0f78f21797fba4d8ded298e4a81f76c34d7977b0e31d361bc738"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/metadata.json": {
+            "bytes": 535,
+            "sha256": "fb55067fb45970319787068a3206110641cceaf0ccbabd706be6d0b3d2afc2ed"
+          },
+          "src/inferrsi/data/skill_library/methods/kernel-writing-guide/SKILL.md": {
+            "bytes": 10029,
+            "sha256": "01fb1712d7eb988c5e412f02340d2e48b4f8564fcf4ab232dc5897869d0d50a0"
+          },
+          "src/inferrsi/data/skill_library/methods/kernel-writing-guide/metadata.json": {
+            "bytes": 653,
+            "sha256": "0189d184337662056fc6952c7beffc1bf6154c4fa173ae3e3e463bd22d70f36c"
+          },
+          "src/inferrsi/data/skill_library/methods/kernel-writing-guide/references/akg-affine-tail-guards.md": {
+            "bytes": 11941,
+            "sha256": "113deb2db10b171db3425bb3f1792bee763706b794c8cc2e222b4cc0d0114118"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/SKILL.md": {
+            "bytes": 12914,
+            "sha256": "882e7850b9b49ad692a2d11618701afe0f4e83727ea78df114be2029923c725a"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/metadata.json": {
+            "bytes": 1403,
+            "sha256": "9f1fd8a2af52bc964021e21f16ecbdb891e08f2057370e700d67f86605d333db"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 3597,
+            "sha256": "157e74ba180426b279b548a729769f152bf18f9a78ffca4ebfcaf26b7d0c5292"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/MECHANISMS.md": {
+            "bytes": 5875,
+            "sha256": "18bb2f5d7e52773c516420a142edfb531d0b9c822a2934d60f3234ac1cb9d40e"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/NEXT_GATE_LEADS.md": {
+            "bytes": 941,
+            "sha256": "16a264c7159f9c2c700ce8ba46e63f3e636b277c987da4970f6d2fa085fcb053"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/knowledge/adaptation-interfaces.md": {
+            "bytes": 3893,
+            "sha256": "5e761957cb5ecd93926e6ffc13deee7c830c6e8f0ee3c3846e9dd3b50dc74446"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/references/source-notes.md": {
+            "bytes": 13370,
+            "sha256": "54d83c3713045d59a48821bcf93598374ba4a630e3891ad39498f919317c4fb8"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/updates/optimization-search-policy.md": {
+            "bytes": 4238,
+            "sha256": "d3db44fa9a6831b86a02c638b323ed943e69fe256866213f10d68c72d762879e"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/SKILL.md": {
+            "bytes": 15324,
+            "sha256": "c5f78961c23e3c8d659f5675b522518ab06e1f2a8ea8bc74c5d428ffb41693d5"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/metadata.json": {
+            "bytes": 3669,
+            "sha256": "60b5b24ead590b0fb222398fba47822bf6723690da683e256a450bb5c0eeb806"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 7349,
+            "sha256": "083e462e0a204dce599362d46712fcfb392d053c5cda56216e22e28d4b8fa0fa"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/MECHANISMS.md": {
+            "bytes": 4003,
+            "sha256": "742c249082ceca230326f4407a40130fe624553b796ae5a157a8a4c15a7e1398"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/README.md": {
+            "bytes": 1204,
+            "sha256": "17b2d1c7c14124d59ac8a0d5ad4200ad2803d991e784a1a98cd97188cc364999"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/READSET.md": {
+            "bytes": 3937,
+            "sha256": "66459ce393858b2b2160f1e2973373bfc4b9902a12644631934b4ebe3140296a"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/STATIC_CHECKS.md": {
+            "bytes": 1111,
+            "sha256": "cb0aa7d148352ebb7b1678fc42a557a099575054716a341dfa25b6a998029064"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/cases/decision-cases.md": {
+            "bytes": 5491,
+            "sha256": "781a675a6533e0a18c044aaaee0d010da7013b6504654b8b8292a35176db801e"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/govern-optimization-experiments/procedure.md": {
+            "bytes": 3823,
+            "sha256": "ce036a4d8a0908da204515c47e2ff9623548d09de264fa65bd58fedce52a78a0"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/govern-optimization-experiments/references/source-notes.md": {
+            "bytes": 21659,
+            "sha256": "f9a8254e45473ddb2b73e576957b6d3db83c62b99dff1922f4d1e606e569559e"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/knowledge/bounds-predictions-and-decisions.md": {
+            "bytes": 5297,
+            "sha256": "5780d6eb5257612ec45deb94f85f334a6ecabf7e466d1678706041f6549bda79"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/redirect-cross-layer-optimization/procedure.md": {
+            "bytes": 3648,
+            "sha256": "6593a63fce9c474b515663a560c94b7df048dc5f6e77a95b8e54e4a64cd10a09"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/redirect-cross-layer-optimization/references/source-notes.md": {
+            "bytes": 26469,
+            "sha256": "6ff3a7bc31f1172db92aedaf7df1f1aac688bf769227b583a4cd82bfad52da67"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/agentic-kernel.md": {
+            "bytes": 41353,
+            "sha256": "7190573b85f94a4391bb0d887ccbd37a31b3f08bb0531114ad7afd72fe1567a7"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/flashvector.md": {
+            "bytes": 9830,
+            "sha256": "3926369acd88ff3c370bee5718e74dd15f3219c3e4f6311095e30e917bffa4cf"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/inferencebench.md": {
+            "bytes": 9294,
+            "sha256": "21f47111197a0b7f8d6384ea918d139dc8318a62ca3313ce060048271f50b639"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/prior-coverage.md": {
+            "bytes": 3254,
+            "sha256": "35d47ac078591221fcb666219bcd824f9400fc17aafecdc0499d8336f7178238"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/vidur.md": {
+            "bytes": 7736,
+            "sha256": "d38602af5219f5feeb49e6cd6bf53d66098c77fad141dae676f1d8a6743cf25d"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/roofline-contract.md": {
+            "bytes": 7091,
+            "sha256": "419ec4bcaec45ac579a133bce083d4f48779f26e664d7a86f720fa65f343b1d1"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/roofline-example.json": {
+            "bytes": 2015,
+            "sha256": "597cd3205df043cff3ae9d15924c77934e8d8eb300c00c4840f6bb2921f90faf"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/scripts/roofline.py": {
+            "bytes": 9004,
+            "sha256": "524de48a6764401695fe1444f1fa3d38f363069b0cd2fba4715a86ca4b28a5d4"
+          },
+          "src/inferrsi/data/skill_library/methods/plan-activation-residency/SKILL.md": {
+            "bytes": 3507,
+            "sha256": "542d8a1e6963f5019986e1fbd79d3a247a045a51487c09313026087ac457c15e"
+          },
+          "src/inferrsi/data/skill_library/methods/plan-activation-residency/metadata.json": {
+            "bytes": 478,
+            "sha256": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7"
+          },
+          "src/inferrsi/data/skill_library/methods/plan-activation-residency/references/source-notes.md": {
+            "bytes": 13907,
+            "sha256": "5ff52655cfd06fcf970ed2a7d4d4fe1c31115b102af2d41497de0c74848ed2b3"
+          },
+          "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/SKILL.md": {
+            "bytes": 13800,
+            "sha256": "7e20d3390f2501263e5e298e026c47cac16c234f2397833131e15ae5d34c0b97"
+          },
+          "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/metadata.json": {
+            "bytes": 1903,
+            "sha256": "d0e4d6950250e4cec43d82e23324c9301a925d4c22be30b4d2713d790cb6504b"
+          },
+          "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/references/capability-presence-audit.md": {
+            "bytes": 11461,
+            "sha256": "c21eb4a6aabe811a49a6540a34c54dbcfdcf7efae3052a79b54034c7838026a5"
+          },
+          "src/inferrsi/data/skill_library/methods/skill-evolution-playbook/SKILL.md": {
+            "bytes": 9448,
+            "sha256": "9cd37f21fe0d862069e4798edca8a04d59f8b840e55d521901594a63dddd9ba3"
+          },
+          "src/inferrsi/data/skill_library/methods/skill-evolution-playbook/metadata.json": {
+            "bytes": 527,
+            "sha256": "92b36338b9257e2e41b95e4370694d7de52ed84bc7971056b3bfced07b0993d8"
+          },
+          "src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md": {
+            "bytes": 6865,
+            "sha256": "64bd2c6b795d966045d1d3cca55976d3d49b08b06df7c7e06a70c2824cf55b7e"
+          },
+          "src/inferrsi/data/skill_library/methods/training-input-pipeline/metadata.json": {
+            "bytes": 587,
+            "sha256": "80659099fce6a6ec05e9084b75452c877a5812e30f60a78f31ef8c7efcdc1bac"
+          },
+          "src/inferrsi/data/skill_library/methods/training-input-pipeline/references/sources.md": {
+            "bytes": 2637,
+            "sha256": "60e0f3e88655d8e1d5eb4797f39ca4370deda01614f265b8d2ea4559d8ac020f"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/SKILL.md": {
+            "bytes": 7200,
+            "sha256": "bdfa5a4ef37d9f303c3a23cd6f98c21137b97ac5ed58e0ed5fac6d71d33e52ef"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/metadata.json": {
+            "bytes": 976,
+            "sha256": "d42b614af9cee6ee0d4a25dc5d57c7b98438feed1e807a7c9f1755109a86e9cf"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/mindspeed-historical.md": {
+            "bytes": 3384,
+            "sha256": "eb97f935ed30517a8fc9edec5e9cddc04f5d679d156c16d75bd29c534593360a"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 7368,
+            "sha256": "c55c6ad2131ef788866411dcc7056f221420ebff30c95c6bc919b09a4818cb73"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/cases/paper-grounded-cases.md": {
+            "bytes": 5649,
+            "sha256": "19bd25b1a944adb2351ebc6b1d795150ea033041a44f6cc261e881677812a101"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/knowledge/offload-delta.md": {
+            "bytes": 2026,
+            "sha256": "f8740fc5c8bbbe5ac13ff1cfec7ca499569d0b46e8cd56cf1f5a44c3a5e72242"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/notes/training-memory-scheduling.md": {
+            "bytes": 18660,
+            "sha256": "9aa6d3e01452dff95c8a8940ff62ead8c146ce215c9175b78c9f2ef508a7b2de"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/references/source-readset.md": {
+            "bytes": 22946,
+            "sha256": "2a5c166767f56fa47e7eba7386f1f36da8cb73702e5b5bbe14d6a456ca9ebfae"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/sources.md": {
+            "bytes": 2752,
+            "sha256": "6d248f58af8b350fc66c22898988d6f451ac60d03bcd805cd6e84df5e55174ac"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/stas-decision-notes.md": {
+            "bytes": 4086,
+            "sha256": "f1a993e6c4ca9a035a2effcaf5500081f227b0cc56cc885a6b22fda0041fc744"
+          },
+          "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md": {
+            "bytes": 25969,
+            "sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213"
+          },
+          "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json": {
+            "bytes": 2418,
+            "sha256": "798cbff6f10b7c24b188080ebf233b188b0e98ef2c6fdb5e00e1fe3eff727c0f"
+          },
+          "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md": {
+            "bytes": 8656,
+            "sha256": "5a39f14b102b0cee304541e011aa31e44e6b13527d1bee9987c3f3b66c064d3d"
+          },
+          "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md": {
+            "bytes": 6392,
+            "sha256": "6c640f28ee3f0253a17b81461c865cb9f947367002cb75bf72550c4572d02e97"
+          },
+          "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/SKILL.md": {
+            "bytes": 7181,
+            "sha256": "a46110891eca860915e314d6921a22856d631453b22b17837fe30262d6780535"
+          },
+          "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/metadata.json": {
+            "bytes": 2175,
+            "sha256": "4e6100dfb9fc7bc00b19915f4fd38ad509704f2a609c658adf4d632d4e05b6c1"
+          },
+          "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/references/afd-plugin-LICENSE.md": {
+            "bytes": 12747,
+            "sha256": "0e8566ea362276878e1f38f97da49964dbea301c15d4aa9d727f97a4c39fd943"
+          },
+          "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/references/source-notes.md": {
+            "bytes": 17240,
+            "sha256": "5073d87450cb60d1b245308dff8807e73749c07de976d50f2fa26694fad81a72"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/LICENSE": {
+            "bytes": 35149,
+            "sha256": "3972dc9744f6499f0f9b2dbf76696f2ae7ad8af9b23dde66d6af86c9dfb36986"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/SKILL.md": {
+            "bytes": 4507,
+            "sha256": "c993bd6ff8e9b5f82f39f175ec8242341136e4ee0808af414e1d94ad3e196c5e"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/metadata.json": {
+            "bytes": 2032,
+            "sha256": "74742851a002a51c208362cdcd2a59fdab5d1734d8492b1f93fbec3cf21e0841"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/references/graph-cache-invalidation.md": {
+            "bytes": 7679,
+            "sha256": "614c9a4e6943d24d7d7be7d632682ce8b1dcc239124a37d9dc1d487113772eac"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/references/model-residency.md": {
+            "bytes": 7191,
+            "sha256": "014c68386e7cdfdb7b4f566836e8b2ff61bd5a5a506e7b4723f83fe80c5d4de7"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/references/repo-provenance.md": {
+            "bytes": 5648,
+            "sha256": "ca996ef515af42a3ce304747fa3f12a31c00533deaace5ba2ce21aefa0a62e2e"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/references/validation-boundaries.md": {
+            "bytes": 3523,
+            "sha256": "1913dedbeb558d872d6a0e015cc11d822910c3cb259a236db29fb60ad8a64263"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md": {
+            "bytes": 12294,
+            "sha256": "4703beb6e5194ef2ff2c0b5ec69f64e35417a4bc0dfb50a10c80a920b77ef3b5"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/metadata.json": {
+            "bytes": 2333,
+            "sha256": "014be24d0e669f4acbc0717dc323eec784026cf185186619d5a19d6d84496131"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/source-notes.md": {
+            "bytes": 12036,
+            "sha256": "fd917bd5ac80bee9c9f4525f8e0300f6d67dc144e2a16632af42a67f3d53d098"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/validation.md": {
+            "bytes": 5402,
+            "sha256": "60e71e70c5519bcdf6c171eec2be2f342e8801bedd746d6005f11adc646dfaff"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/SKILL.md": {
+            "bytes": 11202,
+            "sha256": "f157562d1edc905cb4fc07246567e3c5b844efb9fbf5fe39f0d40553a1b1a307"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/metadata.json": {
+            "bytes": 2156,
+            "sha256": "4d68bf4c1767eaee12ade963bbd551d0e61d1bcd81a984bb02fd3aae46965919"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/references/source-notes.md": {
+            "bytes": 11633,
+            "sha256": "6fa955022f16cd598f44fd6521901d1cb868b06cec545cabe2b82769a641b5d6"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/references/validation.md": {
+            "bytes": 7196,
+            "sha256": "673d93f6700052974b340fab25b3438e0e6344960569ac9abec7d7720f662288"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/LICENSE": {
+            "bytes": 1077,
+            "sha256": "663e51edcd9a3a8f858018ce6924ad2895a24c269c26c39ad75325bcd0f7e5bc"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/SKILL.md": {
+            "bytes": 5432,
+            "sha256": "fdb4d9483c6efa836f9e59617da605fa9c2186e11c2041b05d5c81850ba0475d"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/metadata.json": {
+            "bytes": 4923,
+            "sha256": "fe626a77af3f084f36fd73ffdb6f67976e7dceeafb94ba22078ef03c3a96cfae"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/THIRD_PARTY_NOTICES.md": {
+            "bytes": 1774,
+            "sha256": "a0d77db7d2aca0eda811ab3c386cc32b3143f2c2609a042cf4642954f0f6fb09"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/minimax-h3-apple.md": {
+            "bytes": 12059,
+            "sha256": "57ce01754ba96c378f13a91305ef3d418b926f99b2753f0af99315572bce063a"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/minimax-h3-weight-inventory.json": {
+            "bytes": 10664,
+            "sha256": "fbaacaeec340d3ce893556ee864405a0746af4d13530a38cb3d5aa9d60929271"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/phase-residency.md": {
+            "bytes": 9408,
+            "sha256": "2836a550eb2be7f5a862f51ec7b135ebd9715bf3c4faa84a7ecbc9d9c11f5d8d"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/precision-and-streaming.md": {
+            "bytes": 9440,
+            "sha256": "7e166c89193f4c7aed852ec95346a2de2b5008dac002e2e1862be1b3e57f6093"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/source-identity.json": {
+            "bytes": 7747,
+            "sha256": "730f8dd23648a3f8f47b9726cc613d893a79d1817d515eec609a9b06d8a9ff90"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/source-notes.md": {
+            "bytes": 9448,
+            "sha256": "7304bf9369efc157dc7aa40686dd828cc14db45c49442e908b3d3cc4b329eb86"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/validation.md": {
+            "bytes": 7880,
+            "sha256": "0754873a758e048480938314894e25b504ae0c269549312e7f8e6020d83fe294"
+          },
+          "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/SKILL.md": {
+            "bytes": 9338,
+            "sha256": "95e81b6bc096a954fa73139ea5bc001d1efa6c99747161da62f012b494e5c013"
+          },
+          "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/metadata.json": {
+            "bytes": 1718,
+            "sha256": "1d30b31dc8352fc6a8d7b8abbef8118c840e47e5b90795dae0c758bf9717278c"
+          },
+          "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/references/source-notes.md": {
+            "bytes": 15234,
+            "sha256": "3d12a401e9decb55a2f4f3b2bfa0d360ef5110f63a0469dfe285654e78054190"
+          },
+          "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md": {
+            "bytes": 10812,
+            "sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27"
+          },
+          "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/metadata.json": {
+            "bytes": 572,
+            "sha256": "c221189b8d85fac3bd41ef1c7cd560f1aebc8cfa0381c182d1d571a7169be829"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/SKILL.md": {
+            "bytes": 3788,
+            "sha256": "bf1177cb9985c12a9bf11edb64771018e1e314c4713920de8fd4a989ff70115f"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/metadata.json": {
+            "bytes": 1299,
+            "sha256": "d86cecb615e93ba0f6be09d1ead102cd5a17068c07e1be2549ac5cd8f7519279"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/cache-stop.md": {
+            "bytes": 2026,
+            "sha256": "bba6b5c73261065309cbcf11b1e591eef48916f0c79ceaaf74d98a9fa96b0992"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/generation-batch.md": {
+            "bytes": 1733,
+            "sha256": "f16d3c57d45e63bcaffd3871aa41c76972ee11f7dbf1e12b45d9b170303cad9b"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/source-map.md": {
+            "bytes": 6358,
+            "sha256": "91d26181f246307f2113f94b12ac9587d689b654dcd216e9f14b11cf5ebb650c"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/validation.md": {
+            "bytes": 1665,
+            "sha256": "73b5dd347017baba519ecc85415aead9993c1751d116b9ba4f07689f58c5f783"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/LICENSE": {
+            "bytes": 11352,
+            "sha256": "a54c8553df5d4313d293d1aef3bf2f13952bda8b7c7a1fdd729b2bbe27ee6015"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/SKILL.md": {
+            "bytes": 2304,
+            "sha256": "d75f5b2881b240ec2d9d4f5cc8e3f7d53e4378cb01732a2fb9699da2901ba38c"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/references/repo-provenance.md": {
+            "bytes": 1778,
+            "sha256": "5e8dbe168a8885e209d4b4532944ffa7fead658b23fd6588054d57b22476f63b"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/references/repo-routing-metadata.json": {
+            "bytes": 511,
+            "sha256": "6820707cb9bd877d1635928cf5a0efe38e4fe37dc04036dc70d53e479ed9bf7c"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/references/snapshot.md": {
+            "bytes": 3488,
+            "sha256": "0fb5d776349de845dc314f16ce73a06c9caf5db84dc1c293327d3b3ca1337c80"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-architecture/SKILL.md": {
+            "bytes": 2355,
+            "sha256": "766b2d66d2056dd779abd9e492cb3af6e25899298f44964f565509f130a7cb24"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-architecture/references/runtime.md": {
+            "bytes": 7734,
+            "sha256": "36bba9874fe054acc5217f9c265f5ae309c83943e0d46228370ae1bcdc69421e"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-extension/SKILL.md": {
+            "bytes": 2193,
+            "sha256": "9af38bca840d0004fc2ed049141476fc40fd40dcfa843f38a404a7a314c19425"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-extension/references/integration.md": {
+            "bytes": 5775,
+            "sha256": "6ce9383ca17bacff60a180f8fb8c43ea47eaa37117fb8da6051ad86135d228ef"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-serving/SKILL.md": {
+            "bytes": 2335,
+            "sha256": "990b6bdebfe22612de49dbebcb16810e8abef94d70691aa883b7937c85210eb6"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-serving/references/recipes-and-tuning.md": {
+            "bytes": 8918,
+            "sha256": "94afb762bbe185f58606154b3f0a5af56e7089cb84ea94e26cfcb9f93c95439f"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/LICENSE": {
+            "bytes": 11346,
+            "sha256": "1495e1e757ef4d0925a2350563cf5754bb23c51701a8ec4fb3c5cdcbedae6747"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/SKILL.md": {
+            "bytes": 5466,
+            "sha256": "ae72d1f0cc750699bf79d7a138a97219209bca7b7e5f19cd427d306245eaf841"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/LICENSE": {
+            "bytes": 11346,
+            "sha256": "1495e1e757ef4d0925a2350563cf5754bb23c51701a8ec4fb3c5cdcbedae6747"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/SKILL.md": {
+            "bytes": 5738,
+            "sha256": "b0a4ba4751087126143ef86983798deb0ed00e24b6ff06bf6ef76a9df6e274f2"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/metadata.json": {
+            "bytes": 2844,
+            "sha256": "6ed57dfb4f25ecb34b81f2188b055e05d172ba2581e6a3b605b814b2f733a4d6"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/references/source-notes.md": {
+            "bytes": 19512,
+            "sha256": "d0fcfd1177ae2f351041c5b2abba7f5121cec133006d9e4dca8f74ddb567332d"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/architecture-and-memory.md": {
+            "bytes": 7735,
+            "sha256": "798e747022b7ec135cc4861c2c9ac44fb14b503bb77a0068ee7fbea494b764a7"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/native-candidates.md": {
+            "bytes": 3961,
+            "sha256": "abafcc9a34df196a6365145554887b8d37afb36c6e236c92e50a4a06c8a4ec93"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/parallelism-and-extension.md": {
+            "bytes": 6324,
+            "sha256": "e3be7f4b51264ac68682a64d264e4d96857d5bb9f9d99cae5f327c56984a9e6b"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/repo-provenance.md": {
+            "bytes": 2274,
+            "sha256": "534d95ae8d0b29bcbc8f6707b3d38fb964e31f5b66ba13517a9a6b8f309bbb38"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/repo-routing-metadata.json": {
+            "bytes": 625,
+            "sha256": "b2b879c1cc42ddc2a5b95ae24115eb891562fa9a6e96d151f4ddf6f7719fe73c"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/source-map.md": {
+            "bytes": 8945,
+            "sha256": "c2adad64ceb0c305cdc4859c03012b3e0a4ab26cd0cb289bbd76e192b0cc3794"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/tuning-and-validation.md": {
+            "bytes": 10832,
+            "sha256": "49b3343b3292cc098a787de9d24d968b444d0ef4f73fedb5d8183390e8f39c9c"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-parallel-extension/SKILL.md": {
+            "bytes": 2316,
+            "sha256": "18bd54ebdd629e1ba6dfa589f391c7cc3c5355785c2ae82f741b64ee8022464f"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-runtime-memory/SKILL.md": {
+            "bytes": 2154,
+            "sha256": "33b1a65aba00fca819d2b6cc42202e7708e9155e4d8f4373a2387f4ee2f2fa3e"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-tuning-validation/SKILL.md": {
+            "bytes": 2018,
+            "sha256": "744b2ef9122101d3f50cbbb40fe020e2b3d3888adbe2d86041041e7c7ce424fc"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE": {
+            "bytes": 1062,
+            "sha256": "b53fa08f515cb5a6fff7b9fd8fcd0961a4b80df29d74372d25e1f9171aa042ee"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md": {
+            "bytes": 5504,
+            "sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json": {
+            "bytes": 2202,
+            "sha256": "f7b7581a3f44ba9774480c559408fc66d6a4d6d1dbbeabe4529f1eb7d4dd5ad6"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md": {
+            "bytes": 8136,
+            "sha256": "7b537cb972347dd68e303960f142892346d8b19b884e4d08016ea36a23bdaa7a"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json": {
+            "bytes": 3617,
+            "sha256": "af4830370b6eddc2d176a9ae14da4ab56a90b0fe048a01a89ee051df4365d8e9"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md": {
+            "bytes": 6589,
+            "sha256": "b67affbd7d237d82b7f51be49f1537e9820fe56ec70a71ee76809f9e599ca693"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md": {
+            "bytes": 3394,
+            "sha256": "5dcf5fa7bf734fbe13214f57d2dfb40ff8215d9cd34e072a9e615bf103b82fc2"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/LICENSE": {
+            "bytes": 1093,
+            "sha256": "949b9b0aada7476006469e58ee4ffb0e0f68c2d9137e90a356757b61a09d37f0"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/SKILL.md": {
+            "bytes": 2813,
+            "sha256": "0e5daeb6409c916be45ecd7ce2531ea5f1c60a4a53e2b0ed25dc1e2cdc5c17ed"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/design-space-analysis.md": {
+            "bytes": 27736,
+            "sha256": "11ecafe71530ddd4839268ea1b24b3361a3895012a705859ca06dae29fa72630"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/repo-provenance.md": {
+            "bytes": 2913,
+            "sha256": "05f17d0ae996ad6c12314bdd29f3427574f6aa611ea60effd18598a0f386638b"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/repo-routing-metadata.json": {
+            "bytes": 332,
+            "sha256": "a10ba49b7e6565109b566c105e17551d042075a1ff649c6f371134e3b7402c74"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/trial-format.md": {
+            "bytes": 2291,
+            "sha256": "92d8cf38549268ae4eef382890662d03e18a7dff858ad45ee5eb6f47eb39c8c0"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/troubleshooting.md": {
+            "bytes": 2079,
+            "sha256": "7301cfa5a5f6cfbf9576c3845a96c6b4de8f8a6509f2330970a8e2e4669b3d5a"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/scripts/summarize_trials.py": {
+            "bytes": 5170,
+            "sha256": "6a243a08da15659ced6519f49f52a49a675bd9e952740c3ab03dd21964eefbe2"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/design-space/SKILL.md": {
+            "bytes": 1826,
+            "sha256": "857bea3c64db45dc035728a41523142ac65778b41f63092148b8ab936f971d76"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/design-space/references/design-space.md": {
+            "bytes": 24231,
+            "sha256": "0a428dcd12d5290b554e9da6795a110f94d98f128771602e1ea5f608048b135e"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/tune-execution/SKILL.md": {
+            "bytes": 1145,
+            "sha256": "cdf51c537e7237296d52ef6cda9f2e4c9d9ce3fc87265a369cd9eb7497c53512"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/tune-execution/references/tuning.md": {
+            "bytes": 11310,
+            "sha256": "7b69aa7872d60dd3fb91e41b77a632572070dcf80dd5b5f6dc1722d8540177b1"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/validate-engine/SKILL.md": {
+            "bytes": 4613,
+            "sha256": "90bfd975e25f18895ce0e85f6f35d24149ce026305a4d479aea4924510053460"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/validate-engine/references/validation.md": {
+            "bytes": 16895,
+            "sha256": "dae846f4c47c68731b96ec1d50cb008c03871c72078f1c5b0f4a9423be9ba397"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/LICENSE": {
+            "bytes": 11357,
+            "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/SKILL.md": {
+            "bytes": 4945,
+            "sha256": "df3e63dc28de2dac2aa92824fd88719159138650d96487332e0effca3f0e2a80"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/references/evidence.md": {
+            "bytes": 4689,
+            "sha256": "ae0d51ba56b34ed81fba34de8fd3685bebf2f160b188fa53719936002881d220"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/references/repo-provenance.md": {
+            "bytes": 2221,
+            "sha256": "43bbfe431434ca4ad6157ba35cbb9ce2d89e4065f427abfdd5f372f51ba0c9c9"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/references/repo-routing-metadata.json": {
+            "bytes": 342,
+            "sha256": "7784b4d98d096718c4f4cc42ce86d5ec113c469844e12de4690b0fccca59d329"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-architecture/SKILL.md": {
+            "bytes": 1336,
+            "sha256": "fb3ff50a7b9122a2545a97ef5179011237c4c4ca214dde7236ad7bf79a7c8623"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-architecture/references/architecture.md": {
+            "bytes": 5600,
+            "sha256": "db845eb4207ddeedb8ae381f1b91704a42cb53ba0ee6c7ee581510ba1a412190"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-serving/SKILL.md": {
+            "bytes": 1343,
+            "sha256": "8ab3209b838635395b22deca6aa879172016531d45c2374eb8e498e7f71bb40f"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-serving/references/serving.md": {
+            "bytes": 5049,
+            "sha256": "9bd5eff300d9755137a7600c8f7a7753bb22d09c8853e08062d7e535662ba8a3"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/SKILL.md": {
+            "bytes": 2227,
+            "sha256": "93c83bc4b799b32f50ff54b870cbbde050a790af16d9b7b062039c4987c80148"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/references/abort-acknowledgment.md": {
+            "bytes": 8547,
+            "sha256": "e95ae5b2cc410917ddf5e9066890d0bef9264104529bdca5ac12bdf56fe80ae3"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/references/validation.md": {
+            "bytes": 4632,
+            "sha256": "1542061f9c76fff2acb6529a58238301484b0e23ec0c697d2fdf129e200ea973"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/LICENSE": {
+            "bytes": 11324,
+            "sha256": "50e6751797c50dedd75ef1b8a0d9e42f5f8472e9fbce91f34718e9f97b0c780a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/SKILL.md": {
+            "bytes": 2863,
+            "sha256": "38e522b8abeb0dc140a2194340432ed41dc8b569d33b63d5d4c59aa1092cd2d1"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/SKILL.md": {
+            "bytes": 6664,
+            "sha256": "313733a65cc48442b18dda1c4665139a689963a90e1032a2235bc6f40e3a8e06"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/metadata.json": {
+            "bytes": 781,
+            "sha256": "b43fc3a7ccae59c664fcdf9ff77962edcc07c2a3df8b59324db551642bd23e23"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/references/source-notes.md": {
+            "bytes": 13853,
+            "sha256": "3665b034594fc8cc767c408002f4e16f8803369242405027c7f37ce6c9ecebe9"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/SKILL.md": {
+            "bytes": 6565,
+            "sha256": "5fb9ea1d1373e110d9b506f69963c4857338b7dbf5e38ec9a49b59a0ff683ed0"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/metadata.json": {
+            "bytes": 1372,
+            "sha256": "78459416734f61a8faa819d9a1a3ad7b9a985e8bf8e149813c604d15be6bb92f"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 9006,
+            "sha256": "82b24be82451ee121892f61575d0cf4ecb9920813a25ccb26cce2b45dda6d8b3"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/migration-is-not-free-memory.md": {
+            "bytes": 2145,
+            "sha256": "2ac168f2725cba7fc92dfcc67c01bf2258dff6a2dbe1ddb063bc60c175d82d0a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/parallelism-change-with-state-cost.md": {
+            "bytes": 2069,
+            "sha256": "5accc81086e4500d66f894ef6af9951e1cbb76941f6c631aab5c5561a786d8d8"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/preemption-under-kv-pressure.md": {
+            "bytes": 2630,
+            "sha256": "462200a6d0b193ef01f32374a463e0f2838ca8edf33d9ab898def65ffd0b490a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/knowledge/scheduling-state-boundaries.md": {
+            "bytes": 7488,
+            "sha256": "d4ef29869959b835a5dcb06ddb2fd8e6cff49bb176994dc1d46eb9a852f6228d"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/fastserve.md": {
+            "bytes": 9289,
+            "sha256": "9b4e275843000ef453d508d9c3c9c3f8ea1de5545fbd01d7abe74baa8a4931a7"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/llumnix.md": {
+            "bytes": 10181,
+            "sha256": "f252ce33ac2db017701ca1d49813c5ec1996c4c92900cf35bb6bda7fc59c0f18"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/loongserve.md": {
+            "bytes": 12422,
+            "sha256": "7bc6d1e256d461b04acc8ed856cf99108faceda9cde729ab95bcca596fe019ca"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/orca.md": {
+            "bytes": 8826,
+            "sha256": "f5a3b5870d64226408a8fcda5d69483769d823f8a5efd7f11698c8da483ab195"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/fastserve-readset.md": {
+            "bytes": 7173,
+            "sha256": "9866bd0c78b2b638998b3b632f6bae26aef032462c81abee336d309af5b07831"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/llumnix-readset.md": {
+            "bytes": 9386,
+            "sha256": "3a2819d65ad24087c08530492ef39d8ea046ed1f780d53512f1d23e2b3578487"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/loongserve-readset.md": {
+            "bytes": 12490,
+            "sha256": "f1a383e7bdbbd5a5a472d7f6e1292c8a17d4910ff736ca5a6aa422e1bee74f8c"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/orca-readset.md": {
+            "bytes": 4796,
+            "sha256": "a497ef23b7e60c0fbc15d95b7e07cd20547cfe56f37c24dfe6ed70d79ef7a3cf"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/source-notes.md": {
+            "bytes": 9575,
+            "sha256": "84033fd514773f10c0bddd9914fabb980497ba2b308f226a922c3e1888413827"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/references/evidence-map.json": {
+            "bytes": 14384,
+            "sha256": "dd8bbab5d285ed3a1a19d2f06ad84e8c444251690d8f673def3e2681765ef6c1"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/references/repo-provenance.json": {
+            "bytes": 1313,
+            "sha256": "4e705a964697712e950e02bedc41320806a41311b7f4dcf32f923d60b1d2e919"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/references/repo-provenance.md": {
+            "bytes": 3261,
+            "sha256": "6e841c38c35ff2cd6ff8139cb1a6c8f75701647897f1089a94a5c62928609941"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/references/repo-routing-metadata.json": {
+            "bytes": 432,
+            "sha256": "a20b54f682eb52311e445ac62b95d1935c001296f5cf23b7764cb04477d535c9"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/SKILL.md": {
+            "bytes": 1835,
+            "sha256": "2a945b5fddacfb90d1cd7cb2b2f767067d641061423a65e1c19afecec7f3c06a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/references/capability-boundaries.md": {
+            "bytes": 6560,
+            "sha256": "c8081d03f04f7481080c46c7dcf728962729bc122f49f9502d50cd1e3d04bcf3"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/references/runtime-and-invariants.md": {
+            "bytes": 8265,
+            "sha256": "afa1f967545021f36bc1b59dbf6aa2c367f4d5ece4f08dce755e2017cf50dbe5"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/extend-validate/SKILL.md": {
+            "bytes": 1731,
+            "sha256": "495679e5b99341bf80daeaacd2d4c2699aba16bc16c447cbd8d2fe17f7559798"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/extend-validate/references/change-boundaries-and-validation.md": {
+            "bytes": 7624,
+            "sha256": "f7fdd2d699f09d96fbf5ca9382f9b56113c870db2688d8e1118eb8163519fc4a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/SKILL.md": {
+            "bytes": 1812,
+            "sha256": "0ef4d2e3a96cbb277557e169a3fc8edf0f0504aad03259771b432a33ee008f78"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/references/commands-and-measurements.md": {
+            "bytes": 5667,
+            "sha256": "76ab51083b14595c49fa1c255cb7e5ebb996f14930a51ab3cbe864cc334b5f80"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/references/experiment-decisions.md": {
+            "bytes": 7644,
+            "sha256": "242593c0d386e57104b57a09bc7db97c2bbbcdb905bb68a397cfe612e07fb083"
+          },
+          "src/inferrsi/data/skill_library/skill-index.json": {
+            "bytes": 7514,
+            "sha256": "8ce35e18f7629260d67a95617b9918eecff73ed8d9370dd3d4595c725ada9aa1"
+          },
+          "src/inferrsi/data/skill_library/workflows/cite-check-downstream-pr/SKILL.md": {
+            "bytes": 1093,
+            "sha256": "485932cdf3b1fba5dc69be73e75d38f86a7b79442d26e82a296c963a7598d226"
+          },
+          "src/inferrsi/data/skill_library/workflows/claim-contract/SKILL.md": {
+            "bytes": 1075,
+            "sha256": "3af1dbbe81d91cb05fc2ef15a2ae25d7592a08558e1a5535f49c52f33a488c55"
+          },
+          "src/inferrsi/data/skill_library/workflows/competitor-scan/SKILL.md": {
+            "bytes": 1088,
+            "sha256": "0ede58ace65a9324a538b4816add1684f230ba536d4364fe5a5dd2c0bbed3776"
+          },
+          "src/inferrsi/data/skill_library/workflows/draft-acceptance-issue/SKILL.md": {
+            "bytes": 1230,
+            "sha256": "ce983acfb7d7d3272775e7cf00c92e4ddadd97511b602575a76e5ab7ade01b56"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/SKILL.md": {
+            "bytes": 5681,
+            "sha256": "364beed2ca617e71f4f955bbd024740d411b94eeb9557d7e055d4463cc22caba"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/cross-repository-mechanisms.md": {
+            "bytes": 17050,
+            "sha256": "28efa1b25f548763180b70d8d0ddebb7d4e7227df189f871fd233fe1d4881303"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/import.md": {
+            "bytes": 3463,
+            "sha256": "c8a57db94768cc6cd4193932def75ba67ec453e56ee0e54f176fb82f1af53091"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/model-aware-preparation.md": {
+            "bytes": 9056,
+            "sha256": "787c017aade68ee1133b979d636fa24062302cef6174c001171c963947ee8da4"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/overview.md": {
+            "bytes": 8099,
+            "sha256": "7ed0a966239b9481b82dd349245cfbfe966c030fc916771c91fd0692464ffc56"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/repository-source-scope.md": {
+            "bytes": 11967,
+            "sha256": "85202ca209171c50573c2fc209acb1e937d8211e1652e4ca9c6cc4bce04d03eb"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/source-manifest.json": {
+            "bytes": 1613,
+            "sha256": "0ce1678803d1ede6b485fb6b4b1c88aba909056ee6e3b5af691899ceb2056240"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/SKILL.md": {
+            "bytes": 6502,
+            "sha256": "74cd863df2e80dc2bda1713d24207809173a3116f30325e3487e1e3517cfec1e"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/metadata.json": {
+            "bytes": 3236,
+            "sha256": "12fc9272bd5fbf02dd9b897d8ca40984b72ae54f20492ace5304711bbc4d71c3"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/COMPARISON.md": {
+            "bytes": 4267,
+            "sha256": "38587678a96581f5c2dc697d982127343565504bffa191352811dffcc165aa23"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 5557,
+            "sha256": "fbc4a96b6f6c477706eed83eaa4fcc29a14f881637629d6c004601ed049f2523"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/QA.md": {
+            "bytes": 1853,
+            "sha256": "3f3785ec39a6938a5d1f2e874e99153fe92c0e666dd3e78a499c8b47d891346a"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/README.md": {
+            "bytes": 2151,
+            "sha256": "221785e91c45588f82b3caab6d52f1ea1d102b16139c20a0802921e2789e6d28"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/UNREAD_LEADS.md": {
+            "bytes": 1038,
+            "sha256": "6c1d87ebe491d27666703149eb05ad582c7e5aff77f0ac5c11a7b89c946f96ee"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/cases/methodology-counterexamples.md": {
+            "bytes": 5213,
+            "sha256": "f9fc21753ea41eef77ca020d78d031e945e5dd835b5d09c5ff0e53749ceb2ca4"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/knowledge/evidence-routing.md": {
+            "bytes": 5282,
+            "sha256": "406d5d74c3fef4c1a23d495492e21a084c3c46495a1a84b385703ed104358874"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/learning-to-commit.md": {
+            "bytes": 2847,
+            "sha256": "9c2d1d738e896d05c00f53e6adf14c9df2501ecc3f0020e71b5a1c2192547086"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/protocol-details.md": {
+            "bytes": 6520,
+            "sha256": "41298c6bde3a28f0712fa06a18bbc4ac59aea6d17cc9eb1318744f4121e3bab6"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/readset.md": {
+            "bytes": 7086,
+            "sha256": "6f953d20d8958c68339c9ceeae348604d0fea3e9c6ccb9b64a63d3b210426a22"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/repo-to-skill.md": {
+            "bytes": 3151,
+            "sha256": "5ebdac2a2bbbe6614f9d4096bc9895bd685acdf76414e2c3c1fb7b50d57c738d"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/skillopt.md": {
+            "bytes": 3040,
+            "sha256": "148075daa67e19f4b4351710498438d17c279a9bce56b2d01cab26ff3be33e99"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/source-manifest.json": {
+            "bytes": 2431,
+            "sha256": "716cd49ed01824ee9f6b6fc55ebf2b24fdd5d5a3e677f6e28e34a472616bbe25"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/trace2skill.md": {
+            "bytes": 2917,
+            "sha256": "230d13b77ba819b5e3edee147523e8cad6bfce95afccc140fb4946d6c8747a2e"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/updates/bounded-updates.md": {
+            "bytes": 7630,
+            "sha256": "d02c5bd0109846d1e1987de074810321cb4af277cd3f66ab2ac268aadd5ee142"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/skill-factory-design-20261014.md": {
+            "bytes": 9465,
+            "sha256": "fb8ce1ad83b7a6ddc657e89ee2eebb710873b213cfd85bdd6d823c8691b3cb9f"
+          },
+          "src/inferrsi/data/skill_library/workflows/rsi-evidence/SKILL.md": {
+            "bytes": 3680,
+            "sha256": "a55610874b658b86a39a1a049efb4e4eab0140f6579ed74dd7718a4ca8cedaca"
+          }
+        },
+        "sha256": "6e19365a0ad8e2951af0f3106f93aa545b4ad00b6e0f0317a20f32065b6a05fb"
+      },
+      "environment": {
+        "filesystem_cache": "not controlled; process cold is not OS-cache cold",
+        "implementation": "CPython",
+        "logical_cpu_count": 8,
+        "machine": "AMD64",
+        "perf_counter_resolution_seconds": 1e-07,
+        "python_executable_sha256": "3470f7919170d235d7e6079691462c4b217745ec67ee612e745730e46d98f238",
+        "python_optimization": 0,
+        "python_version": "3.12.6",
+        "release": "10",
+        "site_initializers_disabled": true,
+        "system": "Windows"
+      },
+      "git_head": "4817f92121df73fd9cd5fb4a319e3c56a68b4a5b",
+      "knowledge_record_count": 108,
+      "library": {
+        "bytes": 2294958,
+        "file_count": 342,
+        "sha256": "be8726296e6186601692dc2d95d99bfbba76f0b4c758de3a88a9b2a523eaa37e"
+      },
+      "modules": {
+        "inferrsi": {
+          "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+          "source": "src/inferrsi/__init__.py"
+        },
+        "inferrsi.catalog": {
+          "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+          "source": "src/inferrsi/catalog.py"
+        },
+        "inferrsi.cli": {
+          "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+          "source": "src/inferrsi/cli.py"
+        },
+        "inferrsi.knowledge_operations": {
+          "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+          "source": "src/inferrsi/knowledge_operations.py"
+        }
+      },
+      "packaged_skill_count": 60,
+      "python_sources": {
+        "bytes": 274444,
+        "file_count": 53,
+        "files": {
+          "src/inferrsi/__init__.py": {
+            "bytes": 102,
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47"
+          },
+          "src/inferrsi/__main__.py": {
+            "bytes": 48,
+            "sha256": "935a1c1166b0c1ea35a82256345000bf2c73ded718d77773bc27a71ecce28f7d"
+          },
+          "src/inferrsi/_skill_selection.py": {
+            "bytes": 5790,
+            "sha256": "4640399f4013425589f6702e2db844820619ba35967d1c3519978eb307168440"
+          },
+          "src/inferrsi/arguments.py": {
+            "bytes": 12200,
+            "sha256": "4456600cf4a362cf069594fe36db46932cf5cb14ddd37683aaf3538f5cf76b4d"
+          },
+          "src/inferrsi/catalog.py": {
+            "bytes": 6855,
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5"
+          },
+          "src/inferrsi/claim.py": {
+            "bytes": 805,
+            "sha256": "5f24c4fa91b48f0bb4d4554531607763fc8fc108e90f332816f06cfb520b5cc1"
+          },
+          "src/inferrsi/cli.py": {
+            "bytes": 14205,
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88"
+          },
+          "src/inferrsi/contracts.py": {
+            "bytes": 119,
+            "sha256": "03000cb9246ce8c0967147533f7ad25df918ba2674839801453ffb870f496490"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/scripts/roofline.py": {
+            "bytes": 9004,
+            "sha256": "524de48a6764401695fe1444f1fa3d38f363069b0cd2fba4715a86ca4b28a5d4"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/scripts/summarize_trials.py": {
+            "bytes": 5170,
+            "sha256": "6a243a08da15659ced6519f49f52a49a675bd9e952740c3ab03dd21964eefbe2"
+          },
+          "src/inferrsi/evaluation.py": {
+            "bytes": 183,
+            "sha256": "482c40710d428206c59fe3696703248cd86c4b3c7d19ad41017c1369bd6d780e"
+          },
+          "src/inferrsi/evidence_paths.py": {
+            "bytes": 859,
+            "sha256": "cc46231e0b8334ef5742defdda20e0093e22b30c86af59df1fb3ae4b5caa70fa"
+          },
+          "src/inferrsi/execution.py": {
+            "bytes": 755,
+            "sha256": "f43d8cbb4d43a72a7ff8a10b29a5fb9cf4cb22af54dd35943283c843fba0da92"
+          },
+          "src/inferrsi/harness/__init__.py": {
+            "bytes": 177,
+            "sha256": "1006e4713003f44bfb32b2e385a918e1652b9e9833801e4555674a71e57fb0d5"
+          },
+          "src/inferrsi/harness/evaluation.py": {
+            "bytes": 4160,
+            "sha256": "9b00b2e93959f16b7b76bb4b257df7dcf48520a06fd6adae96e9a165a05e6e80"
+          },
+          "src/inferrsi/harness/improvement.py": {
+            "bytes": 5777,
+            "sha256": "3faa0f5928a707a98f35f45d1ca18e226c898941e6a07b3829eebeb53e70aeb8"
+          },
+          "src/inferrsi/harness/reporting.py": {
+            "bytes": 3566,
+            "sha256": "38c58f02c44e8c8aeacf46527e142be0e3142a83e4b2693246d0ca76c0ddd7c3"
+          },
+          "src/inferrsi/harness/runner.py": {
+            "bytes": 6567,
+            "sha256": "a566307b9b586655941b627f428b5121ed38d7ac80bccd1e1799e443fb0de911"
+          },
+          "src/inferrsi/harness/scenarios/__init__.py": {
+            "bytes": 79,
+            "sha256": "4e97913c806077f0cadfeb29d986d2fa75b55ab8766cb86e6e3e9a195a9a1f5a"
+          },
+          "src/inferrsi/harness/scenarios/external.py": {
+            "bytes": 1016,
+            "sha256": "84d5f6920501a9d882cc4fffbc721d2051fdbf7c90694bde9cbfba3c2b866304"
+          },
+          "src/inferrsi/harness/scenarios/media.py": {
+            "bytes": 2914,
+            "sha256": "67d979a3ed6f13a8e3e1e40a304f78e5a3e6c2556b3aa8a11c41156745d7050a"
+          },
+          "src/inferrsi/harness/skill_evaluation.py": {
+            "bytes": 11093,
+            "sha256": "b710b8e53758757efd68c703449eb4bab7eb7ec50ae3dfc52a010cdebd8c3bf5"
+          },
+          "src/inferrsi/infra/__init__.py": {
+            "bytes": 85,
+            "sha256": "8779897733779266841b0f79b84b1dc43f0fa57c013855c906c809829ecfc77c"
+          },
+          "src/inferrsi/infra/artifacts.py": {
+            "bytes": 1040,
+            "sha256": "e72e1bebdbb88e42de430238489252386851f9e6645ed7780f2aad2e90338120"
+          },
+          "src/inferrsi/infra/campaign.py": {
+            "bytes": 1028,
+            "sha256": "be505fba1592611bd84ceaca55afbac027205413a2639baefa0c3bb343230209"
+          },
+          "src/inferrsi/infra/commands.py": {
+            "bytes": 811,
+            "sha256": "1890f115b3a2c5c6832173bf60432ad595bd6bf1c69783ecd460c6f8961194f1"
+          },
+          "src/inferrsi/infra/environment.py": {
+            "bytes": 2067,
+            "sha256": "60e46c2aefc216db202b41f44542e7b893e08e0721c667b9ccc50411d1ac84ee"
+          },
+          "src/inferrsi/infra/executors/__init__.py": {
+            "bytes": 64,
+            "sha256": "327c3cd5c821c832ff260a41dba0ee1e439cdabefc9edfb9cad1cd38d14bc87a"
+          },
+          "src/inferrsi/infra/executors/local.py": {
+            "bytes": 3654,
+            "sha256": "37bcbca9f3261c8c58a8bf806c9da40720b18d0676c82e8ed24c3901ac9bbe77"
+          },
+          "src/inferrsi/infra/http.py": {
+            "bytes": 634,
+            "sha256": "f65d6c2b55b6a89694caaf6dcdf7d74f417e9ced103eca1b3f218e305c6b6cb2"
+          },
+          "src/inferrsi/infra/inspection.py": {
+            "bytes": 2415,
+            "sha256": "d1c66deb0cced529bca996d2f88b60f51fa3417ff1cae365c650fec808946ecd"
+          },
+          "src/inferrsi/infra/platforms/__init__.py": {
+            "bytes": 1494,
+            "sha256": "9e88e8b9a396ed90c221e1585135c0a1b58da41c1bb73ed9237c53b6684abfed"
+          },
+          "src/inferrsi/infra/platforms/cuda.py": {
+            "bytes": 3309,
+            "sha256": "9e0e4276e1b2984605493b14b82fba3e86f6635204719c318302e624f66f8b6f"
+          },
+          "src/inferrsi/infra/platforms/mlx.py": {
+            "bytes": 1971,
+            "sha256": "f3b81470db28bd6183ccca2e763c5e079686bba6cdfe6c08ebe2cdf64216f2e3"
+          },
+          "src/inferrsi/infra/processes.py": {
+            "bytes": 8151,
+            "sha256": "3b96ebc7a50913e78e86ff3dcba74c8ee8d03d547d9591ca6488f390ac1e6189"
+          },
+          "src/inferrsi/infra/targets/__init__.py": {
+            "bytes": 186,
+            "sha256": "1f3ea75dded7c6cc967754c7b6d4f30ef00da8191a749938aabc3c6bb8f86eb3"
+          },
+          "src/inferrsi/infra/targets/git.py": {
+            "bytes": 2910,
+            "sha256": "6327c4f126086f310b6de309fc7659c9759da28ed76efaeeb94a3d2fba29a0c2"
+          },
+          "src/inferrsi/inspection.py": {
+            "bytes": 692,
+            "sha256": "afa44bf4cdfbb58a318c21f9b292ca142d1bbb22750e35a4cc8244da5350ebb3"
+          },
+          "src/inferrsi/knowledge_operations.py": {
+            "bytes": 27444,
+            "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b"
+          },
+          "src/inferrsi/knowledge_store.py": {
+            "bytes": 13700,
+            "sha256": "5e794e28906ce4687aff7f14a507e267184640fc8a4fac19e0ced279ce6a6cb6"
+          },
+          "src/inferrsi/model/__init__.py": {
+            "bytes": 77,
+            "sha256": "3c1dba4f5159fd9fbc8d4d96e2a83c8441491c1b1dc7b42031a3fc1782ed9d9f"
+          },
+          "src/inferrsi/model/anthropic_agent.py": {
+            "bytes": 24696,
+            "sha256": "b0ef9cd542544292eeb90ab05ef410be9ed6e5d1a40d729890f49562c28c12f0"
+          },
+          "src/inferrsi/model/protocols/__init__.py": {
+            "bytes": 85,
+            "sha256": "5a2f54483b164de04e309dbebb87a700b213682b0f44a50d89a752dd71a4ce47"
+          },
+          "src/inferrsi/model/protocols/image.py": {
+            "bytes": 2047,
+            "sha256": "251a1f8f885da2cd8e9d3baba766c525afe293091606b4e8668c43ba2cb7cb5f"
+          },
+          "src/inferrsi/model/protocols/metrics.py": {
+            "bytes": 610,
+            "sha256": "17db16d6ef15b41fb60d8f09507893accc43f9cb3ad686b8dc4e9839713ebc56"
+          },
+          "src/inferrsi/model/validators/__init__.py": {
+            "bytes": 65,
+            "sha256": "1d4b6ffa968e283219ddcc8a09de9280238a8d65ae2dc8cd2b60889877a8eb0b"
+          },
+          "src/inferrsi/model/validators/flare.py": {
+            "bytes": 19975,
+            "sha256": "3696b0609c772f2222b448833608565a033123d7e92b11fbc545dd356fec590f"
+          },
+          "src/inferrsi/model/validators/image.py": {
+            "bytes": 3204,
+            "sha256": "79470b6b06a22d53ec6c40636a0b3d821a2e6ee2f7bb7aed3dac7241504add1a"
+          },
+          "src/inferrsi/reporting.py": {
+            "bytes": 127,
+            "sha256": "1f771890ab3e69b1aa9246b548a7825596bdf904f6c395dd218208084266853d"
+          },
+          "src/inferrsi/research.py": {
+            "bytes": 5213,
+            "sha256": "9bfe59dd17b5476c18ce50d671711053c62800521c50d6b710e8140a2b4dca04"
+          },
+          "src/inferrsi/runspec.py": {
+            "bytes": 4398,
+            "sha256": "dca0952425caf42095ef0c7e2feb8ef4740cc422ecb61e777f203962d9b87709"
+          },
+          "src/inferrsi/skill_catalog.py": {
+            "bytes": 9938,
+            "sha256": "65f5a1f3f6894009abb98ec2fec30a63cc1bb2f6ce6aeb01f9f7a3dce0439f2d"
+          },
+          "src/inferrsi/summary_chart.py": {
+            "bytes": 40910,
+            "sha256": "d29023d60f6be804c1b01c9eb73765d4824d6c3bb870fe22c99a24544bc1d8bc"
+          }
+        },
+        "sha256": "4faa529dc47f23c24f8c5b8e0110d8e413aab20df5ad61db591fb751564a6448"
+      },
+      "tracked_source_dirty": false
+    },
+    "B": {
+      "corpus": {
+        "bytes": 2883003,
+        "file_count": 458,
+        "files": {
+          "src/inferrsi/data/knowledge_packs/README.md": {
+            "bytes": 14172,
+            "sha256": "b877e67d9cccf0dcbe57419eeb3ff9374e5bd03ff8eec63adef1025c446140e9"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/diffusers/cuda/diffusion/group-offload-transfer-overlap.json": {
+            "bytes": 3277,
+            "sha256": "9bec3d912192efa2f2c853d3fb70bdae21a75875630a02294fcd084b18c30414"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/flashattention/cuda/attention/generation-specific-dispatch.json": {
+            "bytes": 3085,
+            "sha256": "90290d7bc80b1626863e0520c91bb196d93ebfe08190a450e9de59c6505ae8ad"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/flashinfer/cuda/llm/paged-decode-dispatch.json": {
+            "bytes": 3745,
+            "sha256": "0f2986494fcc9b751df71e598fd61caa68e28bc4b0ac401e4cc763bc3f75283b"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/cross-stream-dependency.json": {
+            "bytes": 3809,
+            "sha256": "b829315e55adf8e11f0bd581d5f45dd130d7eb3f81ce7de532f63c7b9c27cbb5"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/evaluation-completion.json": {
+            "bytes": 3841,
+            "sha256": "19d2e506ac4f50895776a83a062059aa7a052175c8e4b21dd0d46e1a5a6716a1"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/post-detachment-buffer-lifetime.json": {
+            "bytes": 4663,
+            "sha256": "18d78c0e542dacb42ca33e5264d92aef80fa06d163fd16d4feafd5ed0827ab99"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/pytorch/cuda/compiler/cudagraph-shapes-and-lifetimes.json": {
+            "bytes": 3623,
+            "sha256": "7fef4122ff794967e61f55d02a4a6b558d7e498d9513607f5ca86db6386928a8"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/transformers/common/llm/kv-cache-strategy-contracts.json": {
+            "bytes": 3072,
+            "sha256": "a2df26052112a9f3fc6242b2bd636ab13991442db57e56646168d6c52032d38d"
+          },
+          "src/inferrsi/data/knowledge_packs/dependency/triton/common/kernel/bandwidth-fusion-resource-limits.json": {
+            "bytes": 2941,
+            "sha256": "6e9673d1a600a6ad34820e4325f2a6417f44f0f96d8ed0bcf21466d998f19ee4"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/cache_dit/common/diffusion/residual-cache-calibration.json": {
+            "bytes": 3841,
+            "sha256": "bb9509ad0225a91b52bc6787477581d05f0ecd5e44fa59f08f718eea36a2ac45"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/bounded-windowed-replay.json": {
+            "bytes": 10959,
+            "sha256": "43613225e9ccdbd34c0f7e3e0767887b9bbfd0f38b2c21684677d99e31a432f0"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/speculative-verification-budget.json": {
+            "bytes": 10177,
+            "sha256": "a729e1fc00be44e1480a5dc65bbcdd58df8599681df74dbc651a5a49f27e9028"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/zero-overhead-deferred-scheduling.json": {
+            "bytes": 8191,
+            "sha256": "ed5426af145f066f6183bd316c4073c46f324f5ee83aa9bd616387c0d0e7149c"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/lmdeploy/cuda/llm/kv-quantization-capacity-cost.json": {
+            "bytes": 3366,
+            "sha256": "873997b38981ccc03ff73e6fb84e96022f293a019843efd8c4aa50f27878ac9a"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/codec-constants-and-state.json": {
+            "bytes": 4250,
+            "sha256": "1cecd7905fef5af3ef8952a4d803900a7df91d662ebae34db47cc0dbf352d4e7"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/fixed-customvoice-prompt.json": {
+            "bytes": 3660,
+            "sha256": "ea5e8fdf54cb731b255b73aa0cf47ef53fac103baa2bd5246be45f95d54cb593"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/graph-batching-state-ownership.json": {
+            "bytes": 3950,
+            "sha256": "130741a9f7ac9d366784d6c4dd42860262235caff0c277de1e8a8f8ec933b37f"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/sglang/cuda/llm/hierarchical-prefix-cache.json": {
+            "bytes": 3641,
+            "sha256": "b50ca58df28d4343545d07adf0da23f8f181297b30aba7870e5690630e7d7315"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/tensorrt_llm/cuda/llm/prefill-decode-disaggregation.json": {
+            "bytes": 2967,
+            "sha256": "349856dd98628f94987e014889cd1b15d0e3ae6ab6c576cdde917f28fbafb102"
+          },
+          "src/inferrsi/data/knowledge_packs/peer/xdit/cuda/diffusion/parallelism-freshness-and-topology.json": {
+            "bytes": 3236,
+            "sha256": "3674b5916cb7b384951bf3e449b6796b161878daabcc1d7e7f6d6f8ff861cc17"
+          },
+          "src/inferrsi/data/repo_packs/afd_plugin/pack.json": {
+            "bytes": 502,
+            "sha256": "00042abeabdc9c0e24f1a4edaba47e233df398aef0ad4e181fc70191d383da5f"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-aclgraph-capture-replay.json": {
+            "bytes": 4311,
+            "sha256": "327414443ea9537ae320aa405912c338351bc06cefdc0b99352e1eea36f61dc3"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-device-health.json": {
+            "bytes": 4115,
+            "sha256": "633fa5fdcea1e1e9edda8a218d6a3fde0d12637848348e1e2fe7d1b0bd744a16"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-distributed-parallel.json": {
+            "bytes": 5841,
+            "sha256": "709ca0d061e4d6917fe8fa049fe93165e71dfecf96c7bf264ced25346381f603"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-graph-mode.json": {
+            "bytes": 4135,
+            "sha256": "bb82ab3d628b91be957e557545810ef5c20790be7d43296096f20a21f45ec391"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-model-adaptation.json": {
+            "bytes": 4137,
+            "sha256": "7ea1ab0f0044afff935a91071b36ac09bc46b5ef044fe2a68ab0608256ddd853"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-profiling.json": {
+            "bytes": 4186,
+            "sha256": "0d27058fd8d0ef8b524b7a60d1cb76950622746c94b5a03915e64e6edb4ab3d4"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-tuning.json": {
+            "bytes": 4439,
+            "sha256": "fcb49d71f59d8b39c1eaa03d24e95e2cbfe5f9d7f6e0f22df03ee0d60f6a568a"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-optimization-casebook.json": {
+            "bytes": 4371,
+            "sha256": "05783630bfaf3c7ef99b848865f4bbab50d0b52e80ad04fcd028c624d7638e9b"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-performance-analysis.json": {
+            "bytes": 7200,
+            "sha256": "b6eea897c12d0c6ac9d172089535c82c25eb71c09d7a4972e64812cc4682b5c9"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascendc-tiling-design.json": {
+            "bytes": 4505,
+            "sha256": "e69b6e89dd342ba55ae54b5fac14aa1b91ee2a361da40cb669c287ce519f4517"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/cann-attribution-overlap.json": {
+            "bytes": 4915,
+            "sha256": "8ac01b60f5cd37fd594d6cb5119e4e6525199d017296fc8779dc68a576f5525c"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/torchnpu-device-memory-lifecycle.json": {
+            "bytes": 4349,
+            "sha256": "7a79d0eb2c50381d40086e6db72d4bdd2cec5d2261bfb00cf73a60701a94da2a"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/torchnpu-pinned-host-memory-lifecycle.json": {
+            "bytes": 4431,
+            "sha256": "8bb34dcfbebb48667a94434b036a1a408b970af9890b047a3de325f364ed5f7c"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/assess-serving-preemption.json": {
+            "bytes": 4198,
+            "sha256": "e3d2a1a72d0d980f31fc85deee27db26bc0daa0c709e8ae99e42deb814a4a7f4"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/assess-split-backward-schedule.json": {
+            "bytes": 3915,
+            "sha256": "a8561794786eb0a897d831f00c92b4fdd3b0492316c1b1d8ad13c1f8f75ac815"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-h3-runtime-residency.json": {
+            "bytes": 12169,
+            "sha256": "602dd625874bd4a89861d5bb25881cc63e4b5a7fdedc2867089f0952943f6eff"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-llama-runtime-state.json": {
+            "bytes": 5816,
+            "sha256": "64f6ab8a95172378350323cead08bc6c201c624b711e6242938a6a21a8e83ad7"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-sd-cpp-residency.json": {
+            "bytes": 8349,
+            "sha256": "95ed97b717255be3b817e5c69debd6d41197559dbf21771507d5da79134a72e1"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-sglang-diffusion-lifetimes.json": {
+            "bytes": 7276,
+            "sha256": "5e5889223878c6cf42a49c6f09c60e9d9add88cb7e7d0df050c4e75d88a05aa6"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-vllm-kv-lifetimes.json": {
+            "bytes": 4275,
+            "sha256": "00424005a5eeb8b15f239edd2cb2c71047787e3ecf55924860cb7a78d137b586"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/bound-windowed-attention-replay.json": {
+            "bytes": 7877,
+            "sha256": "98b2291147a796ef59cf20a075ff662998ee6a529f85b49e9c6fb62fe41fad9f"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/budget-speculative-verification.json": {
+            "bytes": 7490,
+            "sha256": "66448f59460c9dffac46e74160388fab650cf8bbbd5bdc95c1465b70daf0c2dc"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/calibrate-serving-simulation.json": {
+            "bytes": 3992,
+            "sha256": "055b70e0b86f65156775b519e4b8cbf06a956a8993b06a66175167cedb78f2cb"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/cgroup-resource-controls.json": {
+            "bytes": 4801,
+            "sha256": "ad406b6dd5d9415bd28d448e64269a388b50e11f9eb7f5a9adb6b0a6246fd303"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/checkpoint-recovery-consistency.json": {
+            "bytes": 4740,
+            "sha256": "ccb85314a831bde58f8cc515580d001b174c11dca0bbcbc6221303acf629046d"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/choose-parallel-training-plan.json": {
+            "bytes": 3901,
+            "sha256": "40afcb41c59988fc1af8fdd00d2340800c54e96847dd0de16ecddf7ac221d1b4"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/cite-check-downstream-pr.json": {
+            "bytes": 2533,
+            "sha256": "dabff14684d5ae9a2b05a9743ff5b76d00a90fdb29319588aba67f5087aad706"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/claim-contract.json": {
+            "bytes": 2469,
+            "sha256": "78f15f6533434fc29adf1df651f7a15eb8ba81733f8b02c710d90102541d707b"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/comfy-ui.json": {
+            "bytes": 6919,
+            "sha256": "1496807d958df055f720ca93e6dcf7f7e8f506f3f12c94c9a2166f789b00058d"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/competitor-scan.json": {
+            "bytes": 2449,
+            "sha256": "c2e97e49e5185710cdfaa08ac1c2b1c19b958e68ebfca885d5aee0ed8c9f8c8b"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dcp-async-save-performance.json": {
+            "bytes": 4538,
+            "sha256": "359a5acf8304a78fc074688676f9a0be62f56f463d0960a544e409f49843bdaf"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dense-performance-evidence.json": {
+            "bytes": 4227,
+            "sha256": "d87f93a9dc9ddc583e3929baaac85f7a425666dd6dd71665933f18344ccec834"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/diagnose-vllm-admission-pressure.json": {
+            "bytes": 9789,
+            "sha256": "a1de40def775b2e840c27330b67e6f37c3c244db1731fab2bd3e8a787143210e"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/diffusers-invocation-state.json": {
+            "bytes": 6109,
+            "sha256": "914d2ffe9776ba931bed780a2a270c1525e3c6cbaf0735b8531064d3180f3233"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/draft-acceptance-issue.json": {
+            "bytes": 2549,
+            "sha256": "6aaa63dbfe6cda982ccd3a4e368527b80fd2873672c000c66d06d28e40a2bf14"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/evaluate-elastic-serving-placement.json": {
+            "bytes": 4334,
+            "sha256": "fab1fa814e789ec2d86587fcf7de7c38174464b9e9ba0cf53ef3fb470053e7c4"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/generation-cache-lifecycle.json": {
+            "bytes": 5972,
+            "sha256": "e5679cdecb722bbd17280913913da8473052ad09e9e7a13b3a9b95463a139ef6"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/host-memory-management.json": {
+            "bytes": 6530,
+            "sha256": "a36aea21169281866b2468860d3ffeb912e61edd571308b02f82939f17f694a8"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/host-overhead-optimization.json": {
+            "bytes": 4243,
+            "sha256": "05fd7f054966b96892980d585352bc1c3fcd47f439cb34268fdf07feee58a998"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/inference-benchmark-design.json": {
+            "bytes": 10735,
+            "sha256": "091bb086c6f0123dbb5db78968c13f846cad135e8969982efe557644b7c7613b"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/inference-optimization-casebook.json": {
+            "bytes": 6531,
+            "sha256": "c512025d7a59934c91665e24752d9393ca9b6a9dfb624c72789294defae93c0e"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/kernel-writing-guide.json": {
+            "bytes": 4545,
+            "sha256": "f1f0ca1d11d3550c9b8c5fdbed8f6cd09b877ac14341f973d58dca02b5d8995d"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/kv-tier-management.json": {
+            "bytes": 4609,
+            "sha256": "6141f61fb6b0d147b704ab7a2627e9babdb53a90e15b659152a3d565967afddf"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/lazy-evaluation-stream-lifetimes.json": {
+            "bytes": 5934,
+            "sha256": "34f47009a8a33f8907bd78f110b95fcd62409bef1b38e1055acb46b2eab285ff"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/npu-skill-factory.json": {
+            "bytes": 11868,
+            "sha256": "24c1d3bff288e86b99ea674e3252ec6f54d9ba944274aef6a7666197afd6788a"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/optimization-search-policy.json": {
+            "bytes": 7101,
+            "sha256": "d7bf71d5491765f62eec02fe5013dc1a5ef56e1516892da48593c6165a86366d"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/os-process-placement.json": {
+            "bytes": 6304,
+            "sha256": "e6e8fe5d7d66f3e98f05762d94441726dfa757ce7fbffd2b35402b3f327f8fb3"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/performance-ceiling-analysis.json": {
+            "bytes": 13695,
+            "sha256": "47c034d96026e1c2a02848d9c9c3caf4024a5033664bfd089ad30d9f23a80deb"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/plan-activation-residency.json": {
+            "bytes": 3871,
+            "sha256": "436e364020cc854fd52a10ee76e4593c7519bb0546c8aa5238fcfd89158e8ab0"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/repository-evolution-analysis.json": {
+            "bytes": 5865,
+            "sha256": "e0af402c3320ebd9bd36c774cb52c0708721a86ca446c05f6e94306888838e30"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/rsi-evidence.json": {
+            "bytes": 2475,
+            "sha256": "c8b4bb5a8c88009ac5b1526e82639715381c940d3c3eb295f0333cae6c2d3ca9"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/skill-evolution-playbook.json": {
+            "bytes": 4091,
+            "sha256": "0e063c2e7686de0c27c061e1aac0eb2d8c4fa83246e32ddeeb0e28126c176425"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/storage-io-measurement.json": {
+            "bytes": 4783,
+            "sha256": "80e589e3a0b67d3a045e03feeadc45e720ceb277e6124c917eec04863eb67419"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/training-input-pipeline.json": {
+            "bytes": 4296,
+            "sha256": "ae696bea8d2b7bfdaa29b6e5bc0bd861abcc0e42e7d9bec6ae20c6e7d6f27669"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/training-resource-budget.json": {
+            "bytes": 6944,
+            "sha256": "ea5e5e864e3aa34bd00be67c8e741877ae01480f039652073b65b71dfcd54c97"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/weight-loading-optimization.json": {
+            "bytes": 8139,
+            "sha256": "7f2d1fbc6b1cfd80c3d6ed8df107d4ca9b97655626ee078c345445a303b1b367"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/audit-afd-p2p-exchanges.json": {
+            "bytes": 6640,
+            "sha256": "9ac45aa784b430ad1ca9ffddecce8dbae9740cfe9c98c10b330e01d350015a38"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/diagnose-attention-bottlenecks.json": {
+            "bytes": 7602,
+            "sha256": "e525d75578de1cc164e6553d01efbd0a014e74646200580f59b1161c030413ad"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/host-overhead-profiling.json": {
+            "bytes": 3965,
+            "sha256": "e9e6ae8c5b9986e83cade3bd029baae112859e01aa6c96f741a264f3bee44adc"
+          },
+          "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/reason-hopper-kernels.json": {
+            "bytes": 15086,
+            "sha256": "758a649e651d8f9635ac2df45d18043e156c8d54b8f8dbc95f8e68ebcb866a28"
+          },
+          "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/parallelism-and-extension.json": {
+            "bytes": 3399,
+            "sha256": "1253f5f01bcf1498703c7f34960f0c50580df5ba8f140af557a903e296ad81d3"
+          },
+          "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/runtime-radix-hicache.json": {
+            "bytes": 3317,
+            "sha256": "5a0b2aed51649af214cdc1034f57650b7ce195c934a17452ee67a35131f80cdd"
+          },
+          "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/tuning-and-validation.json": {
+            "bytes": 3385,
+            "sha256": "ef4d4f7182da4433d59d72dda0d21b4aeafadd73260be11ab8d4cb36c20fafc6"
+          },
+          "src/inferrsi/data/repo_packs/sglang/pack.json": {
+            "bytes": 620,
+            "sha256": "bce2e0c42796a9ef499d459c37b56d1d56f4b19344841359c534fca29c0359ce"
+          },
+          "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/audio/serving-and-streaming-quality.json": {
+            "bytes": 3440,
+            "sha256": "97fc1d5a11a39a0634dae9e9602a4ec47a4d2348e121076c2d5c836cf0692114"
+          },
+          "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/omni/extension-lifecycle-validation.json": {
+            "bytes": 3482,
+            "sha256": "7da8dc5d2655099de2bcd5f0fa922d9e1bcb0bfd2f571f026a004a7f402a75b6"
+          },
+          "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/omni/pipeline-state-and-transport.json": {
+            "bytes": 3404,
+            "sha256": "0c57692772019504264e39bf0444a69e20cff86f6e112b072c53cf7a4299cc82"
+          },
+          "src/inferrsi/data/repo_packs/sglang_omni/pack.json": {
+            "bytes": 630,
+            "sha256": "8e70336bc2aac0c74732a026337813652b5d5c6e1fc60f5571e9442345077b96"
+          },
+          "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/architecture-search-space.json": {
+            "bytes": 3294,
+            "sha256": "2c5ad72d5bab4f7082fc3be1532c347444cde38f5a84639fe16b22134b68685e"
+          },
+          "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/controlled-execution-tuning.json": {
+            "bytes": 3255,
+            "sha256": "6a2bf611a432213919e1c237687507ceb0ab753cccad0e934dc77f86604805e9"
+          },
+          "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/engine-validation-contract.json": {
+            "bytes": 3295,
+            "sha256": "ef89775bc82965fe87a832c7ec8a5cf55cb9eba4ecfc6279caea9ce3b0410c41"
+          },
+          "src/inferrsi/data/repo_packs/strata/pack.json": {
+            "bytes": 617,
+            "sha256": "643b32defb8db9aa230f624856f4fb4b595b5c23db58d9453091f5420cff71c3"
+          },
+          "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/bounded-serving-measurement.json": {
+            "bytes": 3239,
+            "sha256": "fb00ddac14ef6b06f48da9ba335529bd55d5d83c2a2fe331f0aa1ca16afceb37"
+          },
+          "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/extension-correctness-gates.json": {
+            "bytes": 3343,
+            "sha256": "4ea7083d277df205ea4c361bee87a8207951596e9ef6cfa136cc7d2aeab02f46"
+          },
+          "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/runtime-and-kv-ownership.json": {
+            "bytes": 3280,
+            "sha256": "e6142c1065754716a5f26f42119722908dc8fff63e667c986d57f1227b8610bd"
+          },
+          "src/inferrsi/data/repo_packs/vllm/pack.json": {
+            "bytes": 508,
+            "sha256": "1705d96710249c4514c15d8d011b43c1a1ae7c15fbc9ddf11ef2d0e6b2ab7d07"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/minimax-h3-reference-kv-reuse.json": {
+            "bytes": 4047,
+            "sha256": "80b853fc594d387ab48bed04dca245e0a22b8cdaf7c9b853551ab3bafa302e84"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/minimax-h3-sp-head-buckets.json": {
+            "bytes": 3742,
+            "sha256": "c4d29d1bceb5e2cf86151bf53c9dafb3d83428fb1e9d2a848d852bb8824bf857"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/qwen-image-21-eager-t2i.json": {
+            "bytes": 5268,
+            "sha256": "ef107d43cc448219eda4d64ab961b61cda6350e20d30d94de1ff7c4ffff7e775"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/audio/streaming-serving-measurement.json": {
+            "bytes": 3389,
+            "sha256": "2680f32f28907f6abc7a7e1e3ecab4f46ee412f40f4898ed2f371ffadf47ae4d"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/measurement-and-profiling.json": {
+            "bytes": 4472,
+            "sha256": "c8b25583cd160024c7472a20debea34cb6d4bc79714b4bad2953555adc9f794e"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/production-integration-gates.json": {
+            "bytes": 6112,
+            "sha256": "649b49ff64d3e9e88bd620027dabf9d126ca69bbe91b448f28e845ecbe1f8e2f"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/step-state-and-output-validation.json": {
+            "bytes": 5090,
+            "sha256": "e1ffe28c4fa54af739cdef2fedf4755bba18e078f8380d4fde8138ea712e5778"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/upstream-review-gates.json": {
+            "bytes": 7614,
+            "sha256": "5127709bb85ce31008d19468e1b4fcb5454a71776f69f991841cd9ef7558f9c4"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/omni/stage-ownership-and-search.json": {
+            "bytes": 3427,
+            "sha256": "e75d6c2829ae22881893f5da2ad7edba099fd64ccfb0fc93d30473e06e9d4d80"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/audio/concurrency-scaling-state-and-overlap.json": {
+            "bytes": 25324,
+            "sha256": "cb47fabb232db81ed525cbb54371becc7e4a28b2bae9d8492f60dfb6f31563e8"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-adaln-schedule-cache.json": {
+            "bytes": 4253,
+            "sha256": "ddbba3bb72b5789938595a82d3280e861f75f6ed54f05a1d2b7dcb1089d30f63"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-cuda-mxfp8.json": {
+            "bytes": 4468,
+            "sha256": "e4c538f03f4ab90a4acc8e3d8721bbd226aec1ee9bde45757a7c27d3d970821d"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-fp32-modulation.json": {
+            "bytes": 3625,
+            "sha256": "abb8d94a1b025f0e2c8e085275e96ee85d68aeb27cb12d0918e4647f8255b693"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-qk-norm-rope.json": {
+            "bytes": 3706,
+            "sha256": "4c1c4acfb9cc0baa90bc28fed1903020ea1ea38755b84315711795db45e48bb6"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-ref2va-teacache.json": {
+            "bytes": 3835,
+            "sha256": "8aeb3f99ac4227cf8b3a5cece13d938b19082e265a557eddae27b7e495bcc22e"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-vae-stacked-tiling.json": {
+            "bytes": 3508,
+            "sha256": "8cc18eacbbfd8fc4d3a6b4152c4b584d559fc9b15377da56ef072e542d807c93"
+          },
+          "src/inferrsi/data/repo_packs/vllm_omni/pack.json": {
+            "bytes": 566,
+            "sha256": "58c971e9dcf6ac15752629ed4d9426aeb5b5d857fc37b5bcf921aa2e81c0cdd5"
+          },
+          "src/inferrsi/data/repo_packs/vllm_rlt/pack.json": {
+            "bytes": 649,
+            "sha256": "6097041b8d026740060e527f49b1b4e288315b1c3227f7583805b86d8fa415f7"
+          },
+          "src/inferrsi/data/skill_library/README.md": {
+            "bytes": 18819,
+            "sha256": "ae583f3578425df1dc91ba32b8b6543c0469c9a817ebea873a77051d6b13e1ce"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/SKILL.md": {
+            "bytes": 5494,
+            "sha256": "7581b56336d07213a87d01477793ae1a8acf0b7cf876fce36c9a761ee32c1ff4"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/metadata.json": {
+            "bytes": 1429,
+            "sha256": "68ac7f64b7ab947fc2db664bc3990f102c1cc5539b42350aa5818d0bbf17d86c"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/ownership.md": {
+            "bytes": 2081,
+            "sha256": "40688cc44183027efbf2fca8d3548c299cdf2e33fef54f76ab3549353eb9bd1b"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/source-map.md": {
+            "bytes": 4635,
+            "sha256": "10562a75683bd9bd07ec2e37a1cbd0f93f2534d6f85b55a566adb2d97485215f"
+          },
+          "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/validation.md": {
+            "bytes": 1477,
+            "sha256": "5a0b854b87d091d7d505b0bdd5c03330052507a851ab6edf35b93a6675f2d4f7"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/SKILL.md": {
+            "bytes": 4010,
+            "sha256": "a95569bcc0da6e715fd74783a6a1e0489e92b680c582f15dafbf98df589d505a"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/metadata.json": {
+            "bytes": 1519,
+            "sha256": "9c2293c8956508de987c886c7224d673d22912e7e915b026b05abdd287e97401"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/execution-lifetime.md": {
+            "bytes": 2147,
+            "sha256": "c4bc41e08aca68f661325f61e6f0db1760a079dc362f295c847bdc374fee3bf6"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/source-map.md": {
+            "bytes": 5918,
+            "sha256": "798b9cb89ed6b885eb47464ca9f137d2802a54571b9327e631d7742d3e5abd5d"
+          },
+          "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/validation.md": {
+            "bytes": 1608,
+            "sha256": "ebfc00bf7d6ab0a1dc415953ae84d56ccdc713d5614aff85e433c4a843793bff"
+          },
+          "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md": {
+            "bytes": 7322,
+            "sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845"
+          },
+          "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json": {
+            "bytes": 587,
+            "sha256": "db7ad368394ccbfdeff6a16e3423c300d51305b1f95ff5c3936df91738a8cc35"
+          },
+          "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md": {
+            "bytes": 5314,
+            "sha256": "22afe4f4d77f441096cc657d55d763473bfcbea8fb531c8e493fe49ab84955f7"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/SKILL.md": {
+            "bytes": 6745,
+            "sha256": "d494fc7e539cbb7388189bd7b3f04bb5c4983fdd7a82484348aad1cee7f741f2"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/metadata.json": {
+            "bytes": 558,
+            "sha256": "71c68431663ed2336cccc985a354adab6b13b4cfd6123a18a5d8d53c28268509"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/references/source-notes.md": {
+            "bytes": 13004,
+            "sha256": "6ffbfb68cc7e04923b4fbca2a73a7555b7c3b45a6f40819bb5f3213ef05cf88d"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/SKILL.md": {
+            "bytes": 5977,
+            "sha256": "03829039860ece35519f2e4c339a48b5014ec25e07a2c150a3efb6a3a4e35b4a"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/metadata.json": {
+            "bytes": 559,
+            "sha256": "80d80c15514919826fa72d6f6cf26b76db8accd85d21187fc239232ca454e7bf"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/SKILL.md": {
+            "bytes": 14703,
+            "sha256": "4ce9dcadd1ae3c7a8cc633336760d0a3c382f354f1b7028e3a10d28fa0d0c9fa"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/metadata.json": {
+            "bytes": 1403,
+            "sha256": "b6cb25879b1bca5f5af08b48776f2e664cf5dcd11a0efb1c4f56879e2afa9006"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/references/awesome-ascend-communication-tests.md": {
+            "bytes": 9388,
+            "sha256": "0cb27a3d24409171bb414d7ebf0c49ad62be5f46482f29624695496ba9eb3e63"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/SKILL.md": {
+            "bytes": 5839,
+            "sha256": "9203cc5a14a1d65df5fbf275734b93be023e06527459b2f79864f28a338cb44a"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/metadata.json": {
+            "bytes": 536,
+            "sha256": "72b41e7845915d34b09219624e82976386f3125816293540c2c38143fe6b477f"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/SKILL.md": {
+            "bytes": 6440,
+            "sha256": "8fdf46fb2743b67a66bc069d22a45305a222debfb606777bc451b4b8f31d9cc6"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/metadata.json": {
+            "bytes": 536,
+            "sha256": "be681a3206e5e183b4c9670c9d7784bcffce18c891449e5b15f94ca5f1324b9a"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/SKILL.md": {
+            "bytes": 12988,
+            "sha256": "401981cf7828edcab0c9ea882f7ecd83badfa3c2a95006e912cf99d67b3d6a12"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/metadata.json": {
+            "bytes": 760,
+            "sha256": "efd37bb23acb500dad4ec8ebe7ddfe5c3f196374052825f923ac85da8445f06b"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/SKILL.md": {
+            "bytes": 12236,
+            "sha256": "38b151bc6b4e9734c6fa41efb7a30343559400c0c75bbdd26bd8a968b8ea4592"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/metadata.json": {
+            "bytes": 557,
+            "sha256": "04d2002fc8a0a7becf86b0b73daaf2da8716ab466d7f988da43d41a10f26680e"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/SKILL.md": {
+            "bytes": 9172,
+            "sha256": "647f5e9a288f2c4e1617ba53df59d9d827df749eb4fe86182d21a9c41175caf8"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/metadata.json": {
+            "bytes": 536,
+            "sha256": "f425f7c1abbd616387273e8932c6bd02f4b1da012f099286ad4171e0f6037b22"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/SKILL.md": {
+            "bytes": 9880,
+            "sha256": "a057b7f86a24f34a8f84abbbb4910d58d4e50ff47060e1c404600017e3febe89"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/metadata.json": {
+            "bytes": 2297,
+            "sha256": "9e705625c8308013c287576b0247da6f6116d817a4a535cbecff80abd6ea35ee"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-CANN-2.0-LICENSE.md": {
+            "bytes": 5262,
+            "sha256": "89b1c114df62cfd183004f38d18d07f186436c3b37203dd69c3e84becf9d2282"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-stage-evidence.md": {
+            "bytes": 11789,
+            "sha256": "f56e4125325c9c1b9235a1d64fd732861c7686444492422daff44ba8dd648be4"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md": {
+            "bytes": 11860,
+            "sha256": "93aa1b775ced304cebfe6c4dec0503ddb2260ecb3cb01edc17ce329446c5ce2b"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/metadata.json": {
+            "bytes": 552,
+            "sha256": "f6d94f104cb70969024b2f4542078a240e200712c66279eca8b94f60bee1dd71"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md": {
+            "bytes": 8972,
+            "sha256": "0e6f094dc554abea12e07def50f20b233b47127779a8aefad595b4ab1efd24c2"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/metadata.json": {
+            "bytes": 569,
+            "sha256": "acccc8eb86de0df6caf78a9b3c6db76a2a686adc4a5ba0029a4d62d50ff9a8f8"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md": {
+            "bytes": 5989,
+            "sha256": "8428df7157af024fa17cfe0b3818ed54bd9f468306dda46eb8d4f0edf4f3f230"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/metadata.json": {
+            "bytes": 579,
+            "sha256": "c650dce88c07b636d68e063dbda4d319f6f7f3ecbcdd5e47f393cbfc3af7c71d"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/references/source-notes.md": {
+            "bytes": 11470,
+            "sha256": "80cba4d3a884e7ee2618326cf74b9f9c3a34e4d9cf443707f6bbe673f2d5ba65"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/SKILL.md": {
+            "bytes": 6159,
+            "sha256": "23e93ea7c98f99743509fdc04139d78ed2e7123c63e80b366c6beaba7eb70d0d"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/metadata.json": {
+            "bytes": 581,
+            "sha256": "97836a38465106d816a9ace894a7b277d6d330f9accade8f8493ce6f0fb57ef9"
+          },
+          "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/references/source-notes.md": {
+            "bytes": 13493,
+            "sha256": "435193590b4a2ba32e2e4c6fa9d7e8f889528d412cbbb48e35ec5433f7dab384"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/SKILL.md": {
+            "bytes": 7137,
+            "sha256": "a25738dbb3b9daa15c5827cb536a20fe13f7c8b9dd14aa3fb7cfa7e132879431"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/metadata.json": {
+            "bytes": 1007,
+            "sha256": "3abe626a6e4c73c943429f609ce2c62aa77aae7e1726475ebbbdfbe1794f6f7d"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/mechanisms.md": {
+            "bytes": 17376,
+            "sha256": "fe08b2292fa587be3874774ddf41063b84c579a0085df422ec6710d8aa90b8b7"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 4024,
+            "sha256": "371b01ee201ef51fecbfd6da892d517aaba295930b9395025f03f34e21c030a1"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/NEXT_GATE_LEADS.md": {
+            "bytes": 1623,
+            "sha256": "35b8b58ae5db02b2df51ccbe5cdf870a7383d17f345f9ee5eb7b0a4146dfc74b"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/README.md": {
+            "bytes": 2888,
+            "sha256": "e0b1ec1b39c10067f5ddce58320851ad089568c4b2296a24777389097cb5b388"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/io-reduction-can-plateau.md": {
+            "bytes": 2671,
+            "sha256": "5cc35f514224ac6e6efd2da6db221bba468b2f4db66807295e67b240bcdfdba5"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/overlap-and-fp8-counter-signals.md": {
+            "bytes": 3692,
+            "sha256": "80c8a60a6df55f3d84fab23e03567d8b38e23bcdbd63cc09881e46db1e2888ab"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/partitioning-after-fusion.md": {
+            "bytes": 3132,
+            "sha256": "4d4cabe0991eab554dcfbd4fa9c8199e61fa4767709b0c4cd646e2588bb398a2"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/knowledge/attention-cost-model-boundaries.md": {
+            "bytes": 5223,
+            "sha256": "5ed8f888443e38ccb72545b6025447aab7b92936f78c151fed680e582126239b"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/sources.md": {
+            "bytes": 9069,
+            "sha256": "53436e0de4e6dda2d29426591bf61f8526378cdc6f9ec7967b88b1bcd4c61e6f"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/host-overhead-profiling/SKILL.md": {
+            "bytes": 11199,
+            "sha256": "326e566fbe30d3ba04e4b50c3e87aad9df037f0ac4ad6fe129165ae97dc5f91d"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/host-overhead-profiling/metadata.json": {
+            "bytes": 527,
+            "sha256": "76e34c1c64aba473a9e126b032a776417caa0f913f6ac3cf4dba3a5d123b1ee0"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/SKILL.md": {
+            "bytes": 4905,
+            "sha256": "cec4e3aa1eb64c31c54b5f5ca40af2b61b7919d94be4174968581d5f752c0538"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/metadata.json": {
+            "bytes": 7453,
+            "sha256": "7520da1cc97412631a600957ecb18aa5a7a4461ea21e76c237330180193ed898"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/fp8.md": {
+            "bytes": 6287,
+            "sha256": "31d9c211c11c2224007abbdcd4059913c4a54833fe2d2f926411c5238fcd3501"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/cutlass-bsd.md": {
+            "bytes": 1547,
+            "sha256": "42fec630f410aa308f70a51a89fadcd19586fa620f9831a32bee528a9a10000e"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/deepgemm-mit.md": {
+            "bytes": 1065,
+            "sha256": "6c73a30c14a2938abaa6afe98ebb19e6d81dac5bd4ace0cb05c61c20afa37b85"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/kernelwiki-kda-mit.md": {
+            "bytes": 1063,
+            "sha256": "84aab57158a914f448b38d8be29dd842b634a08c192d844f416a256beed4ae94"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/notices.md": {
+            "bytes": 3240,
+            "sha256": "876ec016644bf5331b3f257985e126c0e61980e3bfd82569b7b0c3f6b2c6c64c"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/resources-and-scheduling.md": {
+            "bytes": 8919,
+            "sha256": "1ce25b426736072435a1ead9a2a8e1f56df5d0abb8198e8c4fdcb34321db4964"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/source-inventory.json": {
+            "bytes": 120506,
+            "sha256": "1dc09c895c970a0c9708158bcc6dd65cdb736274fd2c35d44542c70407d3849d"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/sources.md": {
+            "bytes": 9004,
+            "sha256": "6b74b9e774f2a80db0860510d9e6940d36284430f29c1e124f115a9eaf25833d"
+          },
+          "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/synchronization.md": {
+            "bytes": 6843,
+            "sha256": "4322f6b3c0e4368bf8c192b4ec752529767f50d19c0c38dad8fdf9cd781e41d9"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/SKILL.md": {
+            "bytes": 7348,
+            "sha256": "fd1f5a7e6a55714730044155f5b38c1d700383d4e16c3e91963bbfbe1c773af5"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/metadata.json": {
+            "bytes": 572,
+            "sha256": "19e40763fedc20990651bb989c32a7693306c5c54af46c3d4be77fd4f87b6c03"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/SKILL.md": {
+            "bytes": 23213,
+            "sha256": "2c807f3df2da2f533d25183de5771f3d412058d16410994292ef8f48c49f9b3a"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/metadata.json": {
+            "bytes": 1684,
+            "sha256": "9ac6af135f0b6d76a76bc201b6bbd8ac9afe5201dd1e2ea82c0ac4826de401da"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-LICENSE.md": {
+            "bytes": 11357,
+            "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-native-offload-readset.json": {
+            "bytes": 6267,
+            "sha256": "cc6c2011be746002ba3ff96ccefab1809863748ddd17eb102b75e7b0e134fe55"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-native-offload.md": {
+            "bytes": 8966,
+            "sha256": "6c2a72b5ba3708a01d229e0576570837a617e6a42dbb0b5312a26a65a6c31542"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/SKILL.md": {
+            "bytes": 16072,
+            "sha256": "36ef88826a32ac3147b869001004ded17a8d5a491e25bfe105005ba39ea9d1eb"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/metadata.json": {
+            "bytes": 539,
+            "sha256": "ae71c389dfdb1a19d267feecb4c6f1fbb4c9afad501ff6d8fab7e38b29bcc37a"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/references/deepinfer-zero-overhead-scheduling.md": {
+            "bytes": 10894,
+            "sha256": "f452bf40132c9195d739f1c86b3aa18857138c7ed0f1861af5853d93b9955924"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/os-process-placement/SKILL.md": {
+            "bytes": 10327,
+            "sha256": "2003492e4d89fde0443b180d74933e764bd1979435e109c8d26e20eb8f6e5053"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/os-process-placement/metadata.json": {
+            "bytes": 1399,
+            "sha256": "ec02995c075f0a540921f57b4846f2cafe334eb90f44389f17c3c50457c3e9c5"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/os-process-placement/references/llama-numa-thread-readset.md": {
+            "bytes": 6617,
+            "sha256": "748b254b3b043a5ddb0f480eb4a566854d27eedd7b111499863e9bf0a5a6951e"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md": {
+            "bytes": 8813,
+            "sha256": "63b883c10a8d3e7d918abfc7175b72e2b401ff1e02697f94b8382ec1c8f369b7"
+          },
+          "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/metadata.json": {
+            "bytes": 558,
+            "sha256": "0a41c7b00978332b54ee1e4a7c1ade88ecde3e961128f24d1b4f7708caff5ff5"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md": {
+            "bytes": 6402,
+            "sha256": "0a5f48d9352c5bf460775ef47f9136829978811f4a16e84983e299782cae305d"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-serving-preemption/metadata.json": {
+            "bytes": 537,
+            "sha256": "77ce5ba16245cbd80d46d0abdf80aaf26fc0924fa2d2a5624ff430de68a1daf0"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-serving-preemption/references/source-notes.md": {
+            "bytes": 30405,
+            "sha256": "f29fcecfba0490008a354c11b42403983fe165e56ff907ad61cc6040d2b6407f"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/SKILL.md": {
+            "bytes": 3181,
+            "sha256": "1e64f28ea71672fb53c56f6108b87864738c6d03d6dcd32eedaef9e0625e96e3"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/metadata.json": {
+            "bytes": 478,
+            "sha256": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7"
+          },
+          "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/references/source-notes.md": {
+            "bytes": 10548,
+            "sha256": "61b1609f71c37b61d136b8c4d39f793de87cffb1cf31c8ec737a1c3fa6fb1b4e"
+          },
+          "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/SKILL.md": {
+            "bytes": 3884,
+            "sha256": "ac773c6f1c110a664dfca28a3ae9129b10c07614e60ca3640dea776220e7b352"
+          },
+          "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/metadata.json": {
+            "bytes": 648,
+            "sha256": "261ce9f644ea6eebb90c830f41fb8b320d2b401456e5bd8d494095a44b8620d0"
+          },
+          "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/references/source-notes.md": {
+            "bytes": 8415,
+            "sha256": "9696d3c0ef96b6611787446bd6f96d4d6255ca8d107105de8c1d9af83edb9fa5"
+          },
+          "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/SKILL.md": {
+            "bytes": 6416,
+            "sha256": "6b7db1eb2780dec27858cb7aa0a3c093c5b5dfa4c30261aebfeaa255213ff688"
+          },
+          "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/metadata.json": {
+            "bytes": 582,
+            "sha256": "4dc8659806539b4784ea9348a01d4194ac70d0bb297a916f90ff6419efa5c328"
+          },
+          "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/references/dlrover-persistence-replicas.md": {
+            "bytes": 8398,
+            "sha256": "055f4357f79026fcbcc766bbe77db610376ec353623f8898f5d14a30ee69e7bc"
+          },
+          "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/references/sources.md": {
+            "bytes": 7984,
+            "sha256": "9425b1aab963a2c8d35f3db897bd15ee69699bf5b0d0c8425c3c8bba3e9a0233"
+          },
+          "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/SKILL.md": {
+            "bytes": 3068,
+            "sha256": "25d04aed9782b2a8ff1186828becdc7fce93c4418039ceaef9b8dcc5ec9bcd24"
+          },
+          "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/metadata.json": {
+            "bytes": 478,
+            "sha256": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7"
+          },
+          "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/references/source-notes.md": {
+            "bytes": 8521,
+            "sha256": "9f9184d46b19312138a6e805e439e43dc0594a0cc6b46675255a582dc946ca5b"
+          },
+          "src/inferrsi/data/skill_library/methods/dense-performance-evidence/SKILL.md": {
+            "bytes": 8305,
+            "sha256": "53cc1867a13f54d1006d50517dea8848869ca50df3f0cd610b2424cc8d18c933"
+          },
+          "src/inferrsi/data/skill_library/methods/dense-performance-evidence/metadata.json": {
+            "bytes": 707,
+            "sha256": "a0d1e3b17140295510a764045d1b222e19be3a3406e6d1e7dddaad49ff474490"
+          },
+          "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/SKILL.md": {
+            "bytes": 6667,
+            "sha256": "82d0099d3c33706f46c657d4fa7f76d3d486e019b36760d1de001921328a60cc"
+          },
+          "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/metadata.json": {
+            "bytes": 547,
+            "sha256": "37379a4d4a058ca46da2b20a920dc01be9ec11e557c32b2332702ec76027f4b5"
+          },
+          "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/references/source-notes.md": {
+            "bytes": 44796,
+            "sha256": "ea416831fd7d0874e06d0a76fb0c7196df53072b717673a0cdf635db6bfd917b"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-benchmark-design/SKILL.md": {
+            "bytes": 8974,
+            "sha256": "19ebc1f567aadf49f7a534a18fdb629c72c0e4ce043d360983136491cb72bfab"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-benchmark-design/metadata.json": {
+            "bytes": 5923,
+            "sha256": "18310402a8ade03fd48ac687f79ba9d447871e90337de84c0e1bc15c0b827306"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-benchmark-design/references/benchmark-commands.md": {
+            "bytes": 8466,
+            "sha256": "0898424e0bff3a6246e1eeeda6abd6c5bd974c047fe32f3bbd5ab9d31ce17cd3"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-benchmark-design/references/task-benchmark-selection.md": {
+            "bytes": 25086,
+            "sha256": "0d4bb080a754e26846e81680023db0dff5e6a4c96be4327c411c3b72ac4ef8e2"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/SKILL.md": {
+            "bytes": 6644,
+            "sha256": "89a6a82c692f7f5192f4d7f4c3a5ff08a4df3b1c4bdaa0aafb7e04292d8aa5b4"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/admission-drain-occupancy.json": {
+            "bytes": 1354,
+            "sha256": "3319e23bbe2bd40e2d45fb89dc87bdc92d4df2b5ddd8086fcf84813b43885b7c"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/admission-drain-occupancy.md": {
+            "bytes": 3134,
+            "sha256": "e88bbac5d253a8e7e3a3d5e21047f23f2970c11ac23887d4303ab94fca52dd27"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/drama3-rejected-directions.json": {
+            "bytes": 1459,
+            "sha256": "2cc0a526fe0509551607a03e7626dd3d7844e643c4afb194837f75cddead8b12"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/drama3-rejected-directions.md": {
+            "bytes": 3557,
+            "sha256": "6f143bc363a3290fc008c08e57cb075f1776d1c8cca8a6a59297d5b1fbbf61bf"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-ascend-rejected-directions.json": {
+            "bytes": 1293,
+            "sha256": "b3d538f79c56724ccb2895cbf50cf2e6b0fec3979be62a684a8230cd932dff55"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-ascend-rejected-directions.md": {
+            "bytes": 2564,
+            "sha256": "be396983812262515899a3e2ba4f01b23ff92e1b836881882c7580372b8ae552"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-b300-rejected-directions.json": {
+            "bytes": 1196,
+            "sha256": "146ea1564a46ed03092c7dc5796b318fc05b04b1821105ee00eb7fa7fa1333e2"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-b300-rejected-directions.md": {
+            "bytes": 2656,
+            "sha256": "cce2930f544a0f78f21797fba4d8ded298e4a81f76c34d7977b0e31d361bc738"
+          },
+          "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/metadata.json": {
+            "bytes": 535,
+            "sha256": "fb55067fb45970319787068a3206110641cceaf0ccbabd706be6d0b3d2afc2ed"
+          },
+          "src/inferrsi/data/skill_library/methods/kernel-writing-guide/SKILL.md": {
+            "bytes": 10029,
+            "sha256": "01fb1712d7eb988c5e412f02340d2e48b4f8564fcf4ab232dc5897869d0d50a0"
+          },
+          "src/inferrsi/data/skill_library/methods/kernel-writing-guide/metadata.json": {
+            "bytes": 653,
+            "sha256": "0189d184337662056fc6952c7beffc1bf6154c4fa173ae3e3e463bd22d70f36c"
+          },
+          "src/inferrsi/data/skill_library/methods/kernel-writing-guide/references/akg-affine-tail-guards.md": {
+            "bytes": 11941,
+            "sha256": "113deb2db10b171db3425bb3f1792bee763706b794c8cc2e222b4cc0d0114118"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/SKILL.md": {
+            "bytes": 12914,
+            "sha256": "882e7850b9b49ad692a2d11618701afe0f4e83727ea78df114be2029923c725a"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/metadata.json": {
+            "bytes": 1403,
+            "sha256": "9f1fd8a2af52bc964021e21f16ecbdb891e08f2057370e700d67f86605d333db"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 3597,
+            "sha256": "157e74ba180426b279b548a729769f152bf18f9a78ffca4ebfcaf26b7d0c5292"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/MECHANISMS.md": {
+            "bytes": 5875,
+            "sha256": "18bb2f5d7e52773c516420a142edfb531d0b9c822a2934d60f3234ac1cb9d40e"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/NEXT_GATE_LEADS.md": {
+            "bytes": 941,
+            "sha256": "16a264c7159f9c2c700ce8ba46e63f3e636b277c987da4970f6d2fa085fcb053"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/knowledge/adaptation-interfaces.md": {
+            "bytes": 3893,
+            "sha256": "5e761957cb5ecd93926e6ffc13deee7c830c6e8f0ee3c3846e9dd3b50dc74446"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/references/source-notes.md": {
+            "bytes": 13370,
+            "sha256": "54d83c3713045d59a48821bcf93598374ba4a630e3891ad39498f919317c4fb8"
+          },
+          "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/updates/optimization-search-policy.md": {
+            "bytes": 4238,
+            "sha256": "d3db44fa9a6831b86a02c638b323ed943e69fe256866213f10d68c72d762879e"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/SKILL.md": {
+            "bytes": 15324,
+            "sha256": "c5f78961c23e3c8d659f5675b522518ab06e1f2a8ea8bc74c5d428ffb41693d5"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/metadata.json": {
+            "bytes": 3669,
+            "sha256": "60b5b24ead590b0fb222398fba47822bf6723690da683e256a450bb5c0eeb806"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 7349,
+            "sha256": "083e462e0a204dce599362d46712fcfb392d053c5cda56216e22e28d4b8fa0fa"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/MECHANISMS.md": {
+            "bytes": 4003,
+            "sha256": "742c249082ceca230326f4407a40130fe624553b796ae5a157a8a4c15a7e1398"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/README.md": {
+            "bytes": 1204,
+            "sha256": "17b2d1c7c14124d59ac8a0d5ad4200ad2803d991e784a1a98cd97188cc364999"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/READSET.md": {
+            "bytes": 3937,
+            "sha256": "66459ce393858b2b2160f1e2973373bfc4b9902a12644631934b4ebe3140296a"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/STATIC_CHECKS.md": {
+            "bytes": 1111,
+            "sha256": "cb0aa7d148352ebb7b1678fc42a557a099575054716a341dfa25b6a998029064"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/cases/decision-cases.md": {
+            "bytes": 5491,
+            "sha256": "781a675a6533e0a18c044aaaee0d010da7013b6504654b8b8292a35176db801e"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/govern-optimization-experiments/procedure.md": {
+            "bytes": 3823,
+            "sha256": "ce036a4d8a0908da204515c47e2ff9623548d09de264fa65bd58fedce52a78a0"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/govern-optimization-experiments/references/source-notes.md": {
+            "bytes": 21659,
+            "sha256": "f9a8254e45473ddb2b73e576957b6d3db83c62b99dff1922f4d1e606e569559e"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/knowledge/bounds-predictions-and-decisions.md": {
+            "bytes": 5297,
+            "sha256": "5780d6eb5257612ec45deb94f85f334a6ecabf7e466d1678706041f6549bda79"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/redirect-cross-layer-optimization/procedure.md": {
+            "bytes": 3648,
+            "sha256": "6593a63fce9c474b515663a560c94b7df048dc5f6e77a95b8e54e4a64cd10a09"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/redirect-cross-layer-optimization/references/source-notes.md": {
+            "bytes": 26469,
+            "sha256": "6ff3a7bc31f1172db92aedaf7df1f1aac688bf769227b583a4cd82bfad52da67"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/agentic-kernel.md": {
+            "bytes": 41353,
+            "sha256": "7190573b85f94a4391bb0d887ccbd37a31b3f08bb0531114ad7afd72fe1567a7"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/flashvector.md": {
+            "bytes": 9830,
+            "sha256": "3926369acd88ff3c370bee5718e74dd15f3219c3e4f6311095e30e917bffa4cf"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/inferencebench.md": {
+            "bytes": 9294,
+            "sha256": "21f47111197a0b7f8d6384ea918d139dc8318a62ca3313ce060048271f50b639"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/prior-coverage.md": {
+            "bytes": 3254,
+            "sha256": "35d47ac078591221fcb666219bcd824f9400fc17aafecdc0499d8336f7178238"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/vidur.md": {
+            "bytes": 7736,
+            "sha256": "d38602af5219f5feeb49e6cd6bf53d66098c77fad141dae676f1d8a6743cf25d"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/roofline-contract.md": {
+            "bytes": 7091,
+            "sha256": "419ec4bcaec45ac579a133bce083d4f48779f26e664d7a86f720fa65f343b1d1"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/roofline-example.json": {
+            "bytes": 2015,
+            "sha256": "597cd3205df043cff3ae9d15924c77934e8d8eb300c00c4840f6bb2921f90faf"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/scripts/roofline.py": {
+            "bytes": 9004,
+            "sha256": "524de48a6764401695fe1444f1fa3d38f363069b0cd2fba4715a86ca4b28a5d4"
+          },
+          "src/inferrsi/data/skill_library/methods/plan-activation-residency/SKILL.md": {
+            "bytes": 3507,
+            "sha256": "542d8a1e6963f5019986e1fbd79d3a247a045a51487c09313026087ac457c15e"
+          },
+          "src/inferrsi/data/skill_library/methods/plan-activation-residency/metadata.json": {
+            "bytes": 478,
+            "sha256": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7"
+          },
+          "src/inferrsi/data/skill_library/methods/plan-activation-residency/references/source-notes.md": {
+            "bytes": 13907,
+            "sha256": "5ff52655cfd06fcf970ed2a7d4d4fe1c31115b102af2d41497de0c74848ed2b3"
+          },
+          "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/SKILL.md": {
+            "bytes": 13800,
+            "sha256": "7e20d3390f2501263e5e298e026c47cac16c234f2397833131e15ae5d34c0b97"
+          },
+          "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/metadata.json": {
+            "bytes": 1903,
+            "sha256": "d0e4d6950250e4cec43d82e23324c9301a925d4c22be30b4d2713d790cb6504b"
+          },
+          "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/references/capability-presence-audit.md": {
+            "bytes": 11461,
+            "sha256": "c21eb4a6aabe811a49a6540a34c54dbcfdcf7efae3052a79b54034c7838026a5"
+          },
+          "src/inferrsi/data/skill_library/methods/skill-evolution-playbook/SKILL.md": {
+            "bytes": 9448,
+            "sha256": "9cd37f21fe0d862069e4798edca8a04d59f8b840e55d521901594a63dddd9ba3"
+          },
+          "src/inferrsi/data/skill_library/methods/skill-evolution-playbook/metadata.json": {
+            "bytes": 527,
+            "sha256": "92b36338b9257e2e41b95e4370694d7de52ed84bc7971056b3bfced07b0993d8"
+          },
+          "src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md": {
+            "bytes": 6865,
+            "sha256": "64bd2c6b795d966045d1d3cca55976d3d49b08b06df7c7e06a70c2824cf55b7e"
+          },
+          "src/inferrsi/data/skill_library/methods/training-input-pipeline/metadata.json": {
+            "bytes": 587,
+            "sha256": "80659099fce6a6ec05e9084b75452c877a5812e30f60a78f31ef8c7efcdc1bac"
+          },
+          "src/inferrsi/data/skill_library/methods/training-input-pipeline/references/sources.md": {
+            "bytes": 2637,
+            "sha256": "60e0f3e88655d8e1d5eb4797f39ca4370deda01614f265b8d2ea4559d8ac020f"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/SKILL.md": {
+            "bytes": 7200,
+            "sha256": "bdfa5a4ef37d9f303c3a23cd6f98c21137b97ac5ed58e0ed5fac6d71d33e52ef"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/metadata.json": {
+            "bytes": 976,
+            "sha256": "d42b614af9cee6ee0d4a25dc5d57c7b98438feed1e807a7c9f1755109a86e9cf"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/mindspeed-historical.md": {
+            "bytes": 3384,
+            "sha256": "eb97f935ed30517a8fc9edec5e9cddc04f5d679d156c16d75bd29c534593360a"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 7368,
+            "sha256": "c55c6ad2131ef788866411dcc7056f221420ebff30c95c6bc919b09a4818cb73"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/cases/paper-grounded-cases.md": {
+            "bytes": 5649,
+            "sha256": "19bd25b1a944adb2351ebc6b1d795150ea033041a44f6cc261e881677812a101"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/knowledge/offload-delta.md": {
+            "bytes": 2026,
+            "sha256": "f8740fc5c8bbbe5ac13ff1cfec7ca499569d0b46e8cd56cf1f5a44c3a5e72242"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/notes/training-memory-scheduling.md": {
+            "bytes": 18660,
+            "sha256": "9aa6d3e01452dff95c8a8940ff62ead8c146ce215c9175b78c9f2ef508a7b2de"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/references/source-readset.md": {
+            "bytes": 22946,
+            "sha256": "2a5c166767f56fa47e7eba7386f1f36da8cb73702e5b5bbe14d6a456ca9ebfae"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/sources.md": {
+            "bytes": 2752,
+            "sha256": "6d248f58af8b350fc66c22898988d6f451ac60d03bcd805cd6e84df5e55174ac"
+          },
+          "src/inferrsi/data/skill_library/methods/training-resource-budget/references/stas-decision-notes.md": {
+            "bytes": 4086,
+            "sha256": "f1a993e6c4ca9a035a2effcaf5500081f227b0cc56cc885a6b22fda0041fc744"
+          },
+          "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md": {
+            "bytes": 25969,
+            "sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213"
+          },
+          "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json": {
+            "bytes": 2418,
+            "sha256": "798cbff6f10b7c24b188080ebf233b188b0e98ef2c6fdb5e00e1fe3eff727c0f"
+          },
+          "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md": {
+            "bytes": 8656,
+            "sha256": "5a39f14b102b0cee304541e011aa31e44e6b13527d1bee9987c3f3b66c064d3d"
+          },
+          "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md": {
+            "bytes": 6392,
+            "sha256": "6c640f28ee3f0253a17b81461c865cb9f947367002cb75bf72550c4572d02e97"
+          },
+          "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/SKILL.md": {
+            "bytes": 7181,
+            "sha256": "a46110891eca860915e314d6921a22856d631453b22b17837fe30262d6780535"
+          },
+          "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/metadata.json": {
+            "bytes": 2175,
+            "sha256": "4e6100dfb9fc7bc00b19915f4fd38ad509704f2a609c658adf4d632d4e05b6c1"
+          },
+          "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/references/afd-plugin-LICENSE.md": {
+            "bytes": 12747,
+            "sha256": "0e8566ea362276878e1f38f97da49964dbea301c15d4aa9d727f97a4c39fd943"
+          },
+          "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/references/source-notes.md": {
+            "bytes": 17240,
+            "sha256": "5073d87450cb60d1b245308dff8807e73749c07de976d50f2fa26694fad81a72"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/LICENSE": {
+            "bytes": 35149,
+            "sha256": "3972dc9744f6499f0f9b2dbf76696f2ae7ad8af9b23dde66d6af86c9dfb36986"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/SKILL.md": {
+            "bytes": 4507,
+            "sha256": "c993bd6ff8e9b5f82f39f175ec8242341136e4ee0808af414e1d94ad3e196c5e"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/metadata.json": {
+            "bytes": 2032,
+            "sha256": "74742851a002a51c208362cdcd2a59fdab5d1734d8492b1f93fbec3cf21e0841"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/references/graph-cache-invalidation.md": {
+            "bytes": 7679,
+            "sha256": "614c9a4e6943d24d7d7be7d632682ce8b1dcc239124a37d9dc1d487113772eac"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/references/model-residency.md": {
+            "bytes": 7191,
+            "sha256": "014c68386e7cdfdb7b4f566836e8b2ff61bd5a5a506e7b4723f83fe80c5d4de7"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/references/repo-provenance.md": {
+            "bytes": 5648,
+            "sha256": "ca996ef515af42a3ce304747fa3f12a31c00533deaace5ba2ce21aefa0a62e2e"
+          },
+          "src/inferrsi/data/skill_library/repositories/comfy-ui/references/validation-boundaries.md": {
+            "bytes": 3523,
+            "sha256": "1913dedbeb558d872d6a0e015cc11d822910c3cb259a236db29fb60ad8a64263"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md": {
+            "bytes": 12294,
+            "sha256": "4703beb6e5194ef2ff2c0b5ec69f64e35417a4bc0dfb50a10c80a920b77ef3b5"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/metadata.json": {
+            "bytes": 2333,
+            "sha256": "014be24d0e669f4acbc0717dc323eec784026cf185186619d5a19d6d84496131"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/source-notes.md": {
+            "bytes": 12036,
+            "sha256": "fd917bd5ac80bee9c9f4525f8e0300f6d67dc144e2a16632af42a67f3d53d098"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/validation.md": {
+            "bytes": 5402,
+            "sha256": "60e71e70c5519bcdf6c171eec2be2f342e8801bedd746d6005f11adc646dfaff"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/SKILL.md": {
+            "bytes": 11202,
+            "sha256": "f157562d1edc905cb4fc07246567e3c5b844efb9fbf5fe39f0d40553a1b1a307"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/metadata.json": {
+            "bytes": 2156,
+            "sha256": "4d68bf4c1767eaee12ade963bbd551d0e61d1bcd81a984bb02fd3aae46965919"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/references/source-notes.md": {
+            "bytes": 11633,
+            "sha256": "6fa955022f16cd598f44fd6521901d1cb868b06cec545cabe2b82769a641b5d6"
+          },
+          "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/references/validation.md": {
+            "bytes": 7196,
+            "sha256": "673d93f6700052974b340fab25b3438e0e6344960569ac9abec7d7720f662288"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/LICENSE": {
+            "bytes": 1077,
+            "sha256": "663e51edcd9a3a8f858018ce6924ad2895a24c269c26c39ad75325bcd0f7e5bc"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/SKILL.md": {
+            "bytes": 5432,
+            "sha256": "fdb4d9483c6efa836f9e59617da605fa9c2186e11c2041b05d5c81850ba0475d"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/metadata.json": {
+            "bytes": 4923,
+            "sha256": "fe626a77af3f084f36fd73ffdb6f67976e7dceeafb94ba22078ef03c3a96cfae"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/THIRD_PARTY_NOTICES.md": {
+            "bytes": 1774,
+            "sha256": "a0d77db7d2aca0eda811ab3c386cc32b3143f2c2609a042cf4642954f0f6fb09"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/minimax-h3-apple.md": {
+            "bytes": 12059,
+            "sha256": "57ce01754ba96c378f13a91305ef3d418b926f99b2753f0af99315572bce063a"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/minimax-h3-weight-inventory.json": {
+            "bytes": 10664,
+            "sha256": "fbaacaeec340d3ce893556ee864405a0746af4d13530a38cb3d5aa9d60929271"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/phase-residency.md": {
+            "bytes": 9408,
+            "sha256": "2836a550eb2be7f5a862f51ec7b135ebd9715bf3c4faa84a7ecbc9d9c11f5d8d"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/precision-and-streaming.md": {
+            "bytes": 9440,
+            "sha256": "7e166c89193f4c7aed852ec95346a2de2b5008dac002e2e1862be1b3e57f6093"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/source-identity.json": {
+            "bytes": 7747,
+            "sha256": "730f8dd23648a3f8f47b9726cc613d893a79d1817d515eec609a9b06d8a9ff90"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/source-notes.md": {
+            "bytes": 9448,
+            "sha256": "7304bf9369efc157dc7aa40686dd828cc14db45c49442e908b3d3cc4b329eb86"
+          },
+          "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/validation.md": {
+            "bytes": 7880,
+            "sha256": "0754873a758e048480938314894e25b504ae0c269549312e7f8e6020d83fe294"
+          },
+          "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/SKILL.md": {
+            "bytes": 9338,
+            "sha256": "95e81b6bc096a954fa73139ea5bc001d1efa6c99747161da62f012b494e5c013"
+          },
+          "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/metadata.json": {
+            "bytes": 1718,
+            "sha256": "1d30b31dc8352fc6a8d7b8abbef8118c840e47e5b90795dae0c758bf9717278c"
+          },
+          "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/references/source-notes.md": {
+            "bytes": 15234,
+            "sha256": "3d12a401e9decb55a2f4f3b2bfa0d360ef5110f63a0469dfe285654e78054190"
+          },
+          "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md": {
+            "bytes": 10812,
+            "sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27"
+          },
+          "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/metadata.json": {
+            "bytes": 572,
+            "sha256": "c221189b8d85fac3bd41ef1c7cd560f1aebc8cfa0381c182d1d571a7169be829"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/SKILL.md": {
+            "bytes": 3788,
+            "sha256": "bf1177cb9985c12a9bf11edb64771018e1e314c4713920de8fd4a989ff70115f"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/metadata.json": {
+            "bytes": 1299,
+            "sha256": "d86cecb615e93ba0f6be09d1ead102cd5a17068c07e1be2549ac5cd8f7519279"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/cache-stop.md": {
+            "bytes": 2026,
+            "sha256": "bba6b5c73261065309cbcf11b1e591eef48916f0c79ceaaf74d98a9fa96b0992"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/generation-batch.md": {
+            "bytes": 1733,
+            "sha256": "f16d3c57d45e63bcaffd3871aa41c76972ee11f7dbf1e12b45d9b170303cad9b"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/source-map.md": {
+            "bytes": 6358,
+            "sha256": "91d26181f246307f2113f94b12ac9587d689b654dcd216e9f14b11cf5ebb650c"
+          },
+          "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/validation.md": {
+            "bytes": 1665,
+            "sha256": "73b5dd347017baba519ecc85415aead9993c1751d116b9ba4f07689f58c5f783"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/LICENSE": {
+            "bytes": 11352,
+            "sha256": "a54c8553df5d4313d293d1aef3bf2f13952bda8b7c7a1fdd729b2bbe27ee6015"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/SKILL.md": {
+            "bytes": 2304,
+            "sha256": "d75f5b2881b240ec2d9d4f5cc8e3f7d53e4378cb01732a2fb9699da2901ba38c"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/references/repo-provenance.md": {
+            "bytes": 1778,
+            "sha256": "5e8dbe168a8885e209d4b4532944ffa7fead658b23fd6588054d57b22476f63b"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/references/repo-routing-metadata.json": {
+            "bytes": 511,
+            "sha256": "6820707cb9bd877d1635928cf5a0efe38e4fe37dc04036dc70d53e479ed9bf7c"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/references/snapshot.md": {
+            "bytes": 3488,
+            "sha256": "0fb5d776349de845dc314f16ce73a06c9caf5db84dc1c293327d3b3ca1337c80"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-architecture/SKILL.md": {
+            "bytes": 2355,
+            "sha256": "766b2d66d2056dd779abd9e492cb3af6e25899298f44964f565509f130a7cb24"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-architecture/references/runtime.md": {
+            "bytes": 7734,
+            "sha256": "36bba9874fe054acc5217f9c265f5ae309c83943e0d46228370ae1bcdc69421e"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-extension/SKILL.md": {
+            "bytes": 2193,
+            "sha256": "9af38bca840d0004fc2ed049141476fc40fd40dcfa843f38a404a7a314c19425"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-extension/references/integration.md": {
+            "bytes": 5775,
+            "sha256": "6ce9383ca17bacff60a180f8fb8c43ea47eaa37117fb8da6051ad86135d228ef"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-serving/SKILL.md": {
+            "bytes": 2335,
+            "sha256": "990b6bdebfe22612de49dbebcb16810e8abef94d70691aa883b7937c85210eb6"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-serving/references/recipes-and-tuning.md": {
+            "bytes": 8918,
+            "sha256": "94afb762bbe185f58606154b3f0a5af56e7089cb84ea94e26cfcb9f93c95439f"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/LICENSE": {
+            "bytes": 11346,
+            "sha256": "1495e1e757ef4d0925a2350563cf5754bb23c51701a8ec4fb3c5cdcbedae6747"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/SKILL.md": {
+            "bytes": 5466,
+            "sha256": "ae72d1f0cc750699bf79d7a138a97219209bca7b7e5f19cd427d306245eaf841"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/LICENSE": {
+            "bytes": 11346,
+            "sha256": "1495e1e757ef4d0925a2350563cf5754bb23c51701a8ec4fb3c5cdcbedae6747"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/SKILL.md": {
+            "bytes": 5738,
+            "sha256": "b0a4ba4751087126143ef86983798deb0ed00e24b6ff06bf6ef76a9df6e274f2"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/metadata.json": {
+            "bytes": 2844,
+            "sha256": "6ed57dfb4f25ecb34b81f2188b055e05d172ba2581e6a3b605b814b2f733a4d6"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/references/source-notes.md": {
+            "bytes": 19512,
+            "sha256": "d0fcfd1177ae2f351041c5b2abba7f5121cec133006d9e4dca8f74ddb567332d"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/architecture-and-memory.md": {
+            "bytes": 7735,
+            "sha256": "798e747022b7ec135cc4861c2c9ac44fb14b503bb77a0068ee7fbea494b764a7"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/native-candidates.md": {
+            "bytes": 3961,
+            "sha256": "abafcc9a34df196a6365145554887b8d37afb36c6e236c92e50a4a06c8a4ec93"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/parallelism-and-extension.md": {
+            "bytes": 6324,
+            "sha256": "e3be7f4b51264ac68682a64d264e4d96857d5bb9f9d99cae5f327c56984a9e6b"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/repo-provenance.md": {
+            "bytes": 2274,
+            "sha256": "534d95ae8d0b29bcbc8f6707b3d38fb964e31f5b66ba13517a9a6b8f309bbb38"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/repo-routing-metadata.json": {
+            "bytes": 625,
+            "sha256": "b2b879c1cc42ddc2a5b95ae24115eb891562fa9a6e96d151f4ddf6f7719fe73c"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/source-map.md": {
+            "bytes": 8945,
+            "sha256": "c2adad64ceb0c305cdc4859c03012b3e0a4ab26cd0cb289bbd76e192b0cc3794"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/references/tuning-and-validation.md": {
+            "bytes": 10832,
+            "sha256": "49b3343b3292cc098a787de9d24d968b444d0ef4f73fedb5d8183390e8f39c9c"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-parallel-extension/SKILL.md": {
+            "bytes": 2316,
+            "sha256": "18bd54ebdd629e1ba6dfa589f391c7cc3c5355785c2ae82f741b64ee8022464f"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-runtime-memory/SKILL.md": {
+            "bytes": 2154,
+            "sha256": "33b1a65aba00fca819d2b6cc42202e7708e9155e4d8f4373a2387f4ee2f2fa3e"
+          },
+          "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-tuning-validation/SKILL.md": {
+            "bytes": 2018,
+            "sha256": "744b2ef9122101d3f50cbbb40fe020e2b3d3888adbe2d86041041e7c7ce424fc"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE": {
+            "bytes": 1062,
+            "sha256": "b53fa08f515cb5a6fff7b9fd8fcd0961a4b80df29d74372d25e1f9171aa042ee"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md": {
+            "bytes": 5504,
+            "sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json": {
+            "bytes": 2202,
+            "sha256": "f7b7581a3f44ba9774480c559408fc66d6a4d6d1dbbeabe4529f1eb7d4dd5ad6"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md": {
+            "bytes": 8136,
+            "sha256": "7b537cb972347dd68e303960f142892346d8b19b884e4d08016ea36a23bdaa7a"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json": {
+            "bytes": 3617,
+            "sha256": "af4830370b6eddc2d176a9ae14da4ab56a90b0fe048a01a89ee051df4365d8e9"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md": {
+            "bytes": 6589,
+            "sha256": "b67affbd7d237d82b7f51be49f1537e9820fe56ec70a71ee76809f9e599ca693"
+          },
+          "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md": {
+            "bytes": 3394,
+            "sha256": "5dcf5fa7bf734fbe13214f57d2dfb40ff8215d9cd34e072a9e615bf103b82fc2"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/LICENSE": {
+            "bytes": 1093,
+            "sha256": "949b9b0aada7476006469e58ee4ffb0e0f68c2d9137e90a356757b61a09d37f0"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/SKILL.md": {
+            "bytes": 2813,
+            "sha256": "0e5daeb6409c916be45ecd7ce2531ea5f1c60a4a53e2b0ed25dc1e2cdc5c17ed"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/design-space-analysis.md": {
+            "bytes": 27736,
+            "sha256": "11ecafe71530ddd4839268ea1b24b3361a3895012a705859ca06dae29fa72630"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/repo-provenance.md": {
+            "bytes": 2913,
+            "sha256": "05f17d0ae996ad6c12314bdd29f3427574f6aa611ea60effd18598a0f386638b"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/repo-routing-metadata.json": {
+            "bytes": 332,
+            "sha256": "a10ba49b7e6565109b566c105e17551d042075a1ff649c6f371134e3b7402c74"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/trial-format.md": {
+            "bytes": 2291,
+            "sha256": "92d8cf38549268ae4eef382890662d03e18a7dff858ad45ee5eb6f47eb39c8c0"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/references/troubleshooting.md": {
+            "bytes": 2079,
+            "sha256": "7301cfa5a5f6cfbf9576c3845a96c6b4de8f8a6509f2330970a8e2e4669b3d5a"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/scripts/summarize_trials.py": {
+            "bytes": 5170,
+            "sha256": "6a243a08da15659ced6519f49f52a49a675bd9e952740c3ab03dd21964eefbe2"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/design-space/SKILL.md": {
+            "bytes": 1826,
+            "sha256": "857bea3c64db45dc035728a41523142ac65778b41f63092148b8ab936f971d76"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/design-space/references/design-space.md": {
+            "bytes": 24231,
+            "sha256": "0a428dcd12d5290b554e9da6795a110f94d98f128771602e1ea5f608048b135e"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/tune-execution/SKILL.md": {
+            "bytes": 1145,
+            "sha256": "cdf51c537e7237296d52ef6cda9f2e4c9d9ce3fc87265a369cd9eb7497c53512"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/tune-execution/references/tuning.md": {
+            "bytes": 11310,
+            "sha256": "7b69aa7872d60dd3fb91e41b77a632572070dcf80dd5b5f6dc1722d8540177b1"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/validate-engine/SKILL.md": {
+            "bytes": 4613,
+            "sha256": "90bfd975e25f18895ce0e85f6f35d24149ce026305a4d479aea4924510053460"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/sub-skills/validate-engine/references/validation.md": {
+            "bytes": 16895,
+            "sha256": "dae846f4c47c68731b96ec1d50cb008c03871c72078f1c5b0f4a9423be9ba397"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/LICENSE": {
+            "bytes": 11357,
+            "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/SKILL.md": {
+            "bytes": 4945,
+            "sha256": "df3e63dc28de2dac2aa92824fd88719159138650d96487332e0effca3f0e2a80"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/references/evidence.md": {
+            "bytes": 4689,
+            "sha256": "ae0d51ba56b34ed81fba34de8fd3685bebf2f160b188fa53719936002881d220"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/references/repo-provenance.md": {
+            "bytes": 2221,
+            "sha256": "43bbfe431434ca4ad6157ba35cbb9ce2d89e4065f427abfdd5f372f51ba0c9c9"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/references/repo-routing-metadata.json": {
+            "bytes": 342,
+            "sha256": "7784b4d98d096718c4f4cc42ce86d5ec113c469844e12de4690b0fccca59d329"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-architecture/SKILL.md": {
+            "bytes": 1336,
+            "sha256": "fb3ff50a7b9122a2545a97ef5179011237c4c4ca214dde7236ad7bf79a7c8623"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-architecture/references/architecture.md": {
+            "bytes": 5600,
+            "sha256": "db845eb4207ddeedb8ae381f1b91704a42cb53ba0ee6c7ee581510ba1a412190"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-serving/SKILL.md": {
+            "bytes": 1343,
+            "sha256": "8ab3209b838635395b22deca6aa879172016531d45c2374eb8e498e7f71bb40f"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-serving/references/serving.md": {
+            "bytes": 5049,
+            "sha256": "9bd5eff300d9755137a7600c8f7a7753bb22d09c8853e08062d7e535662ba8a3"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/SKILL.md": {
+            "bytes": 2227,
+            "sha256": "93c83bc4b799b32f50ff54b870cbbde050a790af16d9b7b062039c4987c80148"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/references/abort-acknowledgment.md": {
+            "bytes": 8547,
+            "sha256": "e95ae5b2cc410917ddf5e9066890d0bef9264104529bdca5ac12bdf56fe80ae3"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/references/validation.md": {
+            "bytes": 4632,
+            "sha256": "1542061f9c76fff2acb6529a58238301484b0e23ec0c697d2fdf129e200ea973"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/LICENSE": {
+            "bytes": 11324,
+            "sha256": "50e6751797c50dedd75ef1b8a0d9e42f5f8472e9fbce91f34718e9f97b0c780a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/SKILL.md": {
+            "bytes": 2863,
+            "sha256": "38e522b8abeb0dc140a2194340432ed41dc8b569d33b63d5d4c59aa1092cd2d1"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/SKILL.md": {
+            "bytes": 6664,
+            "sha256": "313733a65cc48442b18dda1c4665139a689963a90e1032a2235bc6f40e3a8e06"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/metadata.json": {
+            "bytes": 781,
+            "sha256": "b43fc3a7ccae59c664fcdf9ff77962edcc07c2a3df8b59324db551642bd23e23"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/references/source-notes.md": {
+            "bytes": 13853,
+            "sha256": "3665b034594fc8cc767c408002f4e16f8803369242405027c7f37ce6c9ecebe9"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/SKILL.md": {
+            "bytes": 6565,
+            "sha256": "5fb9ea1d1373e110d9b506f69963c4857338b7dbf5e38ec9a49b59a0ff683ed0"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/metadata.json": {
+            "bytes": 1372,
+            "sha256": "78459416734f61a8faa819d9a1a3ad7b9a985e8bf8e149813c604d15be6bb92f"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 9006,
+            "sha256": "82b24be82451ee121892f61575d0cf4ecb9920813a25ccb26cce2b45dda6d8b3"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/migration-is-not-free-memory.md": {
+            "bytes": 2145,
+            "sha256": "2ac168f2725cba7fc92dfcc67c01bf2258dff6a2dbe1ddb063bc60c175d82d0a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/parallelism-change-with-state-cost.md": {
+            "bytes": 2069,
+            "sha256": "5accc81086e4500d66f894ef6af9951e1cbb76941f6c631aab5c5561a786d8d8"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/preemption-under-kv-pressure.md": {
+            "bytes": 2630,
+            "sha256": "462200a6d0b193ef01f32374a463e0f2838ca8edf33d9ab898def65ffd0b490a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/knowledge/scheduling-state-boundaries.md": {
+            "bytes": 7488,
+            "sha256": "d4ef29869959b835a5dcb06ddb2fd8e6cff49bb176994dc1d46eb9a852f6228d"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/fastserve.md": {
+            "bytes": 9289,
+            "sha256": "9b4e275843000ef453d508d9c3c9c3f8ea1de5545fbd01d7abe74baa8a4931a7"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/llumnix.md": {
+            "bytes": 10181,
+            "sha256": "f252ce33ac2db017701ca1d49813c5ec1996c4c92900cf35bb6bda7fc59c0f18"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/loongserve.md": {
+            "bytes": 12422,
+            "sha256": "7bc6d1e256d461b04acc8ed856cf99108faceda9cde729ab95bcca596fe019ca"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/orca.md": {
+            "bytes": 8826,
+            "sha256": "f5a3b5870d64226408a8fcda5d69483769d823f8a5efd7f11698c8da483ab195"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/fastserve-readset.md": {
+            "bytes": 7173,
+            "sha256": "9866bd0c78b2b638998b3b632f6bae26aef032462c81abee336d309af5b07831"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/llumnix-readset.md": {
+            "bytes": 9386,
+            "sha256": "3a2819d65ad24087c08530492ef39d8ea046ed1f780d53512f1d23e2b3578487"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/loongserve-readset.md": {
+            "bytes": 12490,
+            "sha256": "f1a383e7bdbbd5a5a472d7f6e1292c8a17d4910ff736ca5a6aa422e1bee74f8c"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/orca-readset.md": {
+            "bytes": 4796,
+            "sha256": "a497ef23b7e60c0fbc15d95b7e07cd20547cfe56f37c24dfe6ed70d79ef7a3cf"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/source-notes.md": {
+            "bytes": 9575,
+            "sha256": "84033fd514773f10c0bddd9914fabb980497ba2b308f226a922c3e1888413827"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/references/evidence-map.json": {
+            "bytes": 14384,
+            "sha256": "dd8bbab5d285ed3a1a19d2f06ad84e8c444251690d8f673def3e2681765ef6c1"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/references/repo-provenance.json": {
+            "bytes": 1313,
+            "sha256": "4e705a964697712e950e02bedc41320806a41311b7f4dcf32f923d60b1d2e919"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/references/repo-provenance.md": {
+            "bytes": 3261,
+            "sha256": "6e841c38c35ff2cd6ff8139cb1a6c8f75701647897f1089a94a5c62928609941"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/references/repo-routing-metadata.json": {
+            "bytes": 432,
+            "sha256": "a20b54f682eb52311e445ac62b95d1935c001296f5cf23b7764cb04477d535c9"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/SKILL.md": {
+            "bytes": 1835,
+            "sha256": "2a945b5fddacfb90d1cd7cb2b2f767067d641061423a65e1c19afecec7f3c06a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/references/capability-boundaries.md": {
+            "bytes": 6560,
+            "sha256": "c8081d03f04f7481080c46c7dcf728962729bc122f49f9502d50cd1e3d04bcf3"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/references/runtime-and-invariants.md": {
+            "bytes": 8265,
+            "sha256": "afa1f967545021f36bc1b59dbf6aa2c367f4d5ece4f08dce755e2017cf50dbe5"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/extend-validate/SKILL.md": {
+            "bytes": 1731,
+            "sha256": "495679e5b99341bf80daeaacd2d4c2699aba16bc16c447cbd8d2fe17f7559798"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/extend-validate/references/change-boundaries-and-validation.md": {
+            "bytes": 7624,
+            "sha256": "f7fdd2d699f09d96fbf5ca9382f9b56113c870db2688d8e1118eb8163519fc4a"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/SKILL.md": {
+            "bytes": 1812,
+            "sha256": "0ef4d2e3a96cbb277557e169a3fc8edf0f0504aad03259771b432a33ee008f78"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/references/commands-and-measurements.md": {
+            "bytes": 5667,
+            "sha256": "76ab51083b14595c49fa1c255cb7e5ebb996f14930a51ab3cbe864cc334b5f80"
+          },
+          "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/references/experiment-decisions.md": {
+            "bytes": 7644,
+            "sha256": "242593c0d386e57104b57a09bc7db97c2bbbcdb905bb68a397cfe612e07fb083"
+          },
+          "src/inferrsi/data/skill_library/skill-index.json": {
+            "bytes": 7514,
+            "sha256": "8ce35e18f7629260d67a95617b9918eecff73ed8d9370dd3d4595c725ada9aa1"
+          },
+          "src/inferrsi/data/skill_library/workflows/cite-check-downstream-pr/SKILL.md": {
+            "bytes": 1093,
+            "sha256": "485932cdf3b1fba5dc69be73e75d38f86a7b79442d26e82a296c963a7598d226"
+          },
+          "src/inferrsi/data/skill_library/workflows/claim-contract/SKILL.md": {
+            "bytes": 1075,
+            "sha256": "3af1dbbe81d91cb05fc2ef15a2ae25d7592a08558e1a5535f49c52f33a488c55"
+          },
+          "src/inferrsi/data/skill_library/workflows/competitor-scan/SKILL.md": {
+            "bytes": 1088,
+            "sha256": "0ede58ace65a9324a538b4816add1684f230ba536d4364fe5a5dd2c0bbed3776"
+          },
+          "src/inferrsi/data/skill_library/workflows/draft-acceptance-issue/SKILL.md": {
+            "bytes": 1230,
+            "sha256": "ce983acfb7d7d3272775e7cf00c92e4ddadd97511b602575a76e5ab7ade01b56"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/SKILL.md": {
+            "bytes": 5681,
+            "sha256": "364beed2ca617e71f4f955bbd024740d411b94eeb9557d7e055d4463cc22caba"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/cross-repository-mechanisms.md": {
+            "bytes": 17050,
+            "sha256": "28efa1b25f548763180b70d8d0ddebb7d4e7227df189f871fd233fe1d4881303"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/import.md": {
+            "bytes": 3463,
+            "sha256": "c8a57db94768cc6cd4193932def75ba67ec453e56ee0e54f176fb82f1af53091"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/model-aware-preparation.md": {
+            "bytes": 9056,
+            "sha256": "787c017aade68ee1133b979d636fa24062302cef6174c001171c963947ee8da4"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/overview.md": {
+            "bytes": 8099,
+            "sha256": "7ed0a966239b9481b82dd349245cfbfe966c030fc916771c91fd0692464ffc56"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/repository-source-scope.md": {
+            "bytes": 11967,
+            "sha256": "85202ca209171c50573c2fc209acb1e937d8211e1652e4ca9c6cc4bce04d03eb"
+          },
+          "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/source-manifest.json": {
+            "bytes": 1613,
+            "sha256": "0ce1678803d1ede6b485fb6b4b1c88aba909056ee6e3b5af691899ceb2056240"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/SKILL.md": {
+            "bytes": 6502,
+            "sha256": "74cd863df2e80dc2bda1713d24207809173a3116f30325e3487e1e3517cfec1e"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/metadata.json": {
+            "bytes": 3236,
+            "sha256": "12fc9272bd5fbf02dd9b897d8ca40984b72ae54f20492ace5304711bbc4d71c3"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/COMPARISON.md": {
+            "bytes": 4267,
+            "sha256": "38587678a96581f5c2dc697d982127343565504bffa191352811dffcc165aa23"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/INTEGRATION_NOTES.md": {
+            "bytes": 5557,
+            "sha256": "fbc4a96b6f6c477706eed83eaa4fcc29a14f881637629d6c004601ed049f2523"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/QA.md": {
+            "bytes": 1853,
+            "sha256": "3f3785ec39a6938a5d1f2e874e99153fe92c0e666dd3e78a499c8b47d891346a"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/README.md": {
+            "bytes": 2151,
+            "sha256": "221785e91c45588f82b3caab6d52f1ea1d102b16139c20a0802921e2789e6d28"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/UNREAD_LEADS.md": {
+            "bytes": 1038,
+            "sha256": "6c1d87ebe491d27666703149eb05ad582c7e5aff77f0ac5c11a7b89c946f96ee"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/cases/methodology-counterexamples.md": {
+            "bytes": 5213,
+            "sha256": "f9fc21753ea41eef77ca020d78d031e945e5dd835b5d09c5ff0e53749ceb2ca4"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/knowledge/evidence-routing.md": {
+            "bytes": 5282,
+            "sha256": "406d5d74c3fef4c1a23d495492e21a084c3c46495a1a84b385703ed104358874"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/learning-to-commit.md": {
+            "bytes": 2847,
+            "sha256": "9c2d1d738e896d05c00f53e6adf14c9df2501ecc3f0020e71b5a1c2192547086"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/protocol-details.md": {
+            "bytes": 6520,
+            "sha256": "41298c6bde3a28f0712fa06a18bbc4ac59aea6d17cc9eb1318744f4121e3bab6"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/readset.md": {
+            "bytes": 7086,
+            "sha256": "6f953d20d8958c68339c9ceeae348604d0fea3e9c6ccb9b64a63d3b210426a22"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/repo-to-skill.md": {
+            "bytes": 3151,
+            "sha256": "5ebdac2a2bbbe6614f9d4096bc9895bd685acdf76414e2c3c1fb7b50d57c738d"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/skillopt.md": {
+            "bytes": 3040,
+            "sha256": "148075daa67e19f4b4351710498438d17c279a9bce56b2d01cab26ff3be33e99"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/source-manifest.json": {
+            "bytes": 2431,
+            "sha256": "716cd49ed01824ee9f6b6fc55ebf2b24fdd5d5a3e677f6e28e34a472616bbe25"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/trace2skill.md": {
+            "bytes": 2917,
+            "sha256": "230d13b77ba819b5e3edee147523e8cad6bfce95afccc140fb4946d6c8747a2e"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/updates/bounded-updates.md": {
+            "bytes": 7630,
+            "sha256": "d02c5bd0109846d1e1987de074810321cb4af277cd3f66ab2ac268aadd5ee142"
+          },
+          "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/skill-factory-design-20261014.md": {
+            "bytes": 9465,
+            "sha256": "fb8ce1ad83b7a6ddc657e89ee2eebb710873b213cfd85bdd6d823c8691b3cb9f"
+          },
+          "src/inferrsi/data/skill_library/workflows/rsi-evidence/SKILL.md": {
+            "bytes": 3680,
+            "sha256": "a55610874b658b86a39a1a049efb4e4eab0140f6579ed74dd7718a4ca8cedaca"
+          }
+        },
+        "sha256": "6e19365a0ad8e2951af0f3106f93aa545b4ad00b6e0f0317a20f32065b6a05fb"
+      },
+      "environment": {
+        "filesystem_cache": "not controlled; process cold is not OS-cache cold",
+        "implementation": "CPython",
+        "logical_cpu_count": 8,
+        "machine": "AMD64",
+        "perf_counter_resolution_seconds": 1e-07,
+        "python_executable_sha256": "3470f7919170d235d7e6079691462c4b217745ec67ee612e745730e46d98f238",
+        "python_optimization": 0,
+        "python_version": "3.12.6",
+        "release": "10",
+        "site_initializers_disabled": true,
+        "system": "Windows"
+      },
+      "git_head": "407bb244651bec1c68f4b3e94b0c697b233c4a09",
+      "knowledge_record_count": 108,
+      "library": {
+        "bytes": 2294958,
+        "file_count": 342,
+        "sha256": "be8726296e6186601692dc2d95d99bfbba76f0b4c758de3a88a9b2a523eaa37e"
+      },
+      "modules": {
+        "inferrsi": {
+          "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+          "source": "src/inferrsi/__init__.py"
+        },
+        "inferrsi.catalog": {
+          "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+          "source": "src/inferrsi/catalog.py"
+        },
+        "inferrsi.cli": {
+          "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+          "source": "src/inferrsi/cli.py"
+        },
+        "inferrsi.knowledge_operations": {
+          "sha256": "c71bb2063ab8d919ca5907ac900a3930ec60730515c614fee7375d9efb77173a",
+          "source": "src/inferrsi/knowledge_operations.py"
+        }
+      },
+      "packaged_skill_count": 60,
+      "python_sources": {
+        "bytes": 274624,
+        "file_count": 53,
+        "files": {
+          "src/inferrsi/__init__.py": {
+            "bytes": 102,
+            "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47"
+          },
+          "src/inferrsi/__main__.py": {
+            "bytes": 48,
+            "sha256": "935a1c1166b0c1ea35a82256345000bf2c73ded718d77773bc27a71ecce28f7d"
+          },
+          "src/inferrsi/_skill_selection.py": {
+            "bytes": 5790,
+            "sha256": "4640399f4013425589f6702e2db844820619ba35967d1c3519978eb307168440"
+          },
+          "src/inferrsi/arguments.py": {
+            "bytes": 12200,
+            "sha256": "4456600cf4a362cf069594fe36db46932cf5cb14ddd37683aaf3538f5cf76b4d"
+          },
+          "src/inferrsi/catalog.py": {
+            "bytes": 6855,
+            "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5"
+          },
+          "src/inferrsi/claim.py": {
+            "bytes": 805,
+            "sha256": "5f24c4fa91b48f0bb4d4554531607763fc8fc108e90f332816f06cfb520b5cc1"
+          },
+          "src/inferrsi/cli.py": {
+            "bytes": 14205,
+            "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88"
+          },
+          "src/inferrsi/contracts.py": {
+            "bytes": 119,
+            "sha256": "03000cb9246ce8c0967147533f7ad25df918ba2674839801453ffb870f496490"
+          },
+          "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/scripts/roofline.py": {
+            "bytes": 9004,
+            "sha256": "524de48a6764401695fe1444f1fa3d38f363069b0cd2fba4715a86ca4b28a5d4"
+          },
+          "src/inferrsi/data/skill_library/repositories/strata/scripts/summarize_trials.py": {
+            "bytes": 5170,
+            "sha256": "6a243a08da15659ced6519f49f52a49a675bd9e952740c3ab03dd21964eefbe2"
+          },
+          "src/inferrsi/evaluation.py": {
+            "bytes": 183,
+            "sha256": "482c40710d428206c59fe3696703248cd86c4b3c7d19ad41017c1369bd6d780e"
+          },
+          "src/inferrsi/evidence_paths.py": {
+            "bytes": 859,
+            "sha256": "cc46231e0b8334ef5742defdda20e0093e22b30c86af59df1fb3ae4b5caa70fa"
+          },
+          "src/inferrsi/execution.py": {
+            "bytes": 755,
+            "sha256": "f43d8cbb4d43a72a7ff8a10b29a5fb9cf4cb22af54dd35943283c843fba0da92"
+          },
+          "src/inferrsi/harness/__init__.py": {
+            "bytes": 177,
+            "sha256": "1006e4713003f44bfb32b2e385a918e1652b9e9833801e4555674a71e57fb0d5"
+          },
+          "src/inferrsi/harness/evaluation.py": {
+            "bytes": 4160,
+            "sha256": "9b00b2e93959f16b7b76bb4b257df7dcf48520a06fd6adae96e9a165a05e6e80"
+          },
+          "src/inferrsi/harness/improvement.py": {
+            "bytes": 5777,
+            "sha256": "3faa0f5928a707a98f35f45d1ca18e226c898941e6a07b3829eebeb53e70aeb8"
+          },
+          "src/inferrsi/harness/reporting.py": {
+            "bytes": 3566,
+            "sha256": "38c58f02c44e8c8aeacf46527e142be0e3142a83e4b2693246d0ca76c0ddd7c3"
+          },
+          "src/inferrsi/harness/runner.py": {
+            "bytes": 6567,
+            "sha256": "a566307b9b586655941b627f428b5121ed38d7ac80bccd1e1799e443fb0de911"
+          },
+          "src/inferrsi/harness/scenarios/__init__.py": {
+            "bytes": 79,
+            "sha256": "4e97913c806077f0cadfeb29d986d2fa75b55ab8766cb86e6e3e9a195a9a1f5a"
+          },
+          "src/inferrsi/harness/scenarios/external.py": {
+            "bytes": 1016,
+            "sha256": "84d5f6920501a9d882cc4fffbc721d2051fdbf7c90694bde9cbfba3c2b866304"
+          },
+          "src/inferrsi/harness/scenarios/media.py": {
+            "bytes": 2914,
+            "sha256": "67d979a3ed6f13a8e3e1e40a304f78e5a3e6c2556b3aa8a11c41156745d7050a"
+          },
+          "src/inferrsi/harness/skill_evaluation.py": {
+            "bytes": 11093,
+            "sha256": "b710b8e53758757efd68c703449eb4bab7eb7ec50ae3dfc52a010cdebd8c3bf5"
+          },
+          "src/inferrsi/infra/__init__.py": {
+            "bytes": 85,
+            "sha256": "8779897733779266841b0f79b84b1dc43f0fa57c013855c906c809829ecfc77c"
+          },
+          "src/inferrsi/infra/artifacts.py": {
+            "bytes": 1040,
+            "sha256": "e72e1bebdbb88e42de430238489252386851f9e6645ed7780f2aad2e90338120"
+          },
+          "src/inferrsi/infra/campaign.py": {
+            "bytes": 1028,
+            "sha256": "be505fba1592611bd84ceaca55afbac027205413a2639baefa0c3bb343230209"
+          },
+          "src/inferrsi/infra/commands.py": {
+            "bytes": 811,
+            "sha256": "1890f115b3a2c5c6832173bf60432ad595bd6bf1c69783ecd460c6f8961194f1"
+          },
+          "src/inferrsi/infra/environment.py": {
+            "bytes": 2067,
+            "sha256": "60e46c2aefc216db202b41f44542e7b893e08e0721c667b9ccc50411d1ac84ee"
+          },
+          "src/inferrsi/infra/executors/__init__.py": {
+            "bytes": 64,
+            "sha256": "327c3cd5c821c832ff260a41dba0ee1e439cdabefc9edfb9cad1cd38d14bc87a"
+          },
+          "src/inferrsi/infra/executors/local.py": {
+            "bytes": 3654,
+            "sha256": "37bcbca9f3261c8c58a8bf806c9da40720b18d0676c82e8ed24c3901ac9bbe77"
+          },
+          "src/inferrsi/infra/http.py": {
+            "bytes": 634,
+            "sha256": "f65d6c2b55b6a89694caaf6dcdf7d74f417e9ced103eca1b3f218e305c6b6cb2"
+          },
+          "src/inferrsi/infra/inspection.py": {
+            "bytes": 2415,
+            "sha256": "d1c66deb0cced529bca996d2f88b60f51fa3417ff1cae365c650fec808946ecd"
+          },
+          "src/inferrsi/infra/platforms/__init__.py": {
+            "bytes": 1494,
+            "sha256": "9e88e8b9a396ed90c221e1585135c0a1b58da41c1bb73ed9237c53b6684abfed"
+          },
+          "src/inferrsi/infra/platforms/cuda.py": {
+            "bytes": 3309,
+            "sha256": "9e0e4276e1b2984605493b14b82fba3e86f6635204719c318302e624f66f8b6f"
+          },
+          "src/inferrsi/infra/platforms/mlx.py": {
+            "bytes": 1971,
+            "sha256": "f3b81470db28bd6183ccca2e763c5e079686bba6cdfe6c08ebe2cdf64216f2e3"
+          },
+          "src/inferrsi/infra/processes.py": {
+            "bytes": 8151,
+            "sha256": "3b96ebc7a50913e78e86ff3dcba74c8ee8d03d547d9591ca6488f390ac1e6189"
+          },
+          "src/inferrsi/infra/targets/__init__.py": {
+            "bytes": 186,
+            "sha256": "1f3ea75dded7c6cc967754c7b6d4f30ef00da8191a749938aabc3c6bb8f86eb3"
+          },
+          "src/inferrsi/infra/targets/git.py": {
+            "bytes": 2910,
+            "sha256": "6327c4f126086f310b6de309fc7659c9759da28ed76efaeeb94a3d2fba29a0c2"
+          },
+          "src/inferrsi/inspection.py": {
+            "bytes": 692,
+            "sha256": "afa44bf4cdfbb58a318c21f9b292ca142d1bbb22750e35a4cc8244da5350ebb3"
+          },
+          "src/inferrsi/knowledge_operations.py": {
+            "bytes": 27624,
+            "sha256": "c71bb2063ab8d919ca5907ac900a3930ec60730515c614fee7375d9efb77173a"
+          },
+          "src/inferrsi/knowledge_store.py": {
+            "bytes": 13700,
+            "sha256": "5e794e28906ce4687aff7f14a507e267184640fc8a4fac19e0ced279ce6a6cb6"
+          },
+          "src/inferrsi/model/__init__.py": {
+            "bytes": 77,
+            "sha256": "3c1dba4f5159fd9fbc8d4d96e2a83c8441491c1b1dc7b42031a3fc1782ed9d9f"
+          },
+          "src/inferrsi/model/anthropic_agent.py": {
+            "bytes": 24696,
+            "sha256": "b0ef9cd542544292eeb90ab05ef410be9ed6e5d1a40d729890f49562c28c12f0"
+          },
+          "src/inferrsi/model/protocols/__init__.py": {
+            "bytes": 85,
+            "sha256": "5a2f54483b164de04e309dbebb87a700b213682b0f44a50d89a752dd71a4ce47"
+          },
+          "src/inferrsi/model/protocols/image.py": {
+            "bytes": 2047,
+            "sha256": "251a1f8f885da2cd8e9d3baba766c525afe293091606b4e8668c43ba2cb7cb5f"
+          },
+          "src/inferrsi/model/protocols/metrics.py": {
+            "bytes": 610,
+            "sha256": "17db16d6ef15b41fb60d8f09507893accc43f9cb3ad686b8dc4e9839713ebc56"
+          },
+          "src/inferrsi/model/validators/__init__.py": {
+            "bytes": 65,
+            "sha256": "1d4b6ffa968e283219ddcc8a09de9280238a8d65ae2dc8cd2b60889877a8eb0b"
+          },
+          "src/inferrsi/model/validators/flare.py": {
+            "bytes": 19975,
+            "sha256": "3696b0609c772f2222b448833608565a033123d7e92b11fbc545dd356fec590f"
+          },
+          "src/inferrsi/model/validators/image.py": {
+            "bytes": 3204,
+            "sha256": "79470b6b06a22d53ec6c40636a0b3d821a2e6ee2f7bb7aed3dac7241504add1a"
+          },
+          "src/inferrsi/reporting.py": {
+            "bytes": 127,
+            "sha256": "1f771890ab3e69b1aa9246b548a7825596bdf904f6c395dd218208084266853d"
+          },
+          "src/inferrsi/research.py": {
+            "bytes": 5213,
+            "sha256": "9bfe59dd17b5476c18ce50d671711053c62800521c50d6b710e8140a2b4dca04"
+          },
+          "src/inferrsi/runspec.py": {
+            "bytes": 4398,
+            "sha256": "dca0952425caf42095ef0c7e2feb8ef4740cc422ecb61e777f203962d9b87709"
+          },
+          "src/inferrsi/skill_catalog.py": {
+            "bytes": 9938,
+            "sha256": "65f5a1f3f6894009abb98ec2fec30a63cc1bb2f6ce6aeb01f9f7a3dce0439f2d"
+          },
+          "src/inferrsi/summary_chart.py": {
+            "bytes": 40910,
+            "sha256": "d29023d60f6be804c1b01c9eb73765d4824d6c3bb870fe22c99a24544bc1d8bc"
+          }
+        },
+        "sha256": "d6edbaf975cf3bc00902d6aa7cf2bb955dc58b3e37c74f59bb226c5f0d2fa261"
+      },
+      "tracked_source_dirty": false
+    }
+  },
+  "fixture": {
+    "cases": 24,
+    "answerable": 20,
+    "negative": 4,
+    "fixture_sha256": "149092f60065b31ad95a744188699ddd3acf60c6200c67983d4b01837cb233ab"
+  },
+  "variants": {
+    "A": {
+      "modules": {
+        "inferrsi": {
+          "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+          "source": "src/inferrsi/__init__.py"
+        },
+        "inferrsi.catalog": {
+          "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+          "source": "src/inferrsi/catalog.py"
+        },
+        "inferrsi.cli": {
+          "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+          "source": "src/inferrsi/cli.py"
+        },
+        "inferrsi.knowledge_operations": {
+          "sha256": "b601568bbd3d7eae9aa31193b9f281483328a46e32496c30a3a4bad5fb001a0b",
+          "source": "src/inferrsi/knowledge_operations.py"
+        }
+      },
+      "outputs": [
+        {
+          "answerable": true,
+          "case": "dev01",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev01",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 428778,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_device_health",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.dcp_async_save_performance",
+                "ref": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "resource_count": 343,
+                "resource_hashes_sha256": "b59637d16a7076477f6011fcceeb3cb5b402aa95af17b841ac9b2ede396216e3",
+                "revision": "sha256:36243cf65b6eaa5f370fd4f7633b8fccb119ff75ed38ba652a6061469de0b5b3"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.checkpoint_recovery_consistency",
+                "ref": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "15c906f896eab2a681eb9bc9e42ee29d11b0594a1e1aeacbf6f414233209d1fa",
+                "revision": "sha256:73f77f89e298ce8724258cb3d74315438c036297bb38c63aac3d7dc95473ce53"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.inference_benchmark_design",
+                "ref": "package:inferrsi.common.skills.inference_benchmark_design",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb743a7fe0662feead981795dc60e2a5c9355ed876339f14ad7a47caab0bb670",
+                "revision": "sha256:6b0cbeaa5d409c04228883884e192f275cad4874c7f6b3039abef9fd443fafb5"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.kv_tier_management",
+                "ref": "package:inferrsi.common.skills.kv_tier_management",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ee42b55fa8a404149ecd1dccc32a892566ee8867710531bc0af553409758a019",
+                "revision": "sha256:5816c2e526a71152486af2230ace24d27b54baeb373bf5d27e353ccd44f640aa"
+              }
+            ],
+            "selection": {
+              "cue": "Why must checkpoint upload finish before cached staging buffers can be reused?",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 434223,
+            "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev01",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 64465,
+            "context_omissions": [
+              {
+                "available_bytes": 11883,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "required_bytes": 14400
+              },
+              {
+                "available_bytes": 11883,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.inference_benchmark_design",
+                "required_bytes": 47014
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md"
+                  }
+                ],
+                "content_bytes": 12636,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md",
+                    "role": "core",
+                    "sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md",
+                    "role": "reference",
+                    "sha256": "22afe4f4d77f441096cc657d55d763473bfcbea8fb531c8e493fe49ab84955f7"
+                  }
+                ],
+                "revision": "sha256:36243cf65b6eaa5f370fd4f7633b8fccb119ff75ed38ba652a6061469de0b5b3",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md"
+                  }
+                ],
+                "content_bytes": 41017,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Normalized provenance covers the existing llama.cpp loading slice at 3e390af18de1131dda27913c00511725b4274c8a and the fastsafetensors distributed-delivery slice at 219e2561a41ede2d5d8eba611c248bb0effbf580; other inherited claims remain unnormalized with original identities and limits retained.",
+                  "The readsets record static mechanisms and source recommendations, not validated startup, memory, placement or backend performance.",
+                  "The fastsafetensors delivery readset does not revalidate inherited I/O/save claims or qualify uneven scatter, accelerator/NPU compatibility, inter-node transport or agent efficacy.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/weight-loading-optimization/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                    "role": "core",
+                    "sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                    "role": "reference",
+                    "sha256": "6c640f28ee3f0253a17b81461c865cb9f947367002cb75bf72550c4572d02e97"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                    "role": "reference",
+                    "sha256": "5a39f14b102b0cee304541e011aa31e44e6b13527d1bee9987c3f3b66c064d3d"
+                  }
+                ],
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md"
+                  }
+                ],
+                "content_bytes": 10812,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/repositories/lmcache/kv-tier-management/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.kv_tier_management",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md",
+                    "role": "core",
+                    "sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27"
+                  }
+                ],
+                "revision": "sha256:5816c2e526a71152486af2230ace24d27b54baeb373bf5d27e353ccd44f640aa",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 458836,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_device_health",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.dcp_async_save_performance",
+                "ref": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "resource_count": 343,
+                "resource_hashes_sha256": "b59637d16a7076477f6011fcceeb3cb5b402aa95af17b841ac9b2ede396216e3",
+                "revision": "sha256:36243cf65b6eaa5f370fd4f7633b8fccb119ff75ed38ba652a6061469de0b5b3"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.checkpoint_recovery_consistency",
+                "ref": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "15c906f896eab2a681eb9bc9e42ee29d11b0594a1e1aeacbf6f414233209d1fa",
+                "revision": "sha256:73f77f89e298ce8724258cb3d74315438c036297bb38c63aac3d7dc95473ce53"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.inference_benchmark_design",
+                "ref": "package:inferrsi.common.skills.inference_benchmark_design",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb743a7fe0662feead981795dc60e2a5c9355ed876339f14ad7a47caab0bb670",
+                "revision": "sha256:6b0cbeaa5d409c04228883884e192f275cad4874c7f6b3039abef9fd443fafb5"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.kv_tier_management",
+                "ref": "package:inferrsi.common.skills.kv_tier_management",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ee42b55fa8a404149ecd1dccc32a892566ee8867710531bc0af553409758a019",
+                "revision": "sha256:5816c2e526a71152486af2230ace24d27b54baeb373bf5d27e353ccd44f640aa"
+              }
+            ],
+            "selection": {
+              "cue": "Why must checkpoint upload finish before cached staging buffers can be reused?",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 540055,
+            "sha256": "1afecd2e68d87dd1ef1e3c1c22ea846fd33c26bd751560219948aae98577f481"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev01",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev02",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev02",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 85545,
+            "no_match": false,
+            "omissions": [],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.dcp_async_save_performance",
+                "ref": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "resource_count": 343,
+                "resource_hashes_sha256": "b59637d16a7076477f6011fcceeb3cb5b402aa95af17b841ac9b2ede396216e3",
+                "revision": "sha256:36243cf65b6eaa5f370fd4f7633b8fccb119ff75ed38ba652a6061469de0b5b3"
+              }
+            ],
+            "selection": {
+              "cue": "DefaultStager",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 85990,
+            "sha256": "27b25a4ec6fa85219601eeb68b85fd54f3dc51c0d529a480f782ec7102a49684"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev02",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 12636,
+            "context_omissions": [],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md"
+                  }
+                ],
+                "content_bytes": 12636,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md",
+                    "role": "core",
+                    "sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md",
+                    "role": "reference",
+                    "sha256": "22afe4f4d77f441096cc657d55d763473bfcbea8fb531c8e493fe49ab84955f7"
+                  }
+                ],
+                "revision": "sha256:36243cf65b6eaa5f370fd4f7633b8fccb119ff75ed38ba652a6061469de0b5b3",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 89815,
+            "no_match": false,
+            "omissions": [],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.dcp_async_save_performance",
+                "ref": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "resource_count": 343,
+                "resource_hashes_sha256": "b59637d16a7076477f6011fcceeb3cb5b402aa95af17b841ac9b2ede396216e3",
+                "revision": "sha256:36243cf65b6eaa5f370fd4f7633b8fccb119ff75ed38ba652a6061469de0b5b3"
+              }
+            ],
+            "selection": {
+              "cue": "DefaultStager",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 106414,
+            "sha256": "672f7b868df0011c492de557341606d2b17761814ad5c5c5b8bfc5107699cd12"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev02",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev03",
+          "mode": "search",
+          "output": {
+            "ordered_items": [
+              {
+                "id": "inferrsi.common.skills.weight_loading_optimization"
+              }
+            ],
+            "serialized_bytes": 7193,
+            "sha256": "eee7ad1236b5121614e35949af184d10c7fed14b9d790befbdd850f12081ef2d"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev03",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 429394,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.training_input_pipeline",
+                "ref": "package:inferrsi.common.skills.training_input_pipeline",
+                "resource_count": 343,
+                "resource_hashes_sha256": "2316be47093b856fff9936ba91f578c2c42c12938ecedc7eae24e16a0fb6385a",
+                "revision": "sha256:d2e9d3993b49d3de2eac8ee492e8467c80bbb26787cee0687bb78a76c111f3d5"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.bound_windowed_attention_replay",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb70e2e746f7ba833a696431c6a4deda830188acf5c4eafce6b0008818f8190a",
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_distributed_parallel",
+                "ref": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ecb6d382307b933beb053b3c380a5bb6ba87c13792c428ad4131c78f5a9bcd04",
+                "revision": "sha256:66dab07081fad44b9cfe69b9c0f7a8f1dbb2c3fa19157ac99e4eee9a6306f605"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascendc_tiling_design",
+                "ref": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "resource_count": 343,
+                "resource_hashes_sha256": "caa4787f895186508b8b6f863b2d03a830364585ffd65f0c098be319c7d54cad",
+                "revision": "sha256:9c90d2bc30d84cf0c99a0382d1e61b01c1210efe51d16ddb81c25afcb7c04e4b"
+              }
+            ],
+            "selection": {
+              "cue": "safetensors index sharded downloads",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 430850,
+            "sha256": "6e8b364dc7518b24f75400cf2a48525b9b047a7b19a316ba8bcc7711a48ff089"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev03",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 62379,
+            "context_omissions": [
+              {
+                "available_bytes": 15017,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "required_bytes": 29732
+              },
+              {
+                "available_bytes": 15017,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "required_bytes": 24091
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md"
+                  }
+                ],
+                "content_bytes": 41017,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Normalized provenance covers the existing llama.cpp loading slice at 3e390af18de1131dda27913c00511725b4274c8a and the fastsafetensors distributed-delivery slice at 219e2561a41ede2d5d8eba611c248bb0effbf580; other inherited claims remain unnormalized with original identities and limits retained.",
+                  "The readsets record static mechanisms and source recommendations, not validated startup, memory, placement or backend performance.",
+                  "The fastsafetensors delivery readset does not revalidate inherited I/O/save claims or qualify uneven scatter, accelerator/NPU compatibility, inter-node transport or agent efficacy.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/weight-loading-optimization/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                    "role": "core",
+                    "sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                    "role": "reference",
+                    "sha256": "6c640f28ee3f0253a17b81461c865cb9f947367002cb75bf72550c4572d02e97"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                    "role": "reference",
+                    "sha256": "5a39f14b102b0cee304541e011aa31e44e6b13527d1bee9987c3f3b66c064d3d"
+                  }
+                ],
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "64bd2c6b795d966045d1d3cca55976d3d49b08b06df7c7e06a70c2824cf55b7e",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/training-input-pipeline/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/training-input-pipeline/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md"
+                  }
+                ],
+                "content_bytes": 9502,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/training-input-pipeline/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.training_input_pipeline",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/training-input-pipeline/SKILL.md",
+                    "role": "core",
+                    "sha256": "64bd2c6b795d966045d1d3cca55976d3d49b08b06df7c7e06a70c2824cf55b7e"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/training-input-pipeline/references/sources.md",
+                    "role": "reference",
+                    "sha256": "60e0f3e88655d8e1d5eb4797f39ca4370deda01614f265b8d2ea4559d8ac020f"
+                  }
+                ],
+                "revision": "sha256:d2e9d3993b49d3de2eac8ee492e8467c80bbb26787cee0687bb78a76c111f3d5",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "64bd2c6b795d966045d1d3cca55976d3d49b08b06df7c7e06a70c2824cf55b7e",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/training-input-pipeline/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/training-input-pipeline/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "93aa1b775ced304cebfe6c4dec0503ddb2260ecb3cb01edc17ce329446c5ce2b",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md"
+                  }
+                ],
+                "content_bytes": 11860,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/ascend/ascendc-tiling-design/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md",
+                    "role": "core",
+                    "sha256": "93aa1b775ced304cebfe6c4dec0503ddb2260ecb3cb01edc17ce329446c5ce2b"
+                  }
+                ],
+                "revision": "sha256:9c90d2bc30d84cf0c99a0382d1e61b01c1210efe51d16ddb81c25afcb7c04e4b",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "93aa1b775ced304cebfe6c4dec0503ddb2260ecb3cb01edc17ce329446c5ce2b",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 458004,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.training_input_pipeline",
+                "ref": "package:inferrsi.common.skills.training_input_pipeline",
+                "resource_count": 343,
+                "resource_hashes_sha256": "2316be47093b856fff9936ba91f578c2c42c12938ecedc7eae24e16a0fb6385a",
+                "revision": "sha256:d2e9d3993b49d3de2eac8ee492e8467c80bbb26787cee0687bb78a76c111f3d5"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.bound_windowed_attention_replay",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb70e2e746f7ba833a696431c6a4deda830188acf5c4eafce6b0008818f8190a",
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_distributed_parallel",
+                "ref": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ecb6d382307b933beb053b3c380a5bb6ba87c13792c428ad4131c78f5a9bcd04",
+                "revision": "sha256:66dab07081fad44b9cfe69b9c0f7a8f1dbb2c3fa19157ac99e4eee9a6306f605"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascendc_tiling_design",
+                "ref": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "resource_count": 343,
+                "resource_hashes_sha256": "caa4787f895186508b8b6f863b2d03a830364585ffd65f0c098be319c7d54cad",
+                "revision": "sha256:9c90d2bc30d84cf0c99a0382d1e61b01c1210efe51d16ddb81c25afcb7c04e4b"
+              }
+            ],
+            "selection": {
+              "cue": "safetensors index sharded downloads",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 532970,
+            "sha256": "6c8cbb5c8b5fcc958f84efa913dbd145e9c2066a2cb8ca16163d8880ff05fb0e"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev03",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 86217,
+            "no_match": false,
+            "omissions": [],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              }
+            ],
+            "serialized_bytes": 86534,
+            "sha256": "4a4c102e42ef0cefd5cb90dc4b7b0b310b8b0fa7f93e35e1f70aaae802e1d305"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev04",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev04",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 433875,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_device_health",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.budget_speculative_verification",
+                "ref": "package:inferrsi.common.skills.budget_speculative_verification",
+                "resource_count": 343,
+                "resource_hashes_sha256": "6ef4bcbc8e2ee8a5fc04c5c970ce5ec223bc85904818d7e006fb6d67548d734b",
+                "revision": "sha256:2686550fa366a367d026447ac1f2bec6b9ad3a81d559b70dccb8098524a27e2d"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_model_adaptation",
+                "ref": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "resource_count": 343,
+                "resource_hashes_sha256": "cd4f45eea332240ebf0229078a47067fe262badd13aae30db4e0e90868340861",
+                "revision": "sha256:1d03cc8e797c6f6dcda862244833019e8bb77941f1f63aecc8e7958bb8265863"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.host_overhead_optimization",
+                "ref": "package:inferrsi.common.skills.host_overhead_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "4fbb3ca8057addd4564893430704c2ffd32c8f7ea622d3d0ddc46ffb37332a86",
+                "revision": "sha256:902b79f76c19ecd9856af1e64b4a0ee916801b7e3e7e7f5bf6c3f105492a9ea0"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.performance_ceiling_analysis",
+                "ref": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c915247877787fcecc949d1d212ae1f18e3ba2a82f02b7c1cf343dedf0a7e885",
+                "revision": "sha256:6f67a5d347f3d05198f92cfce346d30051a544427ec8922b9de64432c3ffe316"
+              }
+            ],
+            "selection": {
+              "cue": "How should model loading overhead be separated from per request latency?",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 439307,
+            "sha256": "6ab5e1e796794df07d09ea5cfaa3d9a3870d601a9d675c2c22026d301ba2e764"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev04",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 63529,
+            "context_omissions": [
+              {
+                "available_bytes": 24519,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.budget_speculative_verification",
+                "required_bytes": 30031
+              },
+              {
+                "available_bytes": 2007,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "required_bytes": 170782
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md"
+                  }
+                ],
+                "content_bytes": 41017,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Normalized provenance covers the existing llama.cpp loading slice at 3e390af18de1131dda27913c00511725b4274c8a and the fastsafetensors distributed-delivery slice at 219e2561a41ede2d5d8eba611c248bb0effbf580; other inherited claims remain unnormalized with original identities and limits retained.",
+                  "The readsets record static mechanisms and source recommendations, not validated startup, memory, placement or backend performance.",
+                  "The fastsafetensors delivery readset does not revalidate inherited I/O/save claims or qualify uneven scatter, accelerator/NPU compatibility, inter-node transport or agent efficacy.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/weight-loading-optimization/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                    "role": "core",
+                    "sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                    "role": "reference",
+                    "sha256": "6c640f28ee3f0253a17b81461c865cb9f947367002cb75bf72550c4572d02e97"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                    "role": "reference",
+                    "sha256": "5a39f14b102b0cee304541e011aa31e44e6b13527d1bee9987c3f3b66c064d3d"
+                  }
+                ],
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "8fdf46fb2743b67a66bc069d22a45305a222debfb606777bc451b4b8f31d9cc6",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/SKILL.md"
+                  }
+                ],
+                "content_bytes": 6440,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/ascend/ascend-model-adaptation/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/ascend-model-adaptation/SKILL.md",
+                    "role": "core",
+                    "sha256": "8fdf46fb2743b67a66bc069d22a45305a222debfb606777bc451b4b8f31d9cc6"
+                  }
+                ],
+                "revision": "sha256:1d03cc8e797c6f6dcda862244833019e8bb77941f1f63aecc8e7958bb8265863",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "8fdf46fb2743b67a66bc069d22a45305a222debfb606777bc451b4b8f31d9cc6",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "36ef88826a32ac3147b869001004ded17a8d5a491e25bfe105005ba39ea9d1eb",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/metadata.json",
+                      "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/references/deepinfer-zero-overhead-scheduling.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/SKILL.md"
+                  }
+                ],
+                "content_bytes": 16072,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/host/host-overhead-optimization/metadata.json",
+                    "reason": "not_context_resource"
+                  },
+                  {
+                    "path": "data/skill_library/hardware/host/host-overhead-optimization/references/deepinfer-zero-overhead-scheduling.md",
+                    "reason": "context_limit"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.host_overhead_optimization",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/host/host-overhead-optimization/SKILL.md",
+                    "role": "core",
+                    "sha256": "36ef88826a32ac3147b869001004ded17a8d5a491e25bfe105005ba39ea9d1eb"
+                  }
+                ],
+                "revision": "sha256:902b79f76c19ecd9856af1e64b4a0ee916801b7e3e7e7f5bf6c3f105492a9ea0",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "36ef88826a32ac3147b869001004ded17a8d5a491e25bfe105005ba39ea9d1eb",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/metadata.json",
+                      "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/references/deepinfer-zero-overhead-scheduling.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 469108,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_device_health",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.budget_speculative_verification",
+                "ref": "package:inferrsi.common.skills.budget_speculative_verification",
+                "resource_count": 343,
+                "resource_hashes_sha256": "6ef4bcbc8e2ee8a5fc04c5c970ce5ec223bc85904818d7e006fb6d67548d734b",
+                "revision": "sha256:2686550fa366a367d026447ac1f2bec6b9ad3a81d559b70dccb8098524a27e2d"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_model_adaptation",
+                "ref": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "resource_count": 343,
+                "resource_hashes_sha256": "cd4f45eea332240ebf0229078a47067fe262badd13aae30db4e0e90868340861",
+                "revision": "sha256:1d03cc8e797c6f6dcda862244833019e8bb77941f1f63aecc8e7958bb8265863"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.host_overhead_optimization",
+                "ref": "package:inferrsi.common.skills.host_overhead_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "4fbb3ca8057addd4564893430704c2ffd32c8f7ea622d3d0ddc46ffb37332a86",
+                "revision": "sha256:902b79f76c19ecd9856af1e64b4a0ee916801b7e3e7e7f5bf6c3f105492a9ea0"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.performance_ceiling_analysis",
+                "ref": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c915247877787fcecc949d1d212ae1f18e3ba2a82f02b7c1cf343dedf0a7e885",
+                "revision": "sha256:6f67a5d347f3d05198f92cfce346d30051a544427ec8922b9de64432c3ffe316"
+              }
+            ],
+            "selection": {
+              "cue": "How should model loading overhead be separated from per request latency?",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 549459,
+            "sha256": "d211e3d91c2a81d9eb94f71c8e5bf368af4073b03f869d87154e196aefb8dcca"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev04",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev05",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev05",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 429290,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.storage_io_measurement",
+                "ref": "package:inferrsi.common.skills.storage_io_measurement",
+                "resource_count": 343,
+                "resource_hashes_sha256": "350d9c963ece5f1e1359ec238f4f804912b7bf0f08743b9cd4b6207583d52515",
+                "revision": "sha256:a57650b2bcaaac84bce2725a49eadb8c17533c82a1125b5ed5d840df5fca201c"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.bound_windowed_attention_replay",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb70e2e746f7ba833a696431c6a4deda830188acf5c4eafce6b0008818f8190a",
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_operator_profiling",
+                "ref": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "resource_count": 343,
+                "resource_hashes_sha256": "6f65a5d36f2fb4ec6097cb8aab45c4b2266f8808da5feed0f1b3800bca65395c",
+                "revision": "sha256:d467d8d6c966efe5dce163a74e15cd8714a6c9a7a008374e839cbe13973f2f88"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_distributed_parallel",
+                "ref": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ecb6d382307b933beb053b3c380a5bb6ba87c13792c428ad4131c78f5a9bcd04",
+                "revision": "sha256:66dab07081fad44b9cfe69b9c0f7a8f1dbb2c3fa19157ac99e4eee9a6306f605"
+              }
+            ],
+            "selection": {
+              "cue": "fio iodepth achieved depth",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 430369,
+            "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev05",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 51533,
+            "context_omissions": [
+              {
+                "available_bytes": 14003,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "required_bytes": 41017
+              },
+              {
+                "available_bytes": 14003,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "required_bytes": 24091
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "63b883c10a8d3e7d918abfc7175b72e2b401ff1e02697f94b8382ec1c8f369b7",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md"
+                  }
+                ],
+                "content_bytes": 8813,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/host/storage-io-measurement/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.storage_io_measurement",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/host/storage-io-measurement/SKILL.md",
+                    "role": "core",
+                    "sha256": "63b883c10a8d3e7d918abfc7175b72e2b401ff1e02697f94b8382ec1c8f369b7"
+                  }
+                ],
+                "revision": "sha256:a57650b2bcaaac84bce2725a49eadb8c17533c82a1125b5ed5d840df5fca201c",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "63b883c10a8d3e7d918abfc7175b72e2b401ff1e02697f94b8382ec1c8f369b7",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "4703beb6e5194ef2ff2c0b5ec69f64e35417a4bc0dfb50a10c80a920b77ef3b5",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/validation.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md"
+                  }
+                ],
+                "content_bytes": 29732,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no equivalence run was performed.",
+                  "Applies to architectures with a demonstrated boundary dependency, not to arbitrary sliding-window stacks.",
+                  "Window sizes, boundary-layer index and ring lengths are model-config specific and are not transferable.",
+                  "Query-row compaction does not imply dropping historical keys; the two decisions are separate.",
+                  "Prompt logprobs are unrecoverable for the clipped region; the reference forces them off with a warning.",
+                  "The truncation path is off by default in the reference source.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md",
+                    "role": "core",
+                    "sha256": "4703beb6e5194ef2ff2c0b5ec69f64e35417a4bc0dfb50a10c80a920b77ef3b5"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "fd917bd5ac80bee9c9f4525f8e0300f6d67dc144e2a16632af42a67f3d53d098"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/validation.md",
+                    "role": "reference",
+                    "sha256": "60e71e70c5519bcdf6c171eec2be2f342e8801bedd746d6005f11adc646dfaff"
+                  }
+                ],
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "4703beb6e5194ef2ff2c0b5ec69f64e35417a4bc0dfb50a10c80a920b77ef3b5",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/validation.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "401981cf7828edcab0c9ea882f7ecd83badfa3c2a95006e912cf99d67b3d6a12",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/SKILL.md"
+                  }
+                ],
+                "content_bytes": 12988,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/ascend/ascend-operator-profiling/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/ascend-operator-profiling/SKILL.md",
+                    "role": "core",
+                    "sha256": "401981cf7828edcab0c9ea882f7ecd83badfa3c2a95006e912cf99d67b3d6a12"
+                  }
+                ],
+                "revision": "sha256:d467d8d6c966efe5dce163a74e15cd8714a6c9a7a008374e839cbe13973f2f88",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "401981cf7828edcab0c9ea882f7ecd83badfa3c2a95006e912cf99d67b3d6a12",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 458087,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.storage_io_measurement",
+                "ref": "package:inferrsi.common.skills.storage_io_measurement",
+                "resource_count": 343,
+                "resource_hashes_sha256": "350d9c963ece5f1e1359ec238f4f804912b7bf0f08743b9cd4b6207583d52515",
+                "revision": "sha256:a57650b2bcaaac84bce2725a49eadb8c17533c82a1125b5ed5d840df5fca201c"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.bound_windowed_attention_replay",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb70e2e746f7ba833a696431c6a4deda830188acf5c4eafce6b0008818f8190a",
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_operator_profiling",
+                "ref": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "resource_count": 343,
+                "resource_hashes_sha256": "6f65a5d36f2fb4ec6097cb8aab45c4b2266f8808da5feed0f1b3800bca65395c",
+                "revision": "sha256:d467d8d6c966efe5dce163a74e15cd8714a6c9a7a008374e839cbe13973f2f88"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_distributed_parallel",
+                "ref": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ecb6d382307b933beb053b3c380a5bb6ba87c13792c428ad4131c78f5a9bcd04",
+                "revision": "sha256:66dab07081fad44b9cfe69b9c0f7a8f1dbb2c3fa19157ac99e4eee9a6306f605"
+              }
+            ],
+            "selection": {
+              "cue": "fio iodepth achieved depth",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 521491,
+            "sha256": "8a3a1fa6c43063fe3165b30705447e22dcdc404bfb018739b401cf81e5f48061"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev05",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev06",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev06",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 431262,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_device_health",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.storage_io_measurement",
+                "ref": "package:inferrsi.common.skills.storage_io_measurement",
+                "resource_count": 343,
+                "resource_hashes_sha256": "350d9c963ece5f1e1359ec238f4f804912b7bf0f08743b9cd4b6207583d52515",
+                "revision": "sha256:a57650b2bcaaac84bce2725a49eadb8c17533c82a1125b5ed5d840df5fca201c"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.cann_attribution_overlap",
+                "ref": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c0bf9f5ab49059fd46fbee3e58bc0fac2819bb9d727e921d2362e618e92f996c",
+                "revision": "sha256:425a1e6d432ff3aa077b6974b41c66fb0a88be4ae88b8ba7a827071dfd8f5f10"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "ref": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "resource_count": 343,
+                "resource_hashes_sha256": "3f8f5e717bf501c4d94bfb3ac40c15393b07dd519c9170343aeb54445fbe933c",
+                "revision": "sha256:9028b6d32dcbe068af6fc777e80162bc815835f63657dc536f93d9e8ce5599f3"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.npu_skill_factory",
+                "ref": "package:inferrsi.common.skills.npu_skill_factory",
+                "resource_count": 343,
+                "resource_hashes_sha256": "7cde43221171eca8e11527545216e348d8c464438bec7458aab055843bb440cf",
+                "revision": "sha256:c789873dd4195bee85724a481228f51852a09c870d630ad1e828901037f875b7"
+              }
+            ],
+            "selection": {
+              "cue": "How can I attribute disk traffic to the loader process rather than the whole device?",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 436719,
+            "sha256": "6ed7e4937400c60c3bda007715df3506107846161efe870350057d4949a91883"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev06",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 35244,
+            "context_omissions": [
+              {
+                "available_bytes": 30292,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "required_bytes": 41017
+              },
+              {
+                "available_bytes": 30292,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.npu_skill_factory",
+                "required_bytes": 65054
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "63b883c10a8d3e7d918abfc7175b72e2b401ff1e02697f94b8382ec1c8f369b7",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md"
+                  }
+                ],
+                "content_bytes": 8813,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/host/storage-io-measurement/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.storage_io_measurement",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/host/storage-io-measurement/SKILL.md",
+                    "role": "core",
+                    "sha256": "63b883c10a8d3e7d918abfc7175b72e2b401ff1e02697f94b8382ec1c8f369b7"
+                  }
+                ],
+                "revision": "sha256:a57650b2bcaaac84bce2725a49eadb8c17533c82a1125b5ed5d840df5fca201c",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "63b883c10a8d3e7d918abfc7175b72e2b401ff1e02697f94b8382ec1c8f369b7",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "0e6f094dc554abea12e07def50f20b233b47127779a8aefad595b4ab1efd24c2",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md"
+                  }
+                ],
+                "content_bytes": 8972,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/ascend/cann-attribution-overlap/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md",
+                    "role": "core",
+                    "sha256": "0e6f094dc554abea12e07def50f20b233b47127779a8aefad595b4ab1efd24c2"
+                  }
+                ],
+                "revision": "sha256:425a1e6d432ff3aa077b6974b41c66fb0a88be4ae88b8ba7a827071dfd8f5f10",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "0e6f094dc554abea12e07def50f20b233b47127779a8aefad595b4ab1efd24c2",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "8428df7157af024fa17cfe0b3818ed54bd9f468306dda46eb8d4f0edf4f3f230",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/metadata.json",
+                      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md"
+                  }
+                ],
+                "content_bytes": 17459,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md",
+                    "role": "core",
+                    "sha256": "8428df7157af024fa17cfe0b3818ed54bd9f468306dda46eb8d4f0edf4f3f230"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "80cba4d3a884e7ee2618326cf74b9f9c3a34e4d9cf443707f6bbe673f2d5ba65"
+                  }
+                ],
+                "revision": "sha256:9028b6d32dcbe068af6fc777e80162bc815835f63657dc536f93d9e8ce5599f3",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "8428df7157af024fa17cfe0b3818ed54bd9f468306dda46eb8d4f0edf4f3f230",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/metadata.json",
+                      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 463110,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_device_health",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.storage_io_measurement",
+                "ref": "package:inferrsi.common.skills.storage_io_measurement",
+                "resource_count": 343,
+                "resource_hashes_sha256": "350d9c963ece5f1e1359ec238f4f804912b7bf0f08743b9cd4b6207583d52515",
+                "revision": "sha256:a57650b2bcaaac84bce2725a49eadb8c17533c82a1125b5ed5d840df5fca201c"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.cann_attribution_overlap",
+                "ref": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c0bf9f5ab49059fd46fbee3e58bc0fac2819bb9d727e921d2362e618e92f996c",
+                "revision": "sha256:425a1e6d432ff3aa077b6974b41c66fb0a88be4ae88b8ba7a827071dfd8f5f10"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "ref": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "resource_count": 343,
+                "resource_hashes_sha256": "3f8f5e717bf501c4d94bfb3ac40c15393b07dd519c9170343aeb54445fbe933c",
+                "revision": "sha256:9028b6d32dcbe068af6fc777e80162bc815835f63657dc536f93d9e8ce5599f3"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.npu_skill_factory",
+                "ref": "package:inferrsi.common.skills.npu_skill_factory",
+                "resource_count": 343,
+                "resource_hashes_sha256": "7cde43221171eca8e11527545216e348d8c464438bec7458aab055843bb440cf",
+                "revision": "sha256:c789873dd4195bee85724a481228f51852a09c870d630ad1e828901037f875b7"
+              }
+            ],
+            "selection": {
+              "cue": "How can I attribute disk traffic to the loader process rather than the whole device?",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 513297,
+            "sha256": "467b97551281ea312611214ece5f7200de1da656c9af15e5d2aa12084eb757df"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev06",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev07",
+          "mode": "search",
+          "output": {
+            "ordered_items": [
+              {
+                "id": "inferrsi.ascend.skills.ascendc_tiling_design"
+              }
+            ],
+            "serialized_bytes": 4033,
+            "sha256": "f87cc18112737157abb8fe1f4c27a0ceaf9656fca6ee8fbc84e760b14fd69a5d"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev07",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 429624,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_device_health",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascendc_tiling_design",
+                "ref": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "resource_count": 343,
+                "resource_hashes_sha256": "caa4787f895186508b8b6f863b2d03a830364585ffd65f0c098be319c7d54cad",
+                "revision": "sha256:9c90d2bc30d84cf0c99a0382d1e61b01c1210efe51d16ddb81c25afcb7c04e4b"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.assess_split_backward_schedule",
+                "ref": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "resource_count": 343,
+                "resource_hashes_sha256": "e3cbdc3297e17bf0af02a32b0fd231cb1f645fd7d21844b7764cd1b61cdc9243",
+                "revision": "sha256:cec60c01e6c992705c02cc68ab17a2a437ecf2cfbd03489b45fc98ac7692b044"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_operator_tuning",
+                "ref": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "resource_count": 343,
+                "resource_hashes_sha256": "1f4ba53716e4b333251713ef826742cb9596f086412d26e2715a3e71c75d6211",
+                "revision": "sha256:01c26a7a1203f0d8fea8e035e875c1c0e59bbbcc68a666d497a44071b05c7323"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "ref": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "resource_count": 343,
+                "resource_hashes_sha256": "7b4026d946b56b7fdafc6e086144d5ab04f6951f233afd5cfbcf003c04d05287",
+                "revision": "sha256:5de043dd66cc20bccdbb002ceee7efffef2136e4296e038406bf1e9884b1ec77"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sd_cpp_residency",
+                "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c83d7a76dffa5a3d97e932a3eac4602ad82e20babb09d9942504fff6b4db53d6",
+                "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde"
+              }
+            ],
+            "selection": {
+              "cue": "UB split buffer branch matrix",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 434145,
+            "sha256": "b176000866b7d392ad408fae0f6fef461f98a6c16a159c9955a0f87ee57a027e"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev07",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 64663,
+            "context_omissions": [],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "93aa1b775ced304cebfe6c4dec0503ddb2260ecb3cb01edc17ce329446c5ce2b",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md"
+                  }
+                ],
+                "content_bytes": 11860,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/ascend/ascendc-tiling-design/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md",
+                    "role": "core",
+                    "sha256": "93aa1b775ced304cebfe6c4dec0503ddb2260ecb3cb01edc17ce329446c5ce2b"
+                  }
+                ],
+                "revision": "sha256:9c90d2bc30d84cf0c99a0382d1e61b01c1210efe51d16ddb81c25afcb7c04e4b",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "93aa1b775ced304cebfe6c4dec0503ddb2260ecb3cb01edc17ce329446c5ce2b",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "1e64f28ea71672fb53c56f6108b87864738c6d03d6dcd32eedaef9e0625e96e3",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/SKILL.md"
+                  }
+                ],
+                "content_bytes": 13729,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/assess-split-backward-schedule/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/assess-split-backward-schedule/SKILL.md",
+                    "role": "core",
+                    "sha256": "1e64f28ea71672fb53c56f6108b87864738c6d03d6dcd32eedaef9e0625e96e3"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/assess-split-backward-schedule/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "61b1609f71c37b61d136b8c4d39f793de87cffb1cf31c8ec737a1c3fa6fb1b4e"
+                  }
+                ],
+                "revision": "sha256:cec60c01e6c992705c02cc68ab17a2a437ecf2cfbd03489b45fc98ac7692b044",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "1e64f28ea71672fb53c56f6108b87864738c6d03d6dcd32eedaef9e0625e96e3",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "38b151bc6b4e9734c6fa41efb7a30343559400c0c75bbdd26bd8a968b8ea4592",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/SKILL.md"
+                  }
+                ],
+                "content_bytes": 12236,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/ascend/ascend-operator-tuning/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/ascend-operator-tuning/SKILL.md",
+                    "role": "core",
+                    "sha256": "38b151bc6b4e9734c6fa41efb7a30343559400c0c75bbdd26bd8a968b8ea4592"
+                  }
+                ],
+                "revision": "sha256:01c26a7a1203f0d8fea8e035e875c1c0e59bbbcc68a666d497a44071b05c7323",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "38b151bc6b4e9734c6fa41efb7a30343559400c0c75bbdd26bd8a968b8ea4592",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "a95569bcc0da6e715fd74783a6a1e0489e92b680c582f15dafbf98df589d505a",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/SKILL.md",
+                      "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/metadata.json",
+                      "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/execution-lifetime.md",
+                      "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/source-map.md",
+                      "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/validation.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/SKILL.md"
+                  }
+                ],
+                "content_bytes": 13683,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local native-backend, model, service or efficacy validation.",
+                  "Static content checks do not promote this candidate or establish installed API conformance.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/SKILL.md",
+                    "role": "core",
+                    "sha256": "a95569bcc0da6e715fd74783a6a1e0489e92b680c582f15dafbf98df589d505a"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/source-map.md",
+                    "role": "reference",
+                    "sha256": "798b9cb89ed6b885eb47464ca9f137d2802a54571b9327e631d7742d3e5abd5d"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/execution-lifetime.md",
+                    "role": "reference",
+                    "sha256": "c4bc41e08aca68f661325f61e6f0db1760a079dc362f295c847bdc374fee3bf6"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/validation.md",
+                    "role": "reference",
+                    "sha256": "ebfc00bf7d6ab0a1dc415953ae84d56ccdc713d5614aff85e433c4a843793bff"
+                  }
+                ],
+                "revision": "sha256:5de043dd66cc20bccdbb002ceee7efffef2136e4296e038406bf1e9884b1ec77",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "a95569bcc0da6e715fd74783a6a1e0489e92b680c582f15dafbf98df589d505a",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/SKILL.md",
+                      "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/metadata.json",
+                      "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/execution-lifetime.md",
+                      "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/source-map.md",
+                      "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/validation.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md"
+                  }
+                ],
+                "content_bytes": 13155,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; native inference, allocator behavior, image quality and performance were not run.",
+                  "Eighteen bounded readset paths plus one separately identity-checked README match the pinned Git commit/tree blob IDs and sizes (19 blobs total); header coverage is limited to selected owner/state declarations, and this does not authenticate every archive member or submodule contents.",
+                  "GGML and other submodules are not expanded; matching private GGML headers are required even for a system library.",
+                  "CPU bypasses the inspected accelerator capacity gate and segmentation eligibility, so CPU checks cannot qualify accelerator memory budgets or prefetch.",
+                  "Video/audio pipelines, detailed model kernels, approximate denoising caches and full multi-device execution remain outside this readset.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json",
+                    "reason": "not_context_resource"
+                  },
+                  {
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md",
+                    "reason": "context_limit"
+                  },
+                  {
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md",
+                    "reason": "context_limit"
+                  },
+                  {
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json",
+                    "reason": "context_limit"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md",
+                    "role": "core",
+                    "sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "b67affbd7d237d82b7f51be49f1537e9820fe56ec70a71ee76809f9e599ca693"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE",
+                    "role": "reference",
+                    "sha256": "b53fa08f515cb5a6fff7b9fd8fcd0961a4b80df29d74372d25e1f9171aa042ee"
+                  }
+                ],
+                "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 454958,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_device_health",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascendc_tiling_design",
+                "ref": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "resource_count": 343,
+                "resource_hashes_sha256": "caa4787f895186508b8b6f863b2d03a830364585ffd65f0c098be319c7d54cad",
+                "revision": "sha256:9c90d2bc30d84cf0c99a0382d1e61b01c1210efe51d16ddb81c25afcb7c04e4b"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.assess_split_backward_schedule",
+                "ref": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "resource_count": 343,
+                "resource_hashes_sha256": "e3cbdc3297e17bf0af02a32b0fd231cb1f645fd7d21844b7764cd1b61cdc9243",
+                "revision": "sha256:cec60c01e6c992705c02cc68ab17a2a437ecf2cfbd03489b45fc98ac7692b044"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_operator_tuning",
+                "ref": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "resource_count": 343,
+                "resource_hashes_sha256": "1f4ba53716e4b333251713ef826742cb9596f086412d26e2715a3e71c75d6211",
+                "revision": "sha256:01c26a7a1203f0d8fea8e035e875c1c0e59bbbcc68a666d497a44071b05c7323"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "ref": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "resource_count": 343,
+                "resource_hashes_sha256": "7b4026d946b56b7fdafc6e086144d5ab04f6951f233afd5cfbcf003c04d05287",
+                "revision": "sha256:5de043dd66cc20bccdbb002ceee7efffef2136e4296e038406bf1e9884b1ec77"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sd_cpp_residency",
+                "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c83d7a76dffa5a3d97e932a3eac4602ad82e20babb09d9942504fff6b4db53d6",
+                "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde"
+              }
+            ],
+            "selection": {
+              "cue": "UB split buffer branch matrix",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 543313,
+            "sha256": "9176aff7fe535f60d6ba945a1c2f49d18d9699b455ae77e26cd5e921fbe676a8"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev07",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 85270,
+            "no_match": false,
+            "omissions": [],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascendc_tiling_design",
+                "ref": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "resource_count": 343,
+                "resource_hashes_sha256": "caa4787f895186508b8b6f863b2d03a830364585ffd65f0c098be319c7d54cad",
+                "revision": "sha256:9c90d2bc30d84cf0c99a0382d1e61b01c1210efe51d16ddb81c25afcb7c04e4b"
+              }
+            ],
+            "serialized_bytes": 85587,
+            "sha256": "ae41b0ffbd7acbffb95d3f5bfdc30e5102d9dce1864073f693da14604f50c774"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev08",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev08",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 85600,
+            "no_match": false,
+            "omissions": [],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "ref": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "resource_count": 343,
+                "resource_hashes_sha256": "3f8f5e717bf501c4d94bfb3ac40c15393b07dd519c9170343aeb54445fbe933c",
+                "revision": "sha256:9028b6d32dcbe068af6fc777e80162bc815835f63657dc536f93d9e8ce5599f3"
+              }
+            ],
+            "selection": {
+              "cue": "multi_stream_lazy_reclaim",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 86057,
+            "sha256": "8e4a81b12208ac5ede8768f9ae00f5ca91e9355d2aac306524106007c18e98c0"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev08",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 17459,
+            "context_omissions": [],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "8428df7157af024fa17cfe0b3818ed54bd9f468306dda46eb8d4f0edf4f3f230",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/metadata.json",
+                      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md"
+                  }
+                ],
+                "content_bytes": 17459,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md",
+                    "role": "core",
+                    "sha256": "8428df7157af024fa17cfe0b3818ed54bd9f468306dda46eb8d4f0edf4f3f230"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "80cba4d3a884e7ee2618326cf74b9f9c3a34e4d9cf443707f6bbe673f2d5ba65"
+                  }
+                ],
+                "revision": "sha256:9028b6d32dcbe068af6fc777e80162bc815835f63657dc536f93d9e8ce5599f3",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "8428df7157af024fa17cfe0b3818ed54bd9f468306dda46eb8d4f0edf4f3f230",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/metadata.json",
+                      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 89687,
+            "no_match": false,
+            "omissions": [],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "ref": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "resource_count": 343,
+                "resource_hashes_sha256": "3f8f5e717bf501c4d94bfb3ac40c15393b07dd519c9170343aeb54445fbe933c",
+                "revision": "sha256:9028b6d32dcbe068af6fc777e80162bc815835f63657dc536f93d9e8ce5599f3"
+              }
+            ],
+            "selection": {
+              "cue": "multi_stream_lazy_reclaim",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 111101,
+            "sha256": "4fd1f7e989385f0ed48d163789ab382436741d2bfb78a7595f81d75b66950330"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev08",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev09",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev09",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 432011,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.kv_tier_management",
+                "ref": "package:inferrsi.common.skills.kv_tier_management",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ee42b55fa8a404149ecd1dccc32a892566ee8867710531bc0af553409758a019",
+                "revision": "sha256:5816c2e526a71152486af2230ace24d27b54baeb373bf5d27e353ccd44f640aa"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.bound_windowed_attention_replay",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb70e2e746f7ba833a696431c6a4deda830188acf5c4eafce6b0008818f8190a",
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sd_cpp_residency",
+                "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c83d7a76dffa5a3d97e932a3eac4602ad82e20babb09d9942504fff6b4db53d6",
+                "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.diffusers_invocation_state",
+                "ref": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "resource_count": 343,
+                "resource_hashes_sha256": "64b10a60e44276bc4a20b3441f65afd929341e8558fde32491ba27a20fde4020",
+                "revision": "sha256:a8529f2c6bf5f84ce48ba0d6426f1bf1bc6fce172cd0a599b6a149878e3b4be4"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.generation_cache_lifecycle",
+                "ref": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "resource_count": 343,
+                "resource_hashes_sha256": "052e200da2a160661cf44647685c70e52feec4b72fc0f2e7b34a10ce54710e03",
+                "revision": "sha256:22411ef79b30ddd4f8b01a7ed5d00fc4fc75a3c9f1e4d31f56f2e440d9b09d2f"
+              }
+            ],
+            "selection": {
+              "cue": "LMCache skip_l1 retain",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 435629,
+            "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev09",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 61835,
+            "context_omissions": [
+              {
+                "available_bytes": 3701,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "required_bytes": 10129
+              },
+              {
+                "available_bytes": 3701,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "required_bytes": 10146
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md"
+                  }
+                ],
+                "content_bytes": 10812,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/repositories/lmcache/kv-tier-management/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.kv_tier_management",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md",
+                    "role": "core",
+                    "sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27"
+                  }
+                ],
+                "revision": "sha256:5816c2e526a71152486af2230ace24d27b54baeb373bf5d27e353ccd44f640aa",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "4703beb6e5194ef2ff2c0b5ec69f64e35417a4bc0dfb50a10c80a920b77ef3b5",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/validation.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md"
+                  }
+                ],
+                "content_bytes": 29732,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no equivalence run was performed.",
+                  "Applies to architectures with a demonstrated boundary dependency, not to arbitrary sliding-window stacks.",
+                  "Window sizes, boundary-layer index and ring lengths are model-config specific and are not transferable.",
+                  "Query-row compaction does not imply dropping historical keys; the two decisions are separate.",
+                  "Prompt logprobs are unrecoverable for the clipped region; the reference forces them off with a warning.",
+                  "The truncation path is off by default in the reference source.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md",
+                    "role": "core",
+                    "sha256": "4703beb6e5194ef2ff2c0b5ec69f64e35417a4bc0dfb50a10c80a920b77ef3b5"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "fd917bd5ac80bee9c9f4525f8e0300f6d67dc144e2a16632af42a67f3d53d098"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/validation.md",
+                    "role": "reference",
+                    "sha256": "60e71e70c5519bcdf6c171eec2be2f342e8801bedd746d6005f11adc646dfaff"
+                  }
+                ],
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "4703beb6e5194ef2ff2c0b5ec69f64e35417a4bc0dfb50a10c80a920b77ef3b5",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/validation.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md"
+                  }
+                ],
+                "content_bytes": 21291,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; native inference, allocator behavior, image quality and performance were not run.",
+                  "Eighteen bounded readset paths plus one separately identity-checked README match the pinned Git commit/tree blob IDs and sizes (19 blobs total); header coverage is limited to selected owner/state declarations, and this does not authenticate every archive member or submodule contents.",
+                  "GGML and other submodules are not expanded; matching private GGML headers are required even for a system library.",
+                  "CPU bypasses the inspected accelerator capacity gate and segmentation eligibility, so CPU checks cannot qualify accelerator memory budgets or prefetch.",
+                  "Video/audio pipelines, detailed model kernels, approximate denoising caches and full multi-device execution remain outside this readset.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json",
+                    "reason": "not_context_resource"
+                  },
+                  {
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json",
+                    "reason": "not_relevant"
+                  },
+                  {
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md",
+                    "reason": "not_relevant"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md",
+                    "role": "core",
+                    "sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "b67affbd7d237d82b7f51be49f1537e9820fe56ec70a71ee76809f9e599ca693"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE",
+                    "role": "reference",
+                    "sha256": "b53fa08f515cb5a6fff7b9fd8fcd0961a4b80df29d74372d25e1f9171aa042ee"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md",
+                    "role": "reference",
+                    "sha256": "7b537cb972347dd68e303960f142892346d8b19b884e4d08016ea36a23bdaa7a"
+                  }
+                ],
+                "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 462705,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.kv_tier_management",
+                "ref": "package:inferrsi.common.skills.kv_tier_management",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ee42b55fa8a404149ecd1dccc32a892566ee8867710531bc0af553409758a019",
+                "revision": "sha256:5816c2e526a71152486af2230ace24d27b54baeb373bf5d27e353ccd44f640aa"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.bound_windowed_attention_replay",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb70e2e746f7ba833a696431c6a4deda830188acf5c4eafce6b0008818f8190a",
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sd_cpp_residency",
+                "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c83d7a76dffa5a3d97e932a3eac4602ad82e20babb09d9942504fff6b4db53d6",
+                "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.diffusers_invocation_state",
+                "ref": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "resource_count": 343,
+                "resource_hashes_sha256": "64b10a60e44276bc4a20b3441f65afd929341e8558fde32491ba27a20fde4020",
+                "revision": "sha256:a8529f2c6bf5f84ce48ba0d6426f1bf1bc6fce172cd0a599b6a149878e3b4be4"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.generation_cache_lifecycle",
+                "ref": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "resource_count": 343,
+                "resource_hashes_sha256": "052e200da2a160661cf44647685c70e52feec4b72fc0f2e7b34a10ce54710e03",
+                "revision": "sha256:22411ef79b30ddd4f8b01a7ed5d00fc4fc75a3c9f1e4d31f56f2e440d9b09d2f"
+              }
+            ],
+            "selection": {
+              "cue": "LMCache skip_l1 retain",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 541917,
+            "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev09",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev10",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev10",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 85740,
+            "no_match": false,
+            "omissions": [],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.checkpoint_recovery_consistency",
+                "ref": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "15c906f896eab2a681eb9bc9e42ee29d11b0594a1e1aeacbf6f414233209d1fa",
+                "revision": "sha256:73f77f89e298ce8724258cb3d74315438c036297bb38c63aac3d7dc95473ce53"
+              }
+            ],
+            "selection": {
+              "cue": "find_latest",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 86183,
+            "sha256": "5ba7d7ecbdb98e25ad2e25f1ce1c8764cef738f47d5657fc9e46a655b886b8a8"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev10",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 14400,
+            "context_omissions": [],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "6b7db1eb2780dec27858cb7aa0a3c093c5b5dfa4c30261aebfeaa255213ff688",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/references/dlrover-persistence-replicas.md",
+                      "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/SKILL.md"
+                  }
+                ],
+                "content_bytes": 14400,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/checkpoint-recovery-consistency/metadata.json",
+                    "reason": "not_context_resource"
+                  },
+                  {
+                    "path": "data/skill_library/methods/checkpoint-recovery-consistency/references/dlrover-persistence-replicas.md",
+                    "reason": "not_relevant"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/checkpoint-recovery-consistency/SKILL.md",
+                    "role": "core",
+                    "sha256": "6b7db1eb2780dec27858cb7aa0a3c093c5b5dfa4c30261aebfeaa255213ff688"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/checkpoint-recovery-consistency/references/sources.md",
+                    "role": "reference",
+                    "sha256": "9425b1aab963a2c8d35f3db897bd15ee69699bf5b0d0c8425c3c8bba3e9a0233"
+                  }
+                ],
+                "revision": "sha256:73f77f89e298ce8724258cb3d74315438c036297bb38c63aac3d7dc95473ce53",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "6b7db1eb2780dec27858cb7aa0a3c093c5b5dfa4c30261aebfeaa255213ff688",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/references/dlrover-persistence-replicas.md",
+                      "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 90211,
+            "no_match": false,
+            "omissions": [],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.checkpoint_recovery_consistency",
+                "ref": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "15c906f896eab2a681eb9bc9e42ee29d11b0594a1e1aeacbf6f414233209d1fa",
+                "revision": "sha256:73f77f89e298ce8724258cb3d74315438c036297bb38c63aac3d7dc95473ce53"
+              }
+            ],
+            "selection": {
+              "cue": "find_latest",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 108906,
+            "sha256": "38c065f634ce4b1297325fc9ccbfff2dac832b28624e31e464ee6995c2b89d10"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev10",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev11",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev11",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 436986,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "ref": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "resource_count": 343,
+                "resource_hashes_sha256": "db84ef8ccf72ef81606670b384a3814297f9c999a5ffa25b05b4f0d10c500ebd",
+                "revision": "sha256:d8a8775c87f022125906e0ec25c0c57a083e272be5b339d5cdeb7a1704c0e4db"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.cuda.skills.reason_hopper_kernels",
+                "ref": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "resource_count": 343,
+                "resource_hashes_sha256": "4dfd3cbff82440ce22e8f16b14e6e75d25a34144661b077e21aed02a856de902",
+                "revision": "sha256:4d0628c3fd9cbcc647ec30d922166044f3544326f54f793779692d731e2a5b77"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.bound_windowed_attention_replay",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb70e2e746f7ba833a696431c6a4deda830188acf5c4eafce6b0008818f8190a",
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_llama_runtime_state",
+                "ref": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "resource_count": 343,
+                "resource_hashes_sha256": "03e86bd3af77f17d6c7ad2933534524ab7fe5993396a7fe9748a6014496a34b5",
+                "revision": "sha256:a3c3d3244cda589e8e7809e56f43b7bfff1852def8dd26fb03a9d29e39bca3dd"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.npu_skill_factory",
+                "ref": "package:inferrsi.common.skills.npu_skill_factory",
+                "resource_count": 343,
+                "resource_hashes_sha256": "7cde43221171eca8e11527545216e348d8c464438bec7458aab055843bb440cf",
+                "revision": "sha256:c789873dd4195bee85724a481228f51852a09c870d630ad1e828901037f875b7"
+              }
+            ],
+            "selection": {
+              "cue": "Why can larger attention tiles cause register spills and lower parallelism?",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 441258,
+            "sha256": "1fd164121d0001fd73173e443cca13d1fa3b40f69b4b39922e0d7c1459331cd3"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev11",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 56835,
+            "context_omissions": [
+              {
+                "available_bytes": 8701,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "required_bytes": 154821
+              },
+              {
+                "available_bytes": 8701,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "required_bytes": 29732
+              },
+              {
+                "available_bytes": 8701,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "required_bytes": 24572
+              },
+              {
+                "available_bytes": 8701,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.npu_skill_factory",
+                "required_bytes": 65054
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "a25738dbb3b9daa15c5827cb536a20fe13f7c8b9dd14aa3fb7cfa7e132879431",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/metadata.json",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/mechanisms.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/INTEGRATION_NOTES.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/NEXT_GATE_LEADS.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/README.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/io-reduction-can-plateau.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/overlap-and-fp8-counter-signals.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/partitioning-after-fusion.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/knowledge/attention-cost-model-boundaries.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/SKILL.md"
+                  }
+                ],
+                "content_bytes": 56835,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/SKILL.md",
+                    "role": "core",
+                    "sha256": "a25738dbb3b9daa15c5827cb536a20fe13f7c8b9dd14aa3fb7cfa7e132879431"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/mechanisms.md",
+                    "role": "reference",
+                    "sha256": "fe08b2292fa587be3874774ddf41063b84c579a0085df422ec6710d8aa90b8b7"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/io-reduction-can-plateau.md",
+                    "role": "reference",
+                    "sha256": "5cc35f514224ac6e6efd2da6db221bba468b2f4db66807295e67b240bcdfdba5"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/overlap-and-fp8-counter-signals.md",
+                    "role": "reference",
+                    "sha256": "80c8a60a6df55f3d84fab23e03567d8b38e23bcdbd63cc09881e46db1e2888ab"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/partitioning-after-fusion.md",
+                    "role": "reference",
+                    "sha256": "4d4cabe0991eab554dcfbd4fa9c8199e61fa4767709b0c4cd646e2588bb398a2"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/INTEGRATION_NOTES.md",
+                    "role": "reference",
+                    "sha256": "371b01ee201ef51fecbfd6da892d517aaba295930b9395025f03f34e21c030a1"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/knowledge/attention-cost-model-boundaries.md",
+                    "role": "reference",
+                    "sha256": "5ed8f888443e38ccb72545b6025447aab7b92936f78c151fed680e582126239b"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/NEXT_GATE_LEADS.md",
+                    "role": "reference",
+                    "sha256": "35b8b58ae5db02b2df51ccbe5cdf870a7383d17f345f9ee5eb7b0a4146dfc74b"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/README.md",
+                    "role": "reference",
+                    "sha256": "e0b1ec1b39c10067f5ddce58320851ad089568c4b2296a24777389097cb5b388"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/sources.md",
+                    "role": "reference",
+                    "sha256": "53436e0de4e6dda2d29426591bf61f8526378cdc6f9ec7967b88b1bcd4c61e6f"
+                  }
+                ],
+                "revision": "sha256:d8a8775c87f022125906e0ec25c0c57a083e272be5b339d5cdeb7a1704c0e4db",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "a25738dbb3b9daa15c5827cb536a20fe13f7c8b9dd14aa3fb7cfa7e132879431",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/metadata.json",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/mechanisms.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/INTEGRATION_NOTES.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/NEXT_GATE_LEADS.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/README.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/io-reduction-can-plateau.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/overlap-and-fp8-counter-signals.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/partitioning-after-fusion.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/knowledge/attention-cost-model-boundaries.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 481490,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "ref": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "resource_count": 343,
+                "resource_hashes_sha256": "db84ef8ccf72ef81606670b384a3814297f9c999a5ffa25b05b4f0d10c500ebd",
+                "revision": "sha256:d8a8775c87f022125906e0ec25c0c57a083e272be5b339d5cdeb7a1704c0e4db"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.cuda.skills.reason_hopper_kernels",
+                "ref": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "resource_count": 343,
+                "resource_hashes_sha256": "4dfd3cbff82440ce22e8f16b14e6e75d25a34144661b077e21aed02a856de902",
+                "revision": "sha256:4d0628c3fd9cbcc647ec30d922166044f3544326f54f793779692d731e2a5b77"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.bound_windowed_attention_replay",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb70e2e746f7ba833a696431c6a4deda830188acf5c4eafce6b0008818f8190a",
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_llama_runtime_state",
+                "ref": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "resource_count": 343,
+                "resource_hashes_sha256": "03e86bd3af77f17d6c7ad2933534524ab7fe5993396a7fe9748a6014496a34b5",
+                "revision": "sha256:a3c3d3244cda589e8e7809e56f43b7bfff1852def8dd26fb03a9d29e39bca3dd"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.npu_skill_factory",
+                "ref": "package:inferrsi.common.skills.npu_skill_factory",
+                "resource_count": 343,
+                "resource_hashes_sha256": "7cde43221171eca8e11527545216e348d8c464438bec7458aab055843bb440cf",
+                "revision": "sha256:c789873dd4195bee85724a481228f51852a09c870d630ad1e828901037f875b7"
+              }
+            ],
+            "selection": {
+              "cue": "Why can larger attention tiles cause register spills and lower parallelism?",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 551101,
+            "sha256": "1d643e68fc4d9833f93a977651a8adccaac197560031521273be3068a034f603"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev11",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev12",
+          "mode": "search",
+          "output": {
+            "ordered_items": [
+              {
+                "id": "inferrsi.common.skills.os_process_placement"
+              }
+            ],
+            "serialized_bytes": 5621,
+            "sha256": "a9bee665f4da822ce6780d24bea786b256567b5e4b9ee6dfdae85de05c2e61e7"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev12",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 430402,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.os_process_placement",
+                "ref": "package:inferrsi.common.skills.os_process_placement",
+                "resource_count": 343,
+                "resource_hashes_sha256": "d0743478a2fb19183d4c1a8ab0cfcebf02a89512bbc35e09ade8f120a1e53afd",
+                "revision": "sha256:3ae1cc0f7d372b46a2e0a697a4485d63ca84d40afdba5414e9c6c1f2db15db7e"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.training_resource_budget",
+                "ref": "package:inferrsi.common.skills.training_resource_budget",
+                "resource_count": 343,
+                "resource_hashes_sha256": "846f754eb2d0d8f76837d8a532ebaed8f53bbbdb5f94695ebf7725de7a951497",
+                "revision": "sha256:0121883f97748a7cab0a53e5831988e4f34ca3d841987598f98d75506cad0734"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.host_memory_management",
+                "ref": "package:inferrsi.common.skills.host_memory_management",
+                "resource_count": 343,
+                "resource_hashes_sha256": "704c39f2041e99d4d66d9e4464c116274638e246db6a4b56e7c447dd711c198f",
+                "revision": "sha256:96c29b3da68c9493ce25421640f23cc2c15258005b6e006b6e9c2af285624fc3"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.cgroup_resource_controls",
+                "ref": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "resource_count": 343,
+                "resource_hashes_sha256": "16e727f435ae4d8b6f349404846e7106e958b83425751c86135d63770802114c",
+                "revision": "sha256:94ef2d4108261a8a911ec5ffc89edcdd6d7bda21f064f7c66701a8f663cdfea3"
+              }
+            ],
+            "selection": {
+              "cue": "NUMA CPU affinity placement",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 433035,
+            "sha256": "f0f0732feaf4882221188f4cbd4aca31c84af3f6f53bf6d86aad09f939315d1e"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev12",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 65309,
+            "context_omissions": [
+              {
+                "available_bytes": 48592,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.training_resource_budget",
+                "required_bytes": 74071
+              },
+              {
+                "available_bytes": 48592,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.host_memory_management",
+                "required_bytes": 49803
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "2003492e4d89fde0443b180d74933e764bd1979435e109c8d26e20eb8f6e5053",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/host/os-process-placement/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/host/os-process-placement/metadata.json",
+                      "src/inferrsi/data/skill_library/hardware/host/os-process-placement/references/llama-numa-thread-readset.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/host/os-process-placement/SKILL.md"
+                  }
+                ],
+                "content_bytes": 16944,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Only the existing llama.cpp NUMA/thread slice is normalized at 3e390af18de1131dda27913c00511725b4274c8a; other original sources remain unnormalized, with their identities and limits retained in the procedure/readsets.",
+                  "Linux NUMA source paths and separate configuration controls do not establish universally optimal thread counts, affinity combinations or measured placement gains.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/host/os-process-placement/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.os_process_placement",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/host/os-process-placement/SKILL.md",
+                    "role": "core",
+                    "sha256": "2003492e4d89fde0443b180d74933e764bd1979435e109c8d26e20eb8f6e5053"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/host/os-process-placement/references/llama-numa-thread-readset.md",
+                    "role": "reference",
+                    "sha256": "748b254b3b043a5ddb0f480eb4a566854d27eedd7b111499863e9bf0a5a6951e"
+                  }
+                ],
+                "revision": "sha256:3ae1cc0f7d372b46a2e0a697a4485d63ca84d40afdba5414e9c6c1f2db15db7e",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "2003492e4d89fde0443b180d74933e764bd1979435e109c8d26e20eb8f6e5053",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/host/os-process-placement/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/host/os-process-placement/metadata.json",
+                      "src/inferrsi/data/skill_library/hardware/host/os-process-placement/references/llama-numa-thread-readset.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/host/os-process-placement/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md"
+                  }
+                ],
+                "content_bytes": 41017,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Normalized provenance covers the existing llama.cpp loading slice at 3e390af18de1131dda27913c00511725b4274c8a and the fastsafetensors distributed-delivery slice at 219e2561a41ede2d5d8eba611c248bb0effbf580; other inherited claims remain unnormalized with original identities and limits retained.",
+                  "The readsets record static mechanisms and source recommendations, not validated startup, memory, placement or backend performance.",
+                  "The fastsafetensors delivery readset does not revalidate inherited I/O/save claims or qualify uneven scatter, accelerator/NPU compatibility, inter-node transport or agent efficacy.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/weight-loading-optimization/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                    "role": "core",
+                    "sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                    "role": "reference",
+                    "sha256": "6c640f28ee3f0253a17b81461c865cb9f947367002cb75bf72550c4572d02e97"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                    "role": "reference",
+                    "sha256": "5a39f14b102b0cee304541e011aa31e44e6b13527d1bee9987c3f3b66c064d3d"
+                  }
+                ],
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "fd1f5a7e6a55714730044155f5b38c1d700383d4e16c3e91963bbfbe1c773af5",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/SKILL.md"
+                  }
+                ],
+                "content_bytes": 7348,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/host/cgroup-resource-controls/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/host/cgroup-resource-controls/SKILL.md",
+                    "role": "core",
+                    "sha256": "fd1f5a7e6a55714730044155f5b38c1d700383d4e16c3e91963bbfbe1c773af5"
+                  }
+                ],
+                "revision": "sha256:94ef2d4108261a8a911ec5ffc89edcdd6d7bda21f064f7c66701a8f663cdfea3",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "fd1f5a7e6a55714730044155f5b38c1d700383d4e16c3e91963bbfbe1c773af5",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 460966,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.os_process_placement",
+                "ref": "package:inferrsi.common.skills.os_process_placement",
+                "resource_count": 343,
+                "resource_hashes_sha256": "d0743478a2fb19183d4c1a8ab0cfcebf02a89512bbc35e09ade8f120a1e53afd",
+                "revision": "sha256:3ae1cc0f7d372b46a2e0a697a4485d63ca84d40afdba5414e9c6c1f2db15db7e"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.training_resource_budget",
+                "ref": "package:inferrsi.common.skills.training_resource_budget",
+                "resource_count": 343,
+                "resource_hashes_sha256": "846f754eb2d0d8f76837d8a532ebaed8f53bbbdb5f94695ebf7725de7a951497",
+                "revision": "sha256:0121883f97748a7cab0a53e5831988e4f34ca3d841987598f98d75506cad0734"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.host_memory_management",
+                "ref": "package:inferrsi.common.skills.host_memory_management",
+                "resource_count": 343,
+                "resource_hashes_sha256": "704c39f2041e99d4d66d9e4464c116274638e246db6a4b56e7c447dd711c198f",
+                "revision": "sha256:96c29b3da68c9493ce25421640f23cc2c15258005b6e006b6e9c2af285624fc3"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.cgroup_resource_controls",
+                "ref": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "resource_count": 343,
+                "resource_hashes_sha256": "16e727f435ae4d8b6f349404846e7106e958b83425751c86135d63770802114c",
+                "revision": "sha256:94ef2d4108261a8a911ec5ffc89edcdd6d7bda21f064f7c66701a8f663cdfea3"
+              }
+            ],
+            "selection": {
+              "cue": "NUMA CPU affinity placement",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 540415,
+            "sha256": "fa83fe34560e480806391c460c276add522b96e93b1c95b26548e0cc0d81b5dc"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev12",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 85733,
+            "no_match": false,
+            "omissions": [],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.os_process_placement",
+                "ref": "package:inferrsi.common.skills.os_process_placement",
+                "resource_count": 343,
+                "resource_hashes_sha256": "d0743478a2fb19183d4c1a8ab0cfcebf02a89512bbc35e09ade8f120a1e53afd",
+                "revision": "sha256:3ae1cc0f7d372b46a2e0a697a4485d63ca84d40afdba5414e9c6c1f2db15db7e"
+              }
+            ],
+            "serialized_bytes": 86050,
+            "sha256": "936169ae69e7b168b2100db3e6913d2034ff12372bdfffbbcf3999b8d9e1e69e"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev13",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev13",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 433439,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.plan_activation_residency",
+                "ref": "package:inferrsi.common.skills.plan_activation_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "8141a002cba8cd3ffa905e29c6517fd935740140b897d2f96b65c724ac3d24b9",
+                "revision": "sha256:06df72aed0c6246c334dada5e335b85e49111753637f4fffa4ecb68da5ec3d43"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sd_cpp_residency",
+                "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c83d7a76dffa5a3d97e932a3eac4602ad82e20babb09d9942504fff6b4db53d6",
+                "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_h3_runtime_residency",
+                "ref": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "64ea29a64c41b0dac809d22e10aa650ac7d60cb008ab15d9e702161ce2ad63c2",
+                "revision": "sha256:601645ce381756514c931d83e0019b1901272e08c9fd4e0ae99318ef0d18405b"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.comfy_ui",
+                "ref": "package:inferrsi.common.skills.comfy_ui",
+                "resource_count": 343,
+                "resource_hashes_sha256": "66d8cb95bd1f2270900a3a81e6274cb2cc3c799229acc75d0676def6ef22f8a8",
+                "revision": "sha256:13e038ab45a19c1ceafc282b3504ddb63bba487b6a2669472f6f580d18ff6599"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "ref": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "resource_count": 343,
+                "resource_hashes_sha256": "8af3478b2f59c5fade35c614a9bfe4e4d53e8d5e9a5961870e8f4e62631cb937",
+                "revision": "sha256:0af6ac4cfd9d26bd90eeba094c3ceeed80e418dbf1872e6b1ffaa23b6b30e397"
+              }
+            ],
+            "selection": {
+              "cue": "gradient activation recomputation residency",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 435641,
+            "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev13",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 45716,
+            "context_omissions": [
+              {
+                "available_bytes": 19820,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "required_bytes": 36142
+              },
+              {
+                "available_bytes": 19820,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.comfy_ui",
+                "required_bytes": 45304
+              },
+              {
+                "available_bytes": 19820,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "required_bytes": 47942
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "542d8a1e6963f5019986e1fbd79d3a247a045a51487c09313026087ac457c15e",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/plan-activation-residency/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/plan-activation-residency/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/plan-activation-residency/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/plan-activation-residency/SKILL.md"
+                  }
+                ],
+                "content_bytes": 17414,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/plan-activation-residency/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.plan_activation_residency",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/plan-activation-residency/SKILL.md",
+                    "role": "core",
+                    "sha256": "542d8a1e6963f5019986e1fbd79d3a247a045a51487c09313026087ac457c15e"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/plan-activation-residency/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "5ff52655cfd06fcf970ed2a7d4d4fe1c31115b102af2d41497de0c74848ed2b3"
+                  }
+                ],
+                "revision": "sha256:06df72aed0c6246c334dada5e335b85e49111753637f4fffa4ecb68da5ec3d43",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "542d8a1e6963f5019986e1fbd79d3a247a045a51487c09313026087ac457c15e",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/plan-activation-residency/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/plan-activation-residency/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/plan-activation-residency/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/plan-activation-residency/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md"
+                  }
+                ],
+                "content_bytes": 28302,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; native inference, allocator behavior, image quality and performance were not run.",
+                  "Eighteen bounded readset paths plus one separately identity-checked README match the pinned Git commit/tree blob IDs and sizes (19 blobs total); header coverage is limited to selected owner/state declarations, and this does not authenticate every archive member or submodule contents.",
+                  "GGML and other submodules are not expanded; matching private GGML headers are required even for a system library.",
+                  "CPU bypasses the inspected accelerator capacity gate and segmentation eligibility, so CPU checks cannot qualify accelerator memory budgets or prefetch.",
+                  "Video/audio pipelines, detailed model kernels, approximate denoising caches and full multi-device execution remain outside this readset.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md",
+                    "role": "core",
+                    "sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "b67affbd7d237d82b7f51be49f1537e9820fe56ec70a71ee76809f9e599ca693"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE",
+                    "role": "reference",
+                    "sha256": "b53fa08f515cb5a6fff7b9fd8fcd0961a4b80df29d74372d25e1f9171aa042ee"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md",
+                    "role": "reference",
+                    "sha256": "7b537cb972347dd68e303960f142892346d8b19b884e4d08016ea36a23bdaa7a"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json",
+                    "role": "reference",
+                    "sha256": "af4830370b6eddc2d176a9ae14da4ab56a90b0fe048a01a89ee051df4365d8e9"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md",
+                    "role": "reference",
+                    "sha256": "5dcf5fa7bf734fbe13214f57d2dfb40ff8215d9cd34e072a9e615bf103b82fc2"
+                  }
+                ],
+                "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 469025,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.plan_activation_residency",
+                "ref": "package:inferrsi.common.skills.plan_activation_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "8141a002cba8cd3ffa905e29c6517fd935740140b897d2f96b65c724ac3d24b9",
+                "revision": "sha256:06df72aed0c6246c334dada5e335b85e49111753637f4fffa4ecb68da5ec3d43"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sd_cpp_residency",
+                "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c83d7a76dffa5a3d97e932a3eac4602ad82e20babb09d9942504fff6b4db53d6",
+                "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_h3_runtime_residency",
+                "ref": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "64ea29a64c41b0dac809d22e10aa650ac7d60cb008ab15d9e702161ce2ad63c2",
+                "revision": "sha256:601645ce381756514c931d83e0019b1901272e08c9fd4e0ae99318ef0d18405b"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.comfy_ui",
+                "ref": "package:inferrsi.common.skills.comfy_ui",
+                "resource_count": 343,
+                "resource_hashes_sha256": "66d8cb95bd1f2270900a3a81e6274cb2cc3c799229acc75d0676def6ef22f8a8",
+                "revision": "sha256:13e038ab45a19c1ceafc282b3504ddb63bba487b6a2669472f6f580d18ff6599"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "ref": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "resource_count": 343,
+                "resource_hashes_sha256": "8af3478b2f59c5fade35c614a9bfe4e4d53e8d5e9a5961870e8f4e62631cb937",
+                "revision": "sha256:0af6ac4cfd9d26bd90eeba094c3ceeed80e418dbf1872e6b1ffaa23b6b30e397"
+              }
+            ],
+            "selection": {
+              "cue": "gradient activation recomputation residency",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 527283,
+            "sha256": "a6c8fd92fd24a7304d4b207634068866b654d2e6d395240dae1bc74eccb9800d"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev13",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev14",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev14",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 432303,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_device_health",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_performance_analysis",
+                "ref": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "resource_count": 343,
+                "resource_hashes_sha256": "255c964b950c3e58c7a55001c35a22d48e386da437cea66ec2486a1f0412dd1f",
+                "revision": "sha256:f7ea7eb4d21d63a62a20b1a8b719277d4d2929907ff255af4acb1e30028b01dc"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.dense_performance_evidence",
+                "ref": "package:inferrsi.common.skills.dense_performance_evidence",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ee7e4bd7570b737eb14c98190545cee0fa11065e1b1a2d0cfe663928cfca0551",
+                "revision": "sha256:82e68013915ca052f96119da0a8d32d40d38d0aab792b04844c2e208d10706dc"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.performance_ceiling_analysis",
+                "ref": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c915247877787fcecc949d1d212ae1f18e3ba2a82f02b7c1cf343dedf0a7e885",
+                "revision": "sha256:6f67a5d347f3d05198f92cfce346d30051a544427ec8922b9de64432c3ffe316"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.claim_contract",
+                "ref": "package:inferrsi.common.skills.claim_contract",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c6481b673e503a8707f897b6e23c355888873b8818458e6a4b8e1fbfee9f9f8f",
+                "revision": "sha256:0f6955d51d86aa329bf643f37bf750084520c666360a64d6c10bfb678e64b8a2"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.inference_benchmark_design",
+                "ref": "package:inferrsi.common.skills.inference_benchmark_design",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb743a7fe0662feead981795dc60e2a5c9355ed876339f14ad7a47caab0bb670",
+                "revision": "sha256:6b0cbeaa5d409c04228883884e192f275cad4874c7f6b3039abef9fd443fafb5"
+              }
+            ],
+            "selection": {
+              "cue": "When does missing validator freeze make a performance claim inconclusive?",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 437750,
+            "sha256": "e074b2d4a400f80ae868b08968631f451801aa928ec7aec0cfe700f250ba4071"
+          },
+          "reciprocal_rank_at_5": 0.25,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev14",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 40799,
+            "context_omissions": [
+              {
+                "available_bytes": 25812,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "required_bytes": 170782
+              },
+              {
+                "available_bytes": 24737,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.inference_benchmark_design",
+                "required_bytes": 47014
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "a057b7f86a24f34a8f84abbbb4910d58d4e50ff47060e1c404600017e3febe89",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/metadata.json",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-CANN-2.0-LICENSE.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-stage-evidence.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/SKILL.md"
+                  }
+                ],
+                "content_bytes": 26931,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Only the incremental canonical CannBot fixed-stage evidence boundary is normalized; inherited MindIE-SD guidance remains unchanged.",
+                  "CannBot callback predicates do not establish numerical accuracy or performance; dependent skills, agent-host integration and NPU execution are unverified.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/ascend/ascend-performance-analysis/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/ascend-performance-analysis/SKILL.md",
+                    "role": "core",
+                    "sha256": "a057b7f86a24f34a8f84abbbb4910d58d4e50ff47060e1c404600017e3febe89"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-stage-evidence.md",
+                    "role": "reference",
+                    "sha256": "f56e4125325c9c1b9235a1d64fd732861c7686444492422daff44ba8dd648be4"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-CANN-2.0-LICENSE.md",
+                    "role": "reference",
+                    "sha256": "89b1c114df62cfd183004f38d18d07f186436c3b37203dd69c3e84becf9d2282"
+                  }
+                ],
+                "revision": "sha256:f7ea7eb4d21d63a62a20b1a8b719277d4d2929907ff255af4acb1e30028b01dc",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "a057b7f86a24f34a8f84abbbb4910d58d4e50ff47060e1c404600017e3febe89",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/metadata.json",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-CANN-2.0-LICENSE.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-stage-evidence.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "53cc1867a13f54d1006d50517dea8848869ca50df3f0cd610b2424cc8d18c933",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/dense-performance-evidence/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/dense-performance-evidence/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/dense-performance-evidence/SKILL.md"
+                  }
+                ],
+                "content_bytes": 12793,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/dense-performance-evidence/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.dense_performance_evidence",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/dense-performance-evidence/SKILL.md",
+                    "role": "core",
+                    "sha256": "53cc1867a13f54d1006d50517dea8848869ca50df3f0cd610b2424cc8d18c933"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/inference-optimization-casebook/cases/admission-drain-occupancy.md",
+                    "role": "reference",
+                    "sha256": "e88bbac5d253a8e7e3a3d5e21047f23f2970c11ac23887d4303ab94fca52dd27"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/inference-optimization-casebook/cases/admission-drain-occupancy.json",
+                    "role": "reference",
+                    "sha256": "3319e23bbe2bd40e2d45fb89dc87bdc92d4df2b5ddd8086fcf84813b43885b7c"
+                  }
+                ],
+                "revision": "sha256:82e68013915ca052f96119da0a8d32d40d38d0aab792b04844c2e208d10706dc",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "53cc1867a13f54d1006d50517dea8848869ca50df3f0cd610b2424cc8d18c933",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/dense-performance-evidence/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/dense-performance-evidence/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/dense-performance-evidence/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "3af1dbbe81d91cb05fc2ef15a2ae25d7592a08558e1a5535f49c52f33a488c55",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/skill-index.json",
+                      "src/inferrsi/data/skill_library/workflows/claim-contract/SKILL.md"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/workflows/claim-contract/SKILL.md"
+                  }
+                ],
+                "content_bytes": 1075,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is unspecified.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [],
+                "ref": "package:inferrsi.common.skills.claim_contract",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/workflows/claim-contract/SKILL.md",
+                    "role": "core",
+                    "sha256": "3af1dbbe81d91cb05fc2ef15a2ae25d7592a08558e1a5535f49c52f33a488c55"
+                  }
+                ],
+                "revision": "sha256:0f6955d51d86aa329bf643f37bf750084520c666360a64d6c10bfb678e64b8a2",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "3af1dbbe81d91cb05fc2ef15a2ae25d7592a08558e1a5535f49c52f33a488c55",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/skill-index.json",
+                      "src/inferrsi/data/skill_library/workflows/claim-contract/SKILL.md"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/workflows/claim-contract/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 467584,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_device_health",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_performance_analysis",
+                "ref": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "resource_count": 343,
+                "resource_hashes_sha256": "255c964b950c3e58c7a55001c35a22d48e386da437cea66ec2486a1f0412dd1f",
+                "revision": "sha256:f7ea7eb4d21d63a62a20b1a8b719277d4d2929907ff255af4acb1e30028b01dc"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.dense_performance_evidence",
+                "ref": "package:inferrsi.common.skills.dense_performance_evidence",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ee7e4bd7570b737eb14c98190545cee0fa11065e1b1a2d0cfe663928cfca0551",
+                "revision": "sha256:82e68013915ca052f96119da0a8d32d40d38d0aab792b04844c2e208d10706dc"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.performance_ceiling_analysis",
+                "ref": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c915247877787fcecc949d1d212ae1f18e3ba2a82f02b7c1cf343dedf0a7e885",
+                "revision": "sha256:6f67a5d347f3d05198f92cfce346d30051a544427ec8922b9de64432c3ffe316"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.claim_contract",
+                "ref": "package:inferrsi.common.skills.claim_contract",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c6481b673e503a8707f897b6e23c355888873b8818458e6a4b8e1fbfee9f9f8f",
+                "revision": "sha256:0f6955d51d86aa329bf643f37bf750084520c666360a64d6c10bfb678e64b8a2"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.inference_benchmark_design",
+                "ref": "package:inferrsi.common.skills.inference_benchmark_design",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb743a7fe0662feead981795dc60e2a5c9355ed876339f14ad7a47caab0bb670",
+                "revision": "sha256:6b0cbeaa5d409c04228883884e192f275cad4874c7f6b3039abef9fd443fafb5"
+              }
+            ],
+            "selection": {
+              "cue": "When does missing validator freeze make a performance claim inconclusive?",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 524270,
+            "sha256": "a2df470f45879e6dadaa8c1af7f4f25bf456365e4948e6921276bb8983de4564"
+          },
+          "reciprocal_rank_at_5": 0.25,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev14",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": false,
+          "case": "dev15",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": false,
+          "case": "dev15",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "selection": {
+              "cue": "quasarbanjo xenofloral v7",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 452,
+            "sha256": "b4e79e7856cb502b1735da83ea5cc08996bdeb3eff8c380192ea6710457471a1"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": false,
+          "case": "dev15",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 0,
+            "context_omissions": [],
+            "contexts": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "selection": {
+              "cue": "quasarbanjo xenofloral v7",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 636,
+            "sha256": "a1344ff20112717b0a2fd180c8399638fee996e40c1fbc27236cd2508010dcf2"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": false,
+          "case": "dev15",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": false,
+          "case": "dev16",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": false,
+          "case": "dev16",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 430698,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.assess_serving_preemption",
+                "ref": "package:inferrsi.common.skills.assess_serving_preemption",
+                "resource_count": 343,
+                "resource_hashes_sha256": "1514c158f4d1bbe57e52a20bb872edbd6c628c9fe34cf5018e0d9c4abcab1a55",
+                "revision": "sha256:11751aa0d285e7fcf21c1c0f4564c46a09683b123895cc95bc22a3eccaae6221"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_h3_runtime_residency",
+                "ref": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "64ea29a64c41b0dac809d22e10aa650ac7d60cb008ab15d9e702161ce2ad63c2",
+                "revision": "sha256:601645ce381756514c931d83e0019b1901272e08c9fd4e0ae99318ef0d18405b"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "ref": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "resource_count": 343,
+                "resource_hashes_sha256": "8af3478b2f59c5fade35c614a9bfe4e4d53e8d5e9a5961870e8f4e62631cb937",
+                "revision": "sha256:0af6ac4cfd9d26bd90eeba094c3ceeed80e418dbf1872e6b1ffaa23b6b30e397"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.draft_acceptance_issue",
+                "ref": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "resource_count": 343,
+                "resource_hashes_sha256": "251e45e8af28e24055a9df480c2d541a1a0d3be419e40b212a5aae5cd61b4761",
+                "revision": "sha256:9f179f995ef669f59414e778a2064a4eece7fda9571ae0c191274b4f718b6b2d"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.host_memory_management",
+                "ref": "package:inferrsi.common.skills.host_memory_management",
+                "resource_count": 343,
+                "resource_hashes_sha256": "704c39f2041e99d4d66d9e4464c116274638e246db6a4b56e7c447dd711c198f",
+                "revision": "sha256:96c29b3da68c9493ce25421640f23cc2c15258005b6e006b6e9c2af285624fc3"
+              }
+            ],
+            "selection": {
+              "cue": "sourdough levain hydration recipe",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 431437,
+            "sha256": "4d442f904cbce4a4764030832d10d4e1a091b349bf15a12a4d347121054a54ef"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": false,
+          "case": "dev16",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 38037,
+            "context_omissions": [
+              {
+                "available_bytes": 28729,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "required_bytes": 36142
+              },
+              {
+                "available_bytes": 28729,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "required_bytes": 47942
+              },
+              {
+                "available_bytes": 27499,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.host_memory_management",
+                "required_bytes": 49803
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "0a5f48d9352c5bf460775ef47f9136829978811f4a16e84983e299782cae305d",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/assess-serving-preemption/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/assess-serving-preemption/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md"
+                  }
+                ],
+                "content_bytes": 36807,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/assess-serving-preemption/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.assess_serving_preemption",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/assess-serving-preemption/SKILL.md",
+                    "role": "core",
+                    "sha256": "0a5f48d9352c5bf460775ef47f9136829978811f4a16e84983e299782cae305d"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/assess-serving-preemption/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "f29fcecfba0490008a354c11b42403983fe165e56ff907ad61cc6040d2b6407f"
+                  }
+                ],
+                "revision": "sha256:11751aa0d285e7fcf21c1c0f4564c46a09683b123895cc95bc22a3eccaae6221",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "0a5f48d9352c5bf460775ef47f9136829978811f4a16e84983e299782cae305d",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/assess-serving-preemption/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/assess-serving-preemption/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "ce983acfb7d7d3272775e7cf00c92e4ddadd97511b602575a76e5ab7ade01b56",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/skill-index.json",
+                      "src/inferrsi/data/skill_library/workflows/draft-acceptance-issue/SKILL.md"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/workflows/draft-acceptance-issue/SKILL.md"
+                  }
+                ],
+                "content_bytes": 1230,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is unspecified.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [],
+                "ref": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/workflows/draft-acceptance-issue/SKILL.md",
+                    "role": "core",
+                    "sha256": "ce983acfb7d7d3272775e7cf00c92e4ddadd97511b602575a76e5ab7ade01b56"
+                  }
+                ],
+                "revision": "sha256:9f179f995ef669f59414e778a2064a4eece7fda9571ae0c191274b4f718b6b2d",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "ce983acfb7d7d3272775e7cf00c92e4ddadd97511b602575a76e5ab7ade01b56",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/skill-index.json",
+                      "src/inferrsi/data/skill_library/workflows/draft-acceptance-issue/SKILL.md"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/workflows/draft-acceptance-issue/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 460977,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.assess_serving_preemption",
+                "ref": "package:inferrsi.common.skills.assess_serving_preemption",
+                "resource_count": 343,
+                "resource_hashes_sha256": "1514c158f4d1bbe57e52a20bb872edbd6c628c9fe34cf5018e0d9c4abcab1a55",
+                "revision": "sha256:11751aa0d285e7fcf21c1c0f4564c46a09683b123895cc95bc22a3eccaae6221"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_h3_runtime_residency",
+                "ref": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "64ea29a64c41b0dac809d22e10aa650ac7d60cb008ab15d9e702161ce2ad63c2",
+                "revision": "sha256:601645ce381756514c931d83e0019b1901272e08c9fd4e0ae99318ef0d18405b"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "ref": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "resource_count": 343,
+                "resource_hashes_sha256": "8af3478b2f59c5fade35c614a9bfe4e4d53e8d5e9a5961870e8f4e62631cb937",
+                "revision": "sha256:0af6ac4cfd9d26bd90eeba094c3ceeed80e418dbf1872e6b1ffaa23b6b30e397"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.draft_acceptance_issue",
+                "ref": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "resource_count": 343,
+                "resource_hashes_sha256": "251e45e8af28e24055a9df480c2d541a1a0d3be419e40b212a5aae5cd61b4761",
+                "revision": "sha256:9f179f995ef669f59414e778a2064a4eece7fda9571ae0c191274b4f718b6b2d"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.host_memory_management",
+                "ref": "package:inferrsi.common.skills.host_memory_management",
+                "resource_count": 343,
+                "resource_hashes_sha256": "704c39f2041e99d4d66d9e4464c116274638e246db6a4b56e7c447dd711c198f",
+                "revision": "sha256:96c29b3da68c9493ce25421640f23cc2c15258005b6e006b6e9c2af285624fc3"
+              }
+            ],
+            "selection": {
+              "cue": "sourdough levain hydration recipe",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 505859,
+            "sha256": "504fb320c520276a163585366787f608153457c6eea9670efc3d394092945afd"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": false,
+          "case": "dev16",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "withheld01",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "withheld01",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 85410,
+            "no_match": false,
+            "omissions": [],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.training_input_pipeline",
+                "ref": "package:inferrsi.common.skills.training_input_pipeline",
+                "resource_count": 343,
+                "resource_hashes_sha256": "2316be47093b856fff9936ba91f578c2c42c12938ecedc7eae24e16a0fb6385a",
+                "revision": "sha256:d2e9d3993b49d3de2eac8ee492e8467c80bbb26787cee0687bb78a76c111f3d5"
+              }
+            ],
+            "selection": {
+              "cue": "prefetch_queue_depth executor_statistics",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 85882,
+            "sha256": "5e3948a70e07c223f3599b3b2b38f92377c00de65aa00efb732b41e6d4e3b383"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "withheld01",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 9502,
+            "context_omissions": [],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "64bd2c6b795d966045d1d3cca55976d3d49b08b06df7c7e06a70c2824cf55b7e",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/training-input-pipeline/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/training-input-pipeline/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md"
+                  }
+                ],
+                "content_bytes": 9502,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/training-input-pipeline/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.training_input_pipeline",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/training-input-pipeline/SKILL.md",
+                    "role": "core",
+                    "sha256": "64bd2c6b795d966045d1d3cca55976d3d49b08b06df7c7e06a70c2824cf55b7e"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/training-input-pipeline/references/sources.md",
+                    "role": "reference",
+                    "sha256": "60e0f3e88655d8e1d5eb4797f39ca4370deda01614f265b8d2ea4559d8ac020f"
+                  }
+                ],
+                "revision": "sha256:d2e9d3993b49d3de2eac8ee492e8467c80bbb26787cee0687bb78a76c111f3d5",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "64bd2c6b795d966045d1d3cca55976d3d49b08b06df7c7e06a70c2824cf55b7e",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/training-input-pipeline/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/training-input-pipeline/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 89438,
+            "no_match": false,
+            "omissions": [],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.training_input_pipeline",
+                "ref": "package:inferrsi.common.skills.training_input_pipeline",
+                "resource_count": 343,
+                "resource_hashes_sha256": "2316be47093b856fff9936ba91f578c2c42c12938ecedc7eae24e16a0fb6385a",
+                "revision": "sha256:d2e9d3993b49d3de2eac8ee492e8467c80bbb26787cee0687bb78a76c111f3d5"
+              }
+            ],
+            "selection": {
+              "cue": "prefetch_queue_depth executor_statistics",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 102738,
+            "sha256": "c9e3edc3cb701e76b16a4eda2257be592c37798f136788546aec03bb24c86665"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "withheld01",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "withheld02",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "withheld02",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 427591,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_device_health",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.calibrate_serving_simulation",
+                "ref": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "resource_count": 343,
+                "resource_hashes_sha256": "b4b818eb48639d4c8d599222a472bd4ec3ae2871b14e86e0c16b47689c03ecbe",
+                "revision": "sha256:4531196bea6c75c6fe11252c560708bbbcccf7b233bfd854e7fa39acfa70a9c0"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.assess_serving_preemption",
+                "ref": "package:inferrsi.common.skills.assess_serving_preemption",
+                "resource_count": 343,
+                "resource_hashes_sha256": "1514c158f4d1bbe57e52a20bb872edbd6c628c9fe34cf5018e0d9c4abcab1a55",
+                "revision": "sha256:11751aa0d285e7fcf21c1c0f4564c46a09683b123895cc95bc22a3eccaae6221"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.cann_attribution_overlap",
+                "ref": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c0bf9f5ab49059fd46fbee3e58bc0fac2819bb9d727e921d2362e618e92f996c",
+                "revision": "sha256:425a1e6d432ff3aa077b6974b41c66fb0a88be4ae88b8ba7a827071dfd8f5f10"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.storage_io_measurement",
+                "ref": "package:inferrsi.common.skills.storage_io_measurement",
+                "resource_count": 343,
+                "resource_hashes_sha256": "350d9c963ece5f1e1359ec238f4f804912b7bf0f08743b9cd4b6207583d52515",
+                "revision": "sha256:a57650b2bcaaac84bce2725a49eadb8c17533c82a1125b5ed5d840df5fca201c"
+              }
+            ],
+            "selection": {
+              "cue": "Can a serving simulator certify latency SLOs near queue capacity?",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 433025,
+            "sha256": "53a7e3db0c4873887c03aa8b14273d2153d214ca52be87912dab231e661915ac"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "withheld02",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 62288,
+            "context_omissions": [
+              {
+                "available_bytes": 12220,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.assess_serving_preemption",
+                "required_bytes": 36807
+              },
+              {
+                "available_bytes": 3248,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.storage_io_measurement",
+                "required_bytes": 8813
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "ac773c6f1c110a664dfca28a3ae9129b10c07614e60ca3640dea776220e7b352",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/SKILL.md"
+                  }
+                ],
+                "content_bytes": 12299,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/calibrate-serving-simulation/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/calibrate-serving-simulation/SKILL.md",
+                    "role": "core",
+                    "sha256": "ac773c6f1c110a664dfca28a3ae9129b10c07614e60ca3640dea776220e7b352"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/calibrate-serving-simulation/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "9696d3c0ef96b6611787446bd6f96d4d6255ca8d107105de8c1d9af83edb9fa5"
+                  }
+                ],
+                "revision": "sha256:4531196bea6c75c6fe11252c560708bbbcccf7b233bfd854e7fa39acfa70a9c0",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "ac773c6f1c110a664dfca28a3ae9129b10c07614e60ca3640dea776220e7b352",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md"
+                  }
+                ],
+                "content_bytes": 41017,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Normalized provenance covers the existing llama.cpp loading slice at 3e390af18de1131dda27913c00511725b4274c8a and the fastsafetensors distributed-delivery slice at 219e2561a41ede2d5d8eba611c248bb0effbf580; other inherited claims remain unnormalized with original identities and limits retained.",
+                  "The readsets record static mechanisms and source recommendations, not validated startup, memory, placement or backend performance.",
+                  "The fastsafetensors delivery readset does not revalidate inherited I/O/save claims or qualify uneven scatter, accelerator/NPU compatibility, inter-node transport or agent efficacy.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/weight-loading-optimization/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                    "role": "core",
+                    "sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                    "role": "reference",
+                    "sha256": "6c640f28ee3f0253a17b81461c865cb9f947367002cb75bf72550c4572d02e97"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                    "role": "reference",
+                    "sha256": "5a39f14b102b0cee304541e011aa31e44e6b13527d1bee9987c3f3b66c064d3d"
+                  }
+                ],
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "0e6f094dc554abea12e07def50f20b233b47127779a8aefad595b4ab1efd24c2",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md"
+                  }
+                ],
+                "content_bytes": 8972,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/ascend/cann-attribution-overlap/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md",
+                    "role": "core",
+                    "sha256": "0e6f094dc554abea12e07def50f20b233b47127779a8aefad595b4ab1efd24c2"
+                  }
+                ],
+                "revision": "sha256:425a1e6d432ff3aa077b6974b41c66fb0a88be4ae88b8ba7a827071dfd8f5f10",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "0e6f094dc554abea12e07def50f20b233b47127779a8aefad595b4ab1efd24c2",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 451987,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_device_health",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.calibrate_serving_simulation",
+                "ref": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "resource_count": 343,
+                "resource_hashes_sha256": "b4b818eb48639d4c8d599222a472bd4ec3ae2871b14e86e0c16b47689c03ecbe",
+                "revision": "sha256:4531196bea6c75c6fe11252c560708bbbcccf7b233bfd854e7fa39acfa70a9c0"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.assess_serving_preemption",
+                "ref": "package:inferrsi.common.skills.assess_serving_preemption",
+                "resource_count": 343,
+                "resource_hashes_sha256": "1514c158f4d1bbe57e52a20bb872edbd6c628c9fe34cf5018e0d9c4abcab1a55",
+                "revision": "sha256:11751aa0d285e7fcf21c1c0f4564c46a09683b123895cc95bc22a3eccaae6221"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.cann_attribution_overlap",
+                "ref": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c0bf9f5ab49059fd46fbee3e58bc0fac2819bb9d727e921d2362e618e92f996c",
+                "revision": "sha256:425a1e6d432ff3aa077b6974b41c66fb0a88be4ae88b8ba7a827071dfd8f5f10"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.storage_io_measurement",
+                "ref": "package:inferrsi.common.skills.storage_io_measurement",
+                "resource_count": 343,
+                "resource_hashes_sha256": "350d9c963ece5f1e1359ec238f4f804912b7bf0f08743b9cd4b6207583d52515",
+                "revision": "sha256:a57650b2bcaaac84bce2725a49eadb8c17533c82a1125b5ed5d840df5fca201c"
+              }
+            ],
+            "selection": {
+              "cue": "Can a serving simulator certify latency SLOs near queue capacity?",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 530723,
+            "sha256": "eb926e0011d0a30484030b3c663e0b4bd8b4142f9facda0e3d8493e20d6e9262"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "withheld02",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "withheld03",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "withheld03",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 432101,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.assess_serving_preemption",
+                "ref": "package:inferrsi.common.skills.assess_serving_preemption",
+                "resource_count": 343,
+                "resource_hashes_sha256": "1514c158f4d1bbe57e52a20bb872edbd6c628c9fe34cf5018e0d9c4abcab1a55",
+                "revision": "sha256:11751aa0d285e7fcf21c1c0f4564c46a09683b123895cc95bc22a3eccaae6221"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "ref": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "resource_count": 343,
+                "resource_hashes_sha256": "b4f60b42dd217de69e7b619ed6d5bdb586ba15b86a0fbb045c905d36669c9fdc",
+                "revision": "sha256:7f70dc8cc627637a561a14f78df55f9b6522803d4652e66cc5be9b686cb64dc0"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "ref": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "resource_count": 343,
+                "resource_hashes_sha256": "75f66b30f6b01297bf08e3dd13e8cf2807c0b7f0fca17e448be531a7bddcd6a1",
+                "revision": "sha256:d2067aa28b25d0d7de079c19c303eaa9b30135737eccd69e492e93f7d53af23b"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.kv_tier_management",
+                "ref": "package:inferrsi.common.skills.kv_tier_management",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ee42b55fa8a404149ecd1dccc32a892566ee8867710531bc0af553409758a019",
+                "revision": "sha256:5816c2e526a71152486af2230ace24d27b54baeb373bf5d27e353ccd44f640aa"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.bound_windowed_attention_replay",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb70e2e746f7ba833a696431c6a4deda830188acf5c4eafce6b0008818f8190a",
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334"
+              }
+            ],
+            "selection": {
+              "cue": "FastServe suspended KV preemption progress",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 434835,
+            "sha256": "bdf2948a61a1f630f22db00b79aa3a7dd29374c2789ad302ac1d33b3f14a1cc9"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "withheld03",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 47619,
+            "context_omissions": [
+              {
+                "available_bytes": 28729,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "required_bytes": 31841
+              },
+              {
+                "available_bytes": 28729,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "required_bytes": 125365
+              },
+              {
+                "available_bytes": 17917,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "required_bytes": 29732
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "0a5f48d9352c5bf460775ef47f9136829978811f4a16e84983e299782cae305d",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/assess-serving-preemption/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/assess-serving-preemption/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md"
+                  }
+                ],
+                "content_bytes": 36807,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/assess-serving-preemption/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.assess_serving_preemption",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/assess-serving-preemption/SKILL.md",
+                    "role": "core",
+                    "sha256": "0a5f48d9352c5bf460775ef47f9136829978811f4a16e84983e299782cae305d"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/assess-serving-preemption/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "f29fcecfba0490008a354c11b42403983fe165e56ff907ad61cc6040d2b6407f"
+                  }
+                ],
+                "revision": "sha256:11751aa0d285e7fcf21c1c0f4564c46a09683b123895cc95bc22a3eccaae6221",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "0a5f48d9352c5bf460775ef47f9136829978811f4a16e84983e299782cae305d",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/assess-serving-preemption/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/assess-serving-preemption/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md"
+                  }
+                ],
+                "content_bytes": 10812,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/repositories/lmcache/kv-tier-management/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.kv_tier_management",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md",
+                    "role": "core",
+                    "sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27"
+                  }
+                ],
+                "revision": "sha256:5816c2e526a71152486af2230ace24d27b54baeb373bf5d27e353ccd44f640aa",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 461107,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.assess_serving_preemption",
+                "ref": "package:inferrsi.common.skills.assess_serving_preemption",
+                "resource_count": 343,
+                "resource_hashes_sha256": "1514c158f4d1bbe57e52a20bb872edbd6c628c9fe34cf5018e0d9c4abcab1a55",
+                "revision": "sha256:11751aa0d285e7fcf21c1c0f4564c46a09683b123895cc95bc22a3eccaae6221"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "ref": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "resource_count": 343,
+                "resource_hashes_sha256": "b4f60b42dd217de69e7b619ed6d5bdb586ba15b86a0fbb045c905d36669c9fdc",
+                "revision": "sha256:7f70dc8cc627637a561a14f78df55f9b6522803d4652e66cc5be9b686cb64dc0"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "ref": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "resource_count": 343,
+                "resource_hashes_sha256": "75f66b30f6b01297bf08e3dd13e8cf2807c0b7f0fca17e448be531a7bddcd6a1",
+                "revision": "sha256:d2067aa28b25d0d7de079c19c303eaa9b30135737eccd69e492e93f7d53af23b"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.kv_tier_management",
+                "ref": "package:inferrsi.common.skills.kv_tier_management",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ee42b55fa8a404149ecd1dccc32a892566ee8867710531bc0af553409758a019",
+                "revision": "sha256:5816c2e526a71152486af2230ace24d27b54baeb373bf5d27e353ccd44f640aa"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.bound_windowed_attention_replay",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb70e2e746f7ba833a696431c6a4deda830188acf5c4eafce6b0008818f8190a",
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334"
+              }
+            ],
+            "selection": {
+              "cue": "FastServe suspended KV preemption progress",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 518251,
+            "sha256": "7ff9648555ebc5513af1c8dc2eb6b8d09a72173071b9d60dfcc7f95341197a1e"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "withheld03",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "withheld04",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "withheld04",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 430471,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "ref": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "resource_count": 343,
+                "resource_hashes_sha256": "8dae7399288be873c56dd379c37268e885c186af8cecde28ec1d5678adb9f204",
+                "revision": "sha256:b2a7f738b1bf78369f99c7e754f293f0b0a94a041156be5831c7f71a4c473eb4"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_graph_mode",
+                "ref": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ab57a9a122977730364bef92b56851835a4a447dc1ccff65874750ef3bac00b1",
+                "revision": "sha256:2b5dc0b9fea661b7bcbe1652fdf8231e7aef22d57736317597d5935d00808663"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sd_cpp_residency",
+                "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c83d7a76dffa5a3d97e932a3eac4602ad82e20babb09d9942504fff6b4db53d6",
+                "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.bound_windowed_attention_replay",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb70e2e746f7ba833a696431c6a4deda830188acf5c4eafce6b0008818f8190a",
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.dcp_async_save_performance",
+                "ref": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "resource_count": 343,
+                "resource_hashes_sha256": "b59637d16a7076477f6011fcceeb3cb5b402aa95af17b841ac9b2ede396216e3",
+                "revision": "sha256:36243cf65b6eaa5f370fd4f7633b8fccb119ff75ed38ba652a6061469de0b5b3"
+              }
+            ],
+            "selection": {
+              "cue": "ACLGraph stable replay buffers",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 433376,
+            "sha256": "246d682026ae637d40def566a27fe95433ba9815666337b7aeed236d5b58fe13"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "withheld04",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 63132,
+            "context_omissions": [
+              {
+                "available_bytes": 15040,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "required_bytes": 29732
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "d494fc7e539cbb7388189bd7b3f04bb5c4983fdd7a82484348aad1cee7f741f2",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/metadata.json",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/SKILL.md"
+                  }
+                ],
+                "content_bytes": 19749,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/SKILL.md",
+                    "role": "core",
+                    "sha256": "d494fc7e539cbb7388189bd7b3f04bb5c4983fdd7a82484348aad1cee7f741f2"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "6ffbfb68cc7e04923b4fbca2a73a7555b7c3b45a6f40819bb5f3213ef05cf88d"
+                  }
+                ],
+                "revision": "sha256:b2a7f738b1bf78369f99c7e754f293f0b0a94a041156be5831c7f71a4c473eb4",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "d494fc7e539cbb7388189bd7b3f04bb5c4983fdd7a82484348aad1cee7f741f2",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/metadata.json",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "9203cc5a14a1d65df5fbf275734b93be023e06527459b2f79864f28a338cb44a",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/SKILL.md"
+                  }
+                ],
+                "content_bytes": 5839,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/ascend/ascend-graph-mode/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/ascend-graph-mode/SKILL.md",
+                    "role": "core",
+                    "sha256": "9203cc5a14a1d65df5fbf275734b93be023e06527459b2f79864f28a338cb44a"
+                  }
+                ],
+                "revision": "sha256:2b5dc0b9fea661b7bcbe1652fdf8231e7aef22d57736317597d5935d00808663",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "9203cc5a14a1d65df5fbf275734b93be023e06527459b2f79864f28a338cb44a",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md"
+                  }
+                ],
+                "content_bytes": 24908,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; native inference, allocator behavior, image quality and performance were not run.",
+                  "Eighteen bounded readset paths plus one separately identity-checked README match the pinned Git commit/tree blob IDs and sizes (19 blobs total); header coverage is limited to selected owner/state declarations, and this does not authenticate every archive member or submodule contents.",
+                  "GGML and other submodules are not expanded; matching private GGML headers are required even for a system library.",
+                  "CPU bypasses the inspected accelerator capacity gate and segmentation eligibility, so CPU checks cannot qualify accelerator memory budgets or prefetch.",
+                  "Video/audio pipelines, detailed model kernels, approximate denoising caches and full multi-device execution remain outside this readset.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json",
+                    "reason": "not_context_resource"
+                  },
+                  {
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md",
+                    "reason": "not_relevant"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md",
+                    "role": "core",
+                    "sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "b67affbd7d237d82b7f51be49f1537e9820fe56ec70a71ee76809f9e599ca693"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE",
+                    "role": "reference",
+                    "sha256": "b53fa08f515cb5a6fff7b9fd8fcd0961a4b80df29d74372d25e1f9171aa042ee"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md",
+                    "role": "reference",
+                    "sha256": "7b537cb972347dd68e303960f142892346d8b19b884e4d08016ea36a23bdaa7a"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json",
+                    "role": "reference",
+                    "sha256": "af4830370b6eddc2d176a9ae14da4ab56a90b0fe048a01a89ee051df4365d8e9"
+                  }
+                ],
+                "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md"
+                  }
+                ],
+                "content_bytes": 12636,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md",
+                    "role": "core",
+                    "sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md",
+                    "role": "reference",
+                    "sha256": "22afe4f4d77f441096cc657d55d763473bfcbea8fb531c8e493fe49ab84955f7"
+                  }
+                ],
+                "revision": "sha256:36243cf65b6eaa5f370fd4f7633b8fccb119ff75ed38ba652a6061469de0b5b3",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 457833,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "ref": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "resource_count": 343,
+                "resource_hashes_sha256": "8dae7399288be873c56dd379c37268e885c186af8cecde28ec1d5678adb9f204",
+                "revision": "sha256:b2a7f738b1bf78369f99c7e754f293f0b0a94a041156be5831c7f71a4c473eb4"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_graph_mode",
+                "ref": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ab57a9a122977730364bef92b56851835a4a447dc1ccff65874750ef3bac00b1",
+                "revision": "sha256:2b5dc0b9fea661b7bcbe1652fdf8231e7aef22d57736317597d5935d00808663"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sd_cpp_residency",
+                "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c83d7a76dffa5a3d97e932a3eac4602ad82e20babb09d9942504fff6b4db53d6",
+                "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.bound_windowed_attention_replay",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb70e2e746f7ba833a696431c6a4deda830188acf5c4eafce6b0008818f8190a",
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.dcp_async_save_performance",
+                "ref": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "resource_count": 343,
+                "resource_hashes_sha256": "b59637d16a7076477f6011fcceeb3cb5b402aa95af17b841ac9b2ede396216e3",
+                "revision": "sha256:36243cf65b6eaa5f370fd4f7633b8fccb119ff75ed38ba652a6061469de0b5b3"
+              }
+            ],
+            "selection": {
+              "cue": "ACLGraph stable replay buffers",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 540079,
+            "sha256": "5f59eb7e5e46575cdd98e9e9fe4418b371e520535d5527315b040710d075236e"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "withheld04",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 172955,
+            "no_match": false,
+            "omissions": [],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sd_cpp_residency",
+                "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c83d7a76dffa5a3d97e932a3eac4602ad82e20babb09d9942504fff6b4db53d6",
+                "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.dcp_async_save_performance",
+                "ref": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "resource_count": 343,
+                "resource_hashes_sha256": "b59637d16a7076477f6011fcceeb3cb5b402aa95af17b841ac9b2ede396216e3",
+                "revision": "sha256:36243cf65b6eaa5f370fd4f7633b8fccb119ff75ed38ba652a6061469de0b5b3"
+              }
+            ],
+            "serialized_bytes": 173274,
+            "sha256": "ad3941abe38ece8301a3ead812ee767ea65c7288e24f57afe0a471f8eac20e20"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "withheld05",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "withheld05",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 433368,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_device_health",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "ref": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "resource_count": 343,
+                "resource_hashes_sha256": "568becaa26f22d5fcb75a81fb68bb6b0f46a9345dbddd9518a5057a62e6c9e28",
+                "revision": "sha256:a2c3f95251ace650d22c0bc2804780c91236da923993a46cbb25a58962ae00d1"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_distributed_parallel",
+                "ref": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ecb6d382307b933beb053b3c380a5bb6ba87c13792c428ad4131c78f5a9bcd04",
+                "revision": "sha256:66dab07081fad44b9cfe69b9c0f7a8f1dbb2c3fa19157ac99e4eee9a6306f605"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.budget_speculative_verification",
+                "ref": "package:inferrsi.common.skills.budget_speculative_verification",
+                "resource_count": 343,
+                "resource_hashes_sha256": "6ef4bcbc8e2ee8a5fc04c5c970ce5ec223bc85904818d7e006fb6d67548d734b",
+                "revision": "sha256:2686550fa366a367d026447ac1f2bec6b9ad3a81d559b70dccb8098524a27e2d"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "ref": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "resource_count": 343,
+                "resource_hashes_sha256": "8af3478b2f59c5fade35c614a9bfe4e4d53e8d5e9a5961870e8f4e62631cb937",
+                "revision": "sha256:0af6ac4cfd9d26bd90eeba094c3ceeed80e418dbf1872e6b1ffaa23b6b30e397"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "ref": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "resource_count": 343,
+                "resource_hashes_sha256": "75f66b30f6b01297bf08e3dd13e8cf2807c0b7f0fca17e448be531a7bddcd6a1",
+                "revision": "sha256:d2067aa28b25d0d7de079c19c303eaa9b30135737eccd69e492e93f7d53af23b"
+              }
+            ],
+            "selection": {
+              "cue": "How is moving a request between replicas different from changing its sequence parallel group?",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 438800,
+            "sha256": "90023483fcd615d2ac31d9dea0f83b559c443293874331264bb7ed429e382a77"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "withheld05",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 51463,
+            "context_omissions": [
+              {
+                "available_bytes": 14073,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "required_bytes": 24091
+              },
+              {
+                "available_bytes": 14073,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.budget_speculative_verification",
+                "required_bytes": 30031
+              },
+              {
+                "available_bytes": 14073,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "required_bytes": 47942
+              },
+              {
+                "available_bytes": 14073,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "required_bytes": 125365
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "82d0099d3c33706f46c657d4fa7f76d3d486e019b36760d1de001921328a60cc",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/SKILL.md"
+                  }
+                ],
+                "content_bytes": 51463,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/evaluate-elastic-serving-placement/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/evaluate-elastic-serving-placement/SKILL.md",
+                    "role": "core",
+                    "sha256": "82d0099d3c33706f46c657d4fa7f76d3d486e019b36760d1de001921328a60cc"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/evaluate-elastic-serving-placement/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "ea416831fd7d0874e06d0a76fb0c7196df53072b717673a0cdf635db6bfd917b"
+                  }
+                ],
+                "revision": "sha256:a2c3f95251ace650d22c0bc2804780c91236da923993a46cbb25a58962ae00d1",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "82d0099d3c33706f46c657d4fa7f76d3d486e019b36760d1de001921328a60cc",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 465980,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_device_health",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "ref": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "resource_count": 343,
+                "resource_hashes_sha256": "568becaa26f22d5fcb75a81fb68bb6b0f46a9345dbddd9518a5057a62e6c9e28",
+                "revision": "sha256:a2c3f95251ace650d22c0bc2804780c91236da923993a46cbb25a58962ae00d1"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_distributed_parallel",
+                "ref": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ecb6d382307b933beb053b3c380a5bb6ba87c13792c428ad4131c78f5a9bcd04",
+                "revision": "sha256:66dab07081fad44b9cfe69b9c0f7a8f1dbb2c3fa19157ac99e4eee9a6306f605"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.budget_speculative_verification",
+                "ref": "package:inferrsi.common.skills.budget_speculative_verification",
+                "resource_count": 343,
+                "resource_hashes_sha256": "6ef4bcbc8e2ee8a5fc04c5c970ce5ec223bc85904818d7e006fb6d67548d734b",
+                "revision": "sha256:2686550fa366a367d026447ac1f2bec6b9ad3a81d559b70dccb8098524a27e2d"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "ref": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "resource_count": 343,
+                "resource_hashes_sha256": "8af3478b2f59c5fade35c614a9bfe4e4d53e8d5e9a5961870e8f4e62631cb937",
+                "revision": "sha256:0af6ac4cfd9d26bd90eeba094c3ceeed80e418dbf1872e6b1ffaa23b6b30e397"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "ref": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "resource_count": 343,
+                "resource_hashes_sha256": "75f66b30f6b01297bf08e3dd13e8cf2807c0b7f0fca17e448be531a7bddcd6a1",
+                "revision": "sha256:d2067aa28b25d0d7de079c19c303eaa9b30135737eccd69e492e93f7d53af23b"
+              }
+            ],
+            "selection": {
+              "cue": "How is moving a request between replicas different from changing its sequence parallel group?",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 527088,
+            "sha256": "23051b189f8457ebd6a0782db69ab66791287b5bd1a232dcf63503fcef96de50"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "withheld05",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "withheld06",
+          "mode": "search",
+          "output": {
+            "ordered_items": [
+              {
+                "id": "inferrsi.common.skills.performance_ceiling_analysis"
+              }
+            ],
+            "serialized_bytes": 12360,
+            "sha256": "878199c975f54b0a6256318c79d619e0d51a18a490ced61d43204c8e4ad2ec56"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "withheld06",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 437219,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.performance_ceiling_analysis",
+                "ref": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c915247877787fcecc949d1d212ae1f18e3ba2a82f02b7c1cf343dedf0a7e885",
+                "revision": "sha256:6f67a5d347f3d05198f92cfce346d30051a544427ec8922b9de64432c3ffe316"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.bound_windowed_attention_replay",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb70e2e746f7ba833a696431c6a4deda830188acf5c4eafce6b0008818f8190a",
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.kernel_writing_guide",
+                "ref": "package:inferrsi.common.skills.kernel_writing_guide",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ff2d799c7b84ff56e3dc76cdb1648283a621bd5c104c8317acc839969fc41208",
+                "revision": "sha256:adf2c9d4ec9e0157388dcb11ff0367773d2967717a8dd944ef92647b9307b9bc"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.npu_skill_factory",
+                "ref": "package:inferrsi.common.skills.npu_skill_factory",
+                "resource_count": 343,
+                "resource_hashes_sha256": "7cde43221171eca8e11527545216e348d8c464438bec7458aab055843bb440cf",
+                "revision": "sha256:c789873dd4195bee85724a481228f51852a09c870d630ad1e828901037f875b7"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.host_memory_management",
+                "ref": "package:inferrsi.common.skills.host_memory_management",
+                "resource_count": 343,
+                "resource_hashes_sha256": "704c39f2041e99d4d66d9e4464c116274638e246db6a4b56e7c447dd711c198f",
+                "revision": "sha256:96c29b3da68c9493ce25421640f23cc2c15258005b6e006b6e9c2af285624fc3"
+              }
+            ],
+            "selection": {
+              "cue": "roofline analytical empirical end-to-end ceiling",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 441466,
+            "sha256": "6c72e2c8df479a21536178c7f1ef1c210eb067b561fb5675ea8dba7c2fa718c4"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "withheld06",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 51702,
+            "context_omissions": [
+              {
+                "available_bytes": 65536,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "required_bytes": 170782
+              },
+              {
+                "available_bytes": 13834,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.npu_skill_factory",
+                "required_bytes": 65054
+              },
+              {
+                "available_bytes": 13834,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.host_memory_management",
+                "required_bytes": 49803
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "4703beb6e5194ef2ff2c0b5ec69f64e35417a4bc0dfb50a10c80a920b77ef3b5",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/validation.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md"
+                  }
+                ],
+                "content_bytes": 29732,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no equivalence run was performed.",
+                  "Applies to architectures with a demonstrated boundary dependency, not to arbitrary sliding-window stacks.",
+                  "Window sizes, boundary-layer index and ring lengths are model-config specific and are not transferable.",
+                  "Query-row compaction does not imply dropping historical keys; the two decisions are separate.",
+                  "Prompt logprobs are unrecoverable for the clipped region; the reference forces them off with a warning.",
+                  "The truncation path is off by default in the reference source.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md",
+                    "role": "core",
+                    "sha256": "4703beb6e5194ef2ff2c0b5ec69f64e35417a4bc0dfb50a10c80a920b77ef3b5"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "fd917bd5ac80bee9c9f4525f8e0300f6d67dc144e2a16632af42a67f3d53d098"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/validation.md",
+                    "role": "reference",
+                    "sha256": "60e71e70c5519bcdf6c171eec2be2f342e8801bedd746d6005f11adc646dfaff"
+                  }
+                ],
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "4703beb6e5194ef2ff2c0b5ec69f64e35417a4bc0dfb50a10c80a920b77ef3b5",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/validation.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "01fb1712d7eb988c5e412f02340d2e48b4f8564fcf4ab232dc5897869d0d50a0",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/kernel-writing-guide/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/kernel-writing-guide/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/kernel-writing-guide/references/akg-affine-tail-guards.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/kernel-writing-guide/SKILL.md"
+                  }
+                ],
+                "content_bytes": 21970,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/kernel-writing-guide/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.kernel_writing_guide",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/kernel-writing-guide/SKILL.md",
+                    "role": "core",
+                    "sha256": "01fb1712d7eb988c5e412f02340d2e48b4f8564fcf4ab232dc5897869d0d50a0"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/kernel-writing-guide/references/akg-affine-tail-guards.md",
+                    "role": "reference",
+                    "sha256": "113deb2db10b171db3425bb3f1792bee763706b794c8cc2e222b4cc0d0114118"
+                  }
+                ],
+                "revision": "sha256:adf2c9d4ec9e0157388dcb11ff0367773d2967717a8dd944ef92647b9307b9bc",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "01fb1712d7eb988c5e412f02340d2e48b4f8564fcf4ab232dc5897869d0d50a0",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/kernel-writing-guide/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/kernel-writing-guide/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/kernel-writing-guide/references/akg-affine-tail-guards.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/kernel-writing-guide/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 478766,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.performance_ceiling_analysis",
+                "ref": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c915247877787fcecc949d1d212ae1f18e3ba2a82f02b7c1cf343dedf0a7e885",
+                "revision": "sha256:6f67a5d347f3d05198f92cfce346d30051a544427ec8922b9de64432c3ffe316"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.bound_windowed_attention_replay",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb70e2e746f7ba833a696431c6a4deda830188acf5c4eafce6b0008818f8190a",
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.kernel_writing_guide",
+                "ref": "package:inferrsi.common.skills.kernel_writing_guide",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ff2d799c7b84ff56e3dc76cdb1648283a621bd5c104c8317acc839969fc41208",
+                "revision": "sha256:adf2c9d4ec9e0157388dcb11ff0367773d2967717a8dd944ef92647b9307b9bc"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.npu_skill_factory",
+                "ref": "package:inferrsi.common.skills.npu_skill_factory",
+                "resource_count": 343,
+                "resource_hashes_sha256": "7cde43221171eca8e11527545216e348d8c464438bec7458aab055843bb440cf",
+                "revision": "sha256:c789873dd4195bee85724a481228f51852a09c870d630ad1e828901037f875b7"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.host_memory_management",
+                "ref": "package:inferrsi.common.skills.host_memory_management",
+                "resource_count": 343,
+                "resource_hashes_sha256": "704c39f2041e99d4d66d9e4464c116274638e246db6a4b56e7c447dd711c198f",
+                "revision": "sha256:96c29b3da68c9493ce25421640f23cc2c15258005b6e006b6e9c2af285624fc3"
+              }
+            ],
+            "selection": {
+              "cue": "roofline analytical empirical end-to-end ceiling",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 543170,
+            "sha256": "0c9097a61b83e59a924ee2be136aae957cd16a393c3533a8b0d76758d1e12097"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "withheld06",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 90114,
+            "no_match": false,
+            "omissions": [],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.performance_ceiling_analysis",
+                "ref": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c915247877787fcecc949d1d212ae1f18e3ba2a82f02b7c1cf343dedf0a7e885",
+                "revision": "sha256:6f67a5d347f3d05198f92cfce346d30051a544427ec8922b9de64432c3ffe316"
+              }
+            ],
+            "serialized_bytes": 90431,
+            "sha256": "70cb18c9f4f5c7f3ab58acab24d362b94f743c2488d270976d84ade1458688ed"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": false,
+          "case": "withheld07",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": false,
+          "case": "withheld07",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "selection": {
+              "cue": "zephyrcodex latticeotter q83",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 455,
+            "sha256": "a20a7f3222447ce17be8178a15f84a96cd11f61b85861f826c9a22c51d49d87b"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": false,
+          "case": "withheld07",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 0,
+            "context_omissions": [],
+            "contexts": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "selection": {
+              "cue": "zephyrcodex latticeotter q83",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 639,
+            "sha256": "121d4fffb7abe4d79ecb33694570a620007712feeb0062b3602eebb7f1d902a7"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": false,
+          "case": "withheld07",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": false,
+          "case": "withheld08",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": false,
+          "case": "withheld08",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 434773,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_llama_runtime_state",
+                "ref": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "resource_count": 343,
+                "resource_hashes_sha256": "03e86bd3af77f17d6c7ad2933534524ab7fe5993396a7fe9748a6014496a34b5",
+                "revision": "sha256:a3c3d3244cda589e8e7809e56f43b7bfff1852def8dd26fb03a9d29e39bca3dd"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sd_cpp_residency",
+                "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c83d7a76dffa5a3d97e932a3eac4602ad82e20babb09d9942504fff6b4db53d6",
+                "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "ref": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "resource_count": 343,
+                "resource_hashes_sha256": "8af3478b2f59c5fade35c614a9bfe4e4d53e8d5e9a5961870e8f4e62631cb937",
+                "revision": "sha256:0af6ac4cfd9d26bd90eeba094c3ceeed80e418dbf1872e6b1ffaa23b6b30e397"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "ref": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "resource_count": 343,
+                "resource_hashes_sha256": "75f66b30f6b01297bf08e3dd13e8cf2807c0b7f0fca17e448be531a7bddcd6a1",
+                "revision": "sha256:d2067aa28b25d0d7de079c19c303eaa9b30135737eccd69e492e93f7d53af23b"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.diffusers_invocation_state",
+                "ref": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "resource_count": 343,
+                "resource_hashes_sha256": "64b10a60e44276bc4a20b3441f65afd929341e8558fde32491ba27a20fde4020",
+                "revision": "sha256:a8529f2c6bf5f84ce48ba0d6426f1bf1bc6fce172cd0a599b6a149878e3b4be4"
+              }
+            ],
+            "selection": {
+              "cue": "baroque harpsichord tuning temperament",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 435955,
+            "sha256": "0b35e7fc7320ef46c35a3eecfff7896a76cb5bc9c78d60b6b65435a9dadb3423"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": false,
+          "case": "withheld08",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 49333,
+            "context_omissions": [
+              {
+                "available_bytes": 27809,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "required_bytes": 47942
+              },
+              {
+                "available_bytes": 27809,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "required_bytes": 125365
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "95e81b6bc096a954fa73139ea5bc001d1efa6c99747161da62f012b494e5c013",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/SKILL.md"
+                  }
+                ],
+                "content_bytes": 24572,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; native runtime and efficacy validation not run.",
+                  "No cross-backend preemption, model-quality or performance claim.",
+                  "KV, recurrent and hybrid state contracts must be checked independently.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/SKILL.md",
+                    "role": "core",
+                    "sha256": "95e81b6bc096a954fa73139ea5bc001d1efa6c99747161da62f012b494e5c013"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "3d12a401e9decb55a2f4f3b2bfa0d360ef5110f63a0469dfe285654e78054190"
+                  }
+                ],
+                "revision": "sha256:a3c3d3244cda589e8e7809e56f43b7bfff1852def8dd26fb03a9d29e39bca3dd",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "95e81b6bc096a954fa73139ea5bc001d1efa6c99747161da62f012b494e5c013",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md"
+                  }
+                ],
+                "content_bytes": 13155,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; native inference, allocator behavior, image quality and performance were not run.",
+                  "Eighteen bounded readset paths plus one separately identity-checked README match the pinned Git commit/tree blob IDs and sizes (19 blobs total); header coverage is limited to selected owner/state declarations, and this does not authenticate every archive member or submodule contents.",
+                  "GGML and other submodules are not expanded; matching private GGML headers are required even for a system library.",
+                  "CPU bypasses the inspected accelerator capacity gate and segmentation eligibility, so CPU checks cannot qualify accelerator memory budgets or prefetch.",
+                  "Video/audio pipelines, detailed model kernels, approximate denoising caches and full multi-device execution remain outside this readset.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json",
+                    "reason": "not_context_resource"
+                  },
+                  {
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md",
+                    "reason": "not_relevant"
+                  },
+                  {
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json",
+                    "reason": "not_relevant"
+                  },
+                  {
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md",
+                    "reason": "not_relevant"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md",
+                    "role": "core",
+                    "sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "b67affbd7d237d82b7f51be49f1537e9820fe56ec70a71ee76809f9e599ca693"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE",
+                    "role": "reference",
+                    "sha256": "b53fa08f515cb5a6fff7b9fd8fcd0961a4b80df29d74372d25e1f9171aa042ee"
+                  }
+                ],
+                "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "7581b56336d07213a87d01477793ae1a8acf0b7cf876fce36c9a761ee32c1ff4",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/SKILL.md",
+                      "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/metadata.json",
+                      "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/ownership.md",
+                      "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/source-map.md",
+                      "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/validation.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/SKILL.md"
+                  }
+                ],
+                "content_bytes": 11606,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no installed API, native scheduler, model, concurrency or service validation.",
+                  "Supplied manifest verifies a selected pipeline_utils Git blob, not the whole source tree.",
+                  "Other pipelines and schedulers require separate contracts; common indexing is not backend qualification.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/frameworks/diffusers/diffusers-invocation-state/metadata.json",
+                    "reason": "not_context_resource"
+                  },
+                  {
+                    "path": "data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/ownership.md",
+                    "reason": "not_relevant"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/diffusers/diffusers-invocation-state/SKILL.md",
+                    "role": "core",
+                    "sha256": "7581b56336d07213a87d01477793ae1a8acf0b7cf876fce36c9a761ee32c1ff4"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/source-map.md",
+                    "role": "reference",
+                    "sha256": "10562a75683bd9bd07ec2e37a1cbd0f93f2534d6f85b55a566adb2d97485215f"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/validation.md",
+                    "role": "reference",
+                    "sha256": "5a0b854b87d091d7d505b0bdd5c03330052507a851ab6edf35b93a6675f2d4f7"
+                  }
+                ],
+                "revision": "sha256:a8529f2c6bf5f84ce48ba0d6426f1bf1bc6fce172cd0a599b6a149878e3b4be4",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "7581b56336d07213a87d01477793ae1a8acf0b7cf876fce36c9a761ee32c1ff4",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/SKILL.md",
+                      "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/metadata.json",
+                      "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/ownership.md",
+                      "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/source-map.md",
+                      "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/validation.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 469565,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_llama_runtime_state",
+                "ref": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "resource_count": 343,
+                "resource_hashes_sha256": "03e86bd3af77f17d6c7ad2933534524ab7fe5993396a7fe9748a6014496a34b5",
+                "revision": "sha256:a3c3d3244cda589e8e7809e56f43b7bfff1852def8dd26fb03a9d29e39bca3dd"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sd_cpp_residency",
+                "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c83d7a76dffa5a3d97e932a3eac4602ad82e20babb09d9942504fff6b4db53d6",
+                "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "ref": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "resource_count": 343,
+                "resource_hashes_sha256": "8af3478b2f59c5fade35c614a9bfe4e4d53e8d5e9a5961870e8f4e62631cb937",
+                "revision": "sha256:0af6ac4cfd9d26bd90eeba094c3ceeed80e418dbf1872e6b1ffaa23b6b30e397"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "ref": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "resource_count": 343,
+                "resource_hashes_sha256": "75f66b30f6b01297bf08e3dd13e8cf2807c0b7f0fca17e448be531a7bddcd6a1",
+                "revision": "sha256:d2067aa28b25d0d7de079c19c303eaa9b30135737eccd69e492e93f7d53af23b"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.diffusers_invocation_state",
+                "ref": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "resource_count": 343,
+                "resource_hashes_sha256": "64b10a60e44276bc4a20b3441f65afd929341e8558fde32491ba27a20fde4020",
+                "revision": "sha256:a8529f2c6bf5f84ce48ba0d6426f1bf1bc6fce172cd0a599b6a149878e3b4be4"
+              }
+            ],
+            "selection": {
+              "cue": "baroque harpsichord tuning temperament",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 533888,
+            "sha256": "bc2e924a1158532fc45063706c569706e559f80e841ded5494168ad894ff1e47"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": false,
+          "case": "withheld08",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        }
+      ]
+    },
+    "B": {
+      "modules": {
+        "inferrsi": {
+          "sha256": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47",
+          "source": "src/inferrsi/__init__.py"
+        },
+        "inferrsi.catalog": {
+          "sha256": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5",
+          "source": "src/inferrsi/catalog.py"
+        },
+        "inferrsi.cli": {
+          "sha256": "91ddb289186ffb6bd76372c03a0ed4714cef2a50a0fee4452a027a67c0155b88",
+          "source": "src/inferrsi/cli.py"
+        },
+        "inferrsi.knowledge_operations": {
+          "sha256": "c71bb2063ab8d919ca5907ac900a3930ec60730515c614fee7375d9efb77173a",
+          "source": "src/inferrsi/knowledge_operations.py"
+        }
+      },
+      "outputs": [
+        {
+          "answerable": true,
+          "case": "dev01",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev01",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 428778,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_device_health",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.dcp_async_save_performance",
+                "ref": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "resource_count": 343,
+                "resource_hashes_sha256": "b59637d16a7076477f6011fcceeb3cb5b402aa95af17b841ac9b2ede396216e3",
+                "revision": "sha256:36243cf65b6eaa5f370fd4f7633b8fccb119ff75ed38ba652a6061469de0b5b3"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.checkpoint_recovery_consistency",
+                "ref": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "15c906f896eab2a681eb9bc9e42ee29d11b0594a1e1aeacbf6f414233209d1fa",
+                "revision": "sha256:73f77f89e298ce8724258cb3d74315438c036297bb38c63aac3d7dc95473ce53"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.inference_benchmark_design",
+                "ref": "package:inferrsi.common.skills.inference_benchmark_design",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb743a7fe0662feead981795dc60e2a5c9355ed876339f14ad7a47caab0bb670",
+                "revision": "sha256:6b0cbeaa5d409c04228883884e192f275cad4874c7f6b3039abef9fd443fafb5"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.kv_tier_management",
+                "ref": "package:inferrsi.common.skills.kv_tier_management",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ee42b55fa8a404149ecd1dccc32a892566ee8867710531bc0af553409758a019",
+                "revision": "sha256:5816c2e526a71152486af2230ace24d27b54baeb373bf5d27e353ccd44f640aa"
+              }
+            ],
+            "selection": {
+              "cue": "Why must checkpoint upload finish before cached staging buffers can be reused?",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 434223,
+            "sha256": "26e1236d18d1fbbdc1a5c4d94d4a8a69ec4a4bd230ec513541580fef1ab21a2b"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev01",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 64465,
+            "context_omissions": [
+              {
+                "available_bytes": 11883,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "required_bytes": 14400
+              },
+              {
+                "available_bytes": 11883,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.inference_benchmark_design",
+                "required_bytes": 47014
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md"
+                  }
+                ],
+                "content_bytes": 12636,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md",
+                    "role": "core",
+                    "sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md",
+                    "role": "reference",
+                    "sha256": "22afe4f4d77f441096cc657d55d763473bfcbea8fb531c8e493fe49ab84955f7"
+                  }
+                ],
+                "revision": "sha256:36243cf65b6eaa5f370fd4f7633b8fccb119ff75ed38ba652a6061469de0b5b3",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md"
+                  }
+                ],
+                "content_bytes": 41017,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Normalized provenance covers the existing llama.cpp loading slice at 3e390af18de1131dda27913c00511725b4274c8a and the fastsafetensors distributed-delivery slice at 219e2561a41ede2d5d8eba611c248bb0effbf580; other inherited claims remain unnormalized with original identities and limits retained.",
+                  "The readsets record static mechanisms and source recommendations, not validated startup, memory, placement or backend performance.",
+                  "The fastsafetensors delivery readset does not revalidate inherited I/O/save claims or qualify uneven scatter, accelerator/NPU compatibility, inter-node transport or agent efficacy.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/weight-loading-optimization/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                    "role": "core",
+                    "sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                    "role": "reference",
+                    "sha256": "6c640f28ee3f0253a17b81461c865cb9f947367002cb75bf72550c4572d02e97"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                    "role": "reference",
+                    "sha256": "5a39f14b102b0cee304541e011aa31e44e6b13527d1bee9987c3f3b66c064d3d"
+                  }
+                ],
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md"
+                  }
+                ],
+                "content_bytes": 10812,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/repositories/lmcache/kv-tier-management/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.kv_tier_management",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md",
+                    "role": "core",
+                    "sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27"
+                  }
+                ],
+                "revision": "sha256:5816c2e526a71152486af2230ace24d27b54baeb373bf5d27e353ccd44f640aa",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 458836,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_device_health",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.dcp_async_save_performance",
+                "ref": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "resource_count": 343,
+                "resource_hashes_sha256": "b59637d16a7076477f6011fcceeb3cb5b402aa95af17b841ac9b2ede396216e3",
+                "revision": "sha256:36243cf65b6eaa5f370fd4f7633b8fccb119ff75ed38ba652a6061469de0b5b3"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.checkpoint_recovery_consistency",
+                "ref": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "15c906f896eab2a681eb9bc9e42ee29d11b0594a1e1aeacbf6f414233209d1fa",
+                "revision": "sha256:73f77f89e298ce8724258cb3d74315438c036297bb38c63aac3d7dc95473ce53"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.inference_benchmark_design",
+                "ref": "package:inferrsi.common.skills.inference_benchmark_design",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb743a7fe0662feead981795dc60e2a5c9355ed876339f14ad7a47caab0bb670",
+                "revision": "sha256:6b0cbeaa5d409c04228883884e192f275cad4874c7f6b3039abef9fd443fafb5"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.kv_tier_management",
+                "ref": "package:inferrsi.common.skills.kv_tier_management",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ee42b55fa8a404149ecd1dccc32a892566ee8867710531bc0af553409758a019",
+                "revision": "sha256:5816c2e526a71152486af2230ace24d27b54baeb373bf5d27e353ccd44f640aa"
+              }
+            ],
+            "selection": {
+              "cue": "Why must checkpoint upload finish before cached staging buffers can be reused?",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 540055,
+            "sha256": "1afecd2e68d87dd1ef1e3c1c22ea846fd33c26bd751560219948aae98577f481"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev01",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev02",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev02",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 85545,
+            "no_match": false,
+            "omissions": [],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.dcp_async_save_performance",
+                "ref": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "resource_count": 343,
+                "resource_hashes_sha256": "b59637d16a7076477f6011fcceeb3cb5b402aa95af17b841ac9b2ede396216e3",
+                "revision": "sha256:36243cf65b6eaa5f370fd4f7633b8fccb119ff75ed38ba652a6061469de0b5b3"
+              }
+            ],
+            "selection": {
+              "cue": "DefaultStager",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 85990,
+            "sha256": "27b25a4ec6fa85219601eeb68b85fd54f3dc51c0d529a480f782ec7102a49684"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev02",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 12636,
+            "context_omissions": [],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md"
+                  }
+                ],
+                "content_bytes": 12636,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md",
+                    "role": "core",
+                    "sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md",
+                    "role": "reference",
+                    "sha256": "22afe4f4d77f441096cc657d55d763473bfcbea8fb531c8e493fe49ab84955f7"
+                  }
+                ],
+                "revision": "sha256:36243cf65b6eaa5f370fd4f7633b8fccb119ff75ed38ba652a6061469de0b5b3",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 89815,
+            "no_match": false,
+            "omissions": [],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.dcp_async_save_performance",
+                "ref": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "resource_count": 343,
+                "resource_hashes_sha256": "b59637d16a7076477f6011fcceeb3cb5b402aa95af17b841ac9b2ede396216e3",
+                "revision": "sha256:36243cf65b6eaa5f370fd4f7633b8fccb119ff75ed38ba652a6061469de0b5b3"
+              }
+            ],
+            "selection": {
+              "cue": "DefaultStager",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 106414,
+            "sha256": "672f7b868df0011c492de557341606d2b17761814ad5c5c5b8bfc5107699cd12"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev02",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev03",
+          "mode": "search",
+          "output": {
+            "ordered_items": [
+              {
+                "id": "inferrsi.common.skills.weight_loading_optimization"
+              }
+            ],
+            "serialized_bytes": 7193,
+            "sha256": "eee7ad1236b5121614e35949af184d10c7fed14b9d790befbdd850f12081ef2d"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev03",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 429394,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.training_input_pipeline",
+                "ref": "package:inferrsi.common.skills.training_input_pipeline",
+                "resource_count": 343,
+                "resource_hashes_sha256": "2316be47093b856fff9936ba91f578c2c42c12938ecedc7eae24e16a0fb6385a",
+                "revision": "sha256:d2e9d3993b49d3de2eac8ee492e8467c80bbb26787cee0687bb78a76c111f3d5"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.bound_windowed_attention_replay",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb70e2e746f7ba833a696431c6a4deda830188acf5c4eafce6b0008818f8190a",
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_distributed_parallel",
+                "ref": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ecb6d382307b933beb053b3c380a5bb6ba87c13792c428ad4131c78f5a9bcd04",
+                "revision": "sha256:66dab07081fad44b9cfe69b9c0f7a8f1dbb2c3fa19157ac99e4eee9a6306f605"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascendc_tiling_design",
+                "ref": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "resource_count": 343,
+                "resource_hashes_sha256": "caa4787f895186508b8b6f863b2d03a830364585ffd65f0c098be319c7d54cad",
+                "revision": "sha256:9c90d2bc30d84cf0c99a0382d1e61b01c1210efe51d16ddb81c25afcb7c04e4b"
+              }
+            ],
+            "selection": {
+              "cue": "safetensors index sharded downloads",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 430850,
+            "sha256": "6e8b364dc7518b24f75400cf2a48525b9b047a7b19a316ba8bcc7711a48ff089"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev03",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 62379,
+            "context_omissions": [
+              {
+                "available_bytes": 15017,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "required_bytes": 29732
+              },
+              {
+                "available_bytes": 15017,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "required_bytes": 24091
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md"
+                  }
+                ],
+                "content_bytes": 41017,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Normalized provenance covers the existing llama.cpp loading slice at 3e390af18de1131dda27913c00511725b4274c8a and the fastsafetensors distributed-delivery slice at 219e2561a41ede2d5d8eba611c248bb0effbf580; other inherited claims remain unnormalized with original identities and limits retained.",
+                  "The readsets record static mechanisms and source recommendations, not validated startup, memory, placement or backend performance.",
+                  "The fastsafetensors delivery readset does not revalidate inherited I/O/save claims or qualify uneven scatter, accelerator/NPU compatibility, inter-node transport or agent efficacy.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/weight-loading-optimization/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                    "role": "core",
+                    "sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                    "role": "reference",
+                    "sha256": "6c640f28ee3f0253a17b81461c865cb9f947367002cb75bf72550c4572d02e97"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                    "role": "reference",
+                    "sha256": "5a39f14b102b0cee304541e011aa31e44e6b13527d1bee9987c3f3b66c064d3d"
+                  }
+                ],
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "64bd2c6b795d966045d1d3cca55976d3d49b08b06df7c7e06a70c2824cf55b7e",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/training-input-pipeline/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/training-input-pipeline/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md"
+                  }
+                ],
+                "content_bytes": 9502,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/training-input-pipeline/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.training_input_pipeline",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/training-input-pipeline/SKILL.md",
+                    "role": "core",
+                    "sha256": "64bd2c6b795d966045d1d3cca55976d3d49b08b06df7c7e06a70c2824cf55b7e"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/training-input-pipeline/references/sources.md",
+                    "role": "reference",
+                    "sha256": "60e0f3e88655d8e1d5eb4797f39ca4370deda01614f265b8d2ea4559d8ac020f"
+                  }
+                ],
+                "revision": "sha256:d2e9d3993b49d3de2eac8ee492e8467c80bbb26787cee0687bb78a76c111f3d5",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "64bd2c6b795d966045d1d3cca55976d3d49b08b06df7c7e06a70c2824cf55b7e",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/training-input-pipeline/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/training-input-pipeline/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "93aa1b775ced304cebfe6c4dec0503ddb2260ecb3cb01edc17ce329446c5ce2b",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md"
+                  }
+                ],
+                "content_bytes": 11860,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/ascend/ascendc-tiling-design/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md",
+                    "role": "core",
+                    "sha256": "93aa1b775ced304cebfe6c4dec0503ddb2260ecb3cb01edc17ce329446c5ce2b"
+                  }
+                ],
+                "revision": "sha256:9c90d2bc30d84cf0c99a0382d1e61b01c1210efe51d16ddb81c25afcb7c04e4b",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "93aa1b775ced304cebfe6c4dec0503ddb2260ecb3cb01edc17ce329446c5ce2b",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 458004,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.training_input_pipeline",
+                "ref": "package:inferrsi.common.skills.training_input_pipeline",
+                "resource_count": 343,
+                "resource_hashes_sha256": "2316be47093b856fff9936ba91f578c2c42c12938ecedc7eae24e16a0fb6385a",
+                "revision": "sha256:d2e9d3993b49d3de2eac8ee492e8467c80bbb26787cee0687bb78a76c111f3d5"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.bound_windowed_attention_replay",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb70e2e746f7ba833a696431c6a4deda830188acf5c4eafce6b0008818f8190a",
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_distributed_parallel",
+                "ref": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ecb6d382307b933beb053b3c380a5bb6ba87c13792c428ad4131c78f5a9bcd04",
+                "revision": "sha256:66dab07081fad44b9cfe69b9c0f7a8f1dbb2c3fa19157ac99e4eee9a6306f605"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascendc_tiling_design",
+                "ref": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "resource_count": 343,
+                "resource_hashes_sha256": "caa4787f895186508b8b6f863b2d03a830364585ffd65f0c098be319c7d54cad",
+                "revision": "sha256:9c90d2bc30d84cf0c99a0382d1e61b01c1210efe51d16ddb81c25afcb7c04e4b"
+              }
+            ],
+            "selection": {
+              "cue": "safetensors index sharded downloads",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 532970,
+            "sha256": "6c8cbb5c8b5fcc958f84efa913dbd145e9c2066a2cb8ca16163d8880ff05fb0e"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev03",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 86217,
+            "no_match": false,
+            "omissions": [],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              }
+            ],
+            "serialized_bytes": 86534,
+            "sha256": "4a4c102e42ef0cefd5cb90dc4b7b0b310b8b0fa7f93e35e1f70aaae802e1d305"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev04",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev04",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 433875,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_device_health",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.budget_speculative_verification",
+                "ref": "package:inferrsi.common.skills.budget_speculative_verification",
+                "resource_count": 343,
+                "resource_hashes_sha256": "6ef4bcbc8e2ee8a5fc04c5c970ce5ec223bc85904818d7e006fb6d67548d734b",
+                "revision": "sha256:2686550fa366a367d026447ac1f2bec6b9ad3a81d559b70dccb8098524a27e2d"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_model_adaptation",
+                "ref": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "resource_count": 343,
+                "resource_hashes_sha256": "cd4f45eea332240ebf0229078a47067fe262badd13aae30db4e0e90868340861",
+                "revision": "sha256:1d03cc8e797c6f6dcda862244833019e8bb77941f1f63aecc8e7958bb8265863"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.host_overhead_optimization",
+                "ref": "package:inferrsi.common.skills.host_overhead_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "4fbb3ca8057addd4564893430704c2ffd32c8f7ea622d3d0ddc46ffb37332a86",
+                "revision": "sha256:902b79f76c19ecd9856af1e64b4a0ee916801b7e3e7e7f5bf6c3f105492a9ea0"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.performance_ceiling_analysis",
+                "ref": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c915247877787fcecc949d1d212ae1f18e3ba2a82f02b7c1cf343dedf0a7e885",
+                "revision": "sha256:6f67a5d347f3d05198f92cfce346d30051a544427ec8922b9de64432c3ffe316"
+              }
+            ],
+            "selection": {
+              "cue": "How should model loading overhead be separated from per request latency?",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 439307,
+            "sha256": "6ab5e1e796794df07d09ea5cfaa3d9a3870d601a9d675c2c22026d301ba2e764"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev04",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 63529,
+            "context_omissions": [
+              {
+                "available_bytes": 24519,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.budget_speculative_verification",
+                "required_bytes": 30031
+              },
+              {
+                "available_bytes": 2007,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "required_bytes": 170782
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md"
+                  }
+                ],
+                "content_bytes": 41017,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Normalized provenance covers the existing llama.cpp loading slice at 3e390af18de1131dda27913c00511725b4274c8a and the fastsafetensors distributed-delivery slice at 219e2561a41ede2d5d8eba611c248bb0effbf580; other inherited claims remain unnormalized with original identities and limits retained.",
+                  "The readsets record static mechanisms and source recommendations, not validated startup, memory, placement or backend performance.",
+                  "The fastsafetensors delivery readset does not revalidate inherited I/O/save claims or qualify uneven scatter, accelerator/NPU compatibility, inter-node transport or agent efficacy.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/weight-loading-optimization/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                    "role": "core",
+                    "sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                    "role": "reference",
+                    "sha256": "6c640f28ee3f0253a17b81461c865cb9f947367002cb75bf72550c4572d02e97"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                    "role": "reference",
+                    "sha256": "5a39f14b102b0cee304541e011aa31e44e6b13527d1bee9987c3f3b66c064d3d"
+                  }
+                ],
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "8fdf46fb2743b67a66bc069d22a45305a222debfb606777bc451b4b8f31d9cc6",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/SKILL.md"
+                  }
+                ],
+                "content_bytes": 6440,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/ascend/ascend-model-adaptation/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/ascend-model-adaptation/SKILL.md",
+                    "role": "core",
+                    "sha256": "8fdf46fb2743b67a66bc069d22a45305a222debfb606777bc451b4b8f31d9cc6"
+                  }
+                ],
+                "revision": "sha256:1d03cc8e797c6f6dcda862244833019e8bb77941f1f63aecc8e7958bb8265863",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "8fdf46fb2743b67a66bc069d22a45305a222debfb606777bc451b4b8f31d9cc6",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "36ef88826a32ac3147b869001004ded17a8d5a491e25bfe105005ba39ea9d1eb",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/metadata.json",
+                      "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/references/deepinfer-zero-overhead-scheduling.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/SKILL.md"
+                  }
+                ],
+                "content_bytes": 16072,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/host/host-overhead-optimization/metadata.json",
+                    "reason": "not_context_resource"
+                  },
+                  {
+                    "path": "data/skill_library/hardware/host/host-overhead-optimization/references/deepinfer-zero-overhead-scheduling.md",
+                    "reason": "context_limit"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.host_overhead_optimization",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/host/host-overhead-optimization/SKILL.md",
+                    "role": "core",
+                    "sha256": "36ef88826a32ac3147b869001004ded17a8d5a491e25bfe105005ba39ea9d1eb"
+                  }
+                ],
+                "revision": "sha256:902b79f76c19ecd9856af1e64b4a0ee916801b7e3e7e7f5bf6c3f105492a9ea0",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "36ef88826a32ac3147b869001004ded17a8d5a491e25bfe105005ba39ea9d1eb",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/metadata.json",
+                      "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/references/deepinfer-zero-overhead-scheduling.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 469108,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_device_health",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.budget_speculative_verification",
+                "ref": "package:inferrsi.common.skills.budget_speculative_verification",
+                "resource_count": 343,
+                "resource_hashes_sha256": "6ef4bcbc8e2ee8a5fc04c5c970ce5ec223bc85904818d7e006fb6d67548d734b",
+                "revision": "sha256:2686550fa366a367d026447ac1f2bec6b9ad3a81d559b70dccb8098524a27e2d"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_model_adaptation",
+                "ref": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "resource_count": 343,
+                "resource_hashes_sha256": "cd4f45eea332240ebf0229078a47067fe262badd13aae30db4e0e90868340861",
+                "revision": "sha256:1d03cc8e797c6f6dcda862244833019e8bb77941f1f63aecc8e7958bb8265863"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.host_overhead_optimization",
+                "ref": "package:inferrsi.common.skills.host_overhead_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "4fbb3ca8057addd4564893430704c2ffd32c8f7ea622d3d0ddc46ffb37332a86",
+                "revision": "sha256:902b79f76c19ecd9856af1e64b4a0ee916801b7e3e7e7f5bf6c3f105492a9ea0"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.performance_ceiling_analysis",
+                "ref": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c915247877787fcecc949d1d212ae1f18e3ba2a82f02b7c1cf343dedf0a7e885",
+                "revision": "sha256:6f67a5d347f3d05198f92cfce346d30051a544427ec8922b9de64432c3ffe316"
+              }
+            ],
+            "selection": {
+              "cue": "How should model loading overhead be separated from per request latency?",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 549459,
+            "sha256": "d211e3d91c2a81d9eb94f71c8e5bf368af4073b03f869d87154e196aefb8dcca"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev04",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev05",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev05",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 429290,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.storage_io_measurement",
+                "ref": "package:inferrsi.common.skills.storage_io_measurement",
+                "resource_count": 343,
+                "resource_hashes_sha256": "350d9c963ece5f1e1359ec238f4f804912b7bf0f08743b9cd4b6207583d52515",
+                "revision": "sha256:a57650b2bcaaac84bce2725a49eadb8c17533c82a1125b5ed5d840df5fca201c"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.bound_windowed_attention_replay",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb70e2e746f7ba833a696431c6a4deda830188acf5c4eafce6b0008818f8190a",
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_operator_profiling",
+                "ref": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "resource_count": 343,
+                "resource_hashes_sha256": "6f65a5d36f2fb4ec6097cb8aab45c4b2266f8808da5feed0f1b3800bca65395c",
+                "revision": "sha256:d467d8d6c966efe5dce163a74e15cd8714a6c9a7a008374e839cbe13973f2f88"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_distributed_parallel",
+                "ref": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ecb6d382307b933beb053b3c380a5bb6ba87c13792c428ad4131c78f5a9bcd04",
+                "revision": "sha256:66dab07081fad44b9cfe69b9c0f7a8f1dbb2c3fa19157ac99e4eee9a6306f605"
+              }
+            ],
+            "selection": {
+              "cue": "fio iodepth achieved depth",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 430369,
+            "sha256": "5c6e7284460b166b03a1bdded874ed826f0d5fe743a037881bc4ac333d392fcc"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev05",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 51533,
+            "context_omissions": [
+              {
+                "available_bytes": 14003,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "required_bytes": 41017
+              },
+              {
+                "available_bytes": 14003,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "required_bytes": 24091
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "63b883c10a8d3e7d918abfc7175b72e2b401ff1e02697f94b8382ec1c8f369b7",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md"
+                  }
+                ],
+                "content_bytes": 8813,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/host/storage-io-measurement/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.storage_io_measurement",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/host/storage-io-measurement/SKILL.md",
+                    "role": "core",
+                    "sha256": "63b883c10a8d3e7d918abfc7175b72e2b401ff1e02697f94b8382ec1c8f369b7"
+                  }
+                ],
+                "revision": "sha256:a57650b2bcaaac84bce2725a49eadb8c17533c82a1125b5ed5d840df5fca201c",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "63b883c10a8d3e7d918abfc7175b72e2b401ff1e02697f94b8382ec1c8f369b7",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "4703beb6e5194ef2ff2c0b5ec69f64e35417a4bc0dfb50a10c80a920b77ef3b5",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/validation.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md"
+                  }
+                ],
+                "content_bytes": 29732,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no equivalence run was performed.",
+                  "Applies to architectures with a demonstrated boundary dependency, not to arbitrary sliding-window stacks.",
+                  "Window sizes, boundary-layer index and ring lengths are model-config specific and are not transferable.",
+                  "Query-row compaction does not imply dropping historical keys; the two decisions are separate.",
+                  "Prompt logprobs are unrecoverable for the clipped region; the reference forces them off with a warning.",
+                  "The truncation path is off by default in the reference source.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md",
+                    "role": "core",
+                    "sha256": "4703beb6e5194ef2ff2c0b5ec69f64e35417a4bc0dfb50a10c80a920b77ef3b5"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "fd917bd5ac80bee9c9f4525f8e0300f6d67dc144e2a16632af42a67f3d53d098"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/validation.md",
+                    "role": "reference",
+                    "sha256": "60e71e70c5519bcdf6c171eec2be2f342e8801bedd746d6005f11adc646dfaff"
+                  }
+                ],
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "4703beb6e5194ef2ff2c0b5ec69f64e35417a4bc0dfb50a10c80a920b77ef3b5",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/validation.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "401981cf7828edcab0c9ea882f7ecd83badfa3c2a95006e912cf99d67b3d6a12",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/SKILL.md"
+                  }
+                ],
+                "content_bytes": 12988,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/ascend/ascend-operator-profiling/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/ascend-operator-profiling/SKILL.md",
+                    "role": "core",
+                    "sha256": "401981cf7828edcab0c9ea882f7ecd83badfa3c2a95006e912cf99d67b3d6a12"
+                  }
+                ],
+                "revision": "sha256:d467d8d6c966efe5dce163a74e15cd8714a6c9a7a008374e839cbe13973f2f88",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "401981cf7828edcab0c9ea882f7ecd83badfa3c2a95006e912cf99d67b3d6a12",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 458087,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.storage_io_measurement",
+                "ref": "package:inferrsi.common.skills.storage_io_measurement",
+                "resource_count": 343,
+                "resource_hashes_sha256": "350d9c963ece5f1e1359ec238f4f804912b7bf0f08743b9cd4b6207583d52515",
+                "revision": "sha256:a57650b2bcaaac84bce2725a49eadb8c17533c82a1125b5ed5d840df5fca201c"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.bound_windowed_attention_replay",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb70e2e746f7ba833a696431c6a4deda830188acf5c4eafce6b0008818f8190a",
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_operator_profiling",
+                "ref": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "resource_count": 343,
+                "resource_hashes_sha256": "6f65a5d36f2fb4ec6097cb8aab45c4b2266f8808da5feed0f1b3800bca65395c",
+                "revision": "sha256:d467d8d6c966efe5dce163a74e15cd8714a6c9a7a008374e839cbe13973f2f88"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_distributed_parallel",
+                "ref": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ecb6d382307b933beb053b3c380a5bb6ba87c13792c428ad4131c78f5a9bcd04",
+                "revision": "sha256:66dab07081fad44b9cfe69b9c0f7a8f1dbb2c3fa19157ac99e4eee9a6306f605"
+              }
+            ],
+            "selection": {
+              "cue": "fio iodepth achieved depth",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 521491,
+            "sha256": "8a3a1fa6c43063fe3165b30705447e22dcdc404bfb018739b401cf81e5f48061"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev05",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev06",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev06",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 431262,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_device_health",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.storage_io_measurement",
+                "ref": "package:inferrsi.common.skills.storage_io_measurement",
+                "resource_count": 343,
+                "resource_hashes_sha256": "350d9c963ece5f1e1359ec238f4f804912b7bf0f08743b9cd4b6207583d52515",
+                "revision": "sha256:a57650b2bcaaac84bce2725a49eadb8c17533c82a1125b5ed5d840df5fca201c"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.cann_attribution_overlap",
+                "ref": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c0bf9f5ab49059fd46fbee3e58bc0fac2819bb9d727e921d2362e618e92f996c",
+                "revision": "sha256:425a1e6d432ff3aa077b6974b41c66fb0a88be4ae88b8ba7a827071dfd8f5f10"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "ref": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "resource_count": 343,
+                "resource_hashes_sha256": "3f8f5e717bf501c4d94bfb3ac40c15393b07dd519c9170343aeb54445fbe933c",
+                "revision": "sha256:9028b6d32dcbe068af6fc777e80162bc815835f63657dc536f93d9e8ce5599f3"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.npu_skill_factory",
+                "ref": "package:inferrsi.common.skills.npu_skill_factory",
+                "resource_count": 343,
+                "resource_hashes_sha256": "7cde43221171eca8e11527545216e348d8c464438bec7458aab055843bb440cf",
+                "revision": "sha256:c789873dd4195bee85724a481228f51852a09c870d630ad1e828901037f875b7"
+              }
+            ],
+            "selection": {
+              "cue": "How can I attribute disk traffic to the loader process rather than the whole device?",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 436719,
+            "sha256": "6ed7e4937400c60c3bda007715df3506107846161efe870350057d4949a91883"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev06",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 35244,
+            "context_omissions": [
+              {
+                "available_bytes": 30292,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "required_bytes": 41017
+              },
+              {
+                "available_bytes": 30292,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.npu_skill_factory",
+                "required_bytes": 65054
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "63b883c10a8d3e7d918abfc7175b72e2b401ff1e02697f94b8382ec1c8f369b7",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md"
+                  }
+                ],
+                "content_bytes": 8813,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/host/storage-io-measurement/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.storage_io_measurement",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/host/storage-io-measurement/SKILL.md",
+                    "role": "core",
+                    "sha256": "63b883c10a8d3e7d918abfc7175b72e2b401ff1e02697f94b8382ec1c8f369b7"
+                  }
+                ],
+                "revision": "sha256:a57650b2bcaaac84bce2725a49eadb8c17533c82a1125b5ed5d840df5fca201c",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "63b883c10a8d3e7d918abfc7175b72e2b401ff1e02697f94b8382ec1c8f369b7",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "0e6f094dc554abea12e07def50f20b233b47127779a8aefad595b4ab1efd24c2",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md"
+                  }
+                ],
+                "content_bytes": 8972,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/ascend/cann-attribution-overlap/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md",
+                    "role": "core",
+                    "sha256": "0e6f094dc554abea12e07def50f20b233b47127779a8aefad595b4ab1efd24c2"
+                  }
+                ],
+                "revision": "sha256:425a1e6d432ff3aa077b6974b41c66fb0a88be4ae88b8ba7a827071dfd8f5f10",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "0e6f094dc554abea12e07def50f20b233b47127779a8aefad595b4ab1efd24c2",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "8428df7157af024fa17cfe0b3818ed54bd9f468306dda46eb8d4f0edf4f3f230",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/metadata.json",
+                      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md"
+                  }
+                ],
+                "content_bytes": 17459,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md",
+                    "role": "core",
+                    "sha256": "8428df7157af024fa17cfe0b3818ed54bd9f468306dda46eb8d4f0edf4f3f230"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "80cba4d3a884e7ee2618326cf74b9f9c3a34e4d9cf443707f6bbe673f2d5ba65"
+                  }
+                ],
+                "revision": "sha256:9028b6d32dcbe068af6fc777e80162bc815835f63657dc536f93d9e8ce5599f3",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "8428df7157af024fa17cfe0b3818ed54bd9f468306dda46eb8d4f0edf4f3f230",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/metadata.json",
+                      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 463110,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_device_health",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.storage_io_measurement",
+                "ref": "package:inferrsi.common.skills.storage_io_measurement",
+                "resource_count": 343,
+                "resource_hashes_sha256": "350d9c963ece5f1e1359ec238f4f804912b7bf0f08743b9cd4b6207583d52515",
+                "revision": "sha256:a57650b2bcaaac84bce2725a49eadb8c17533c82a1125b5ed5d840df5fca201c"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.cann_attribution_overlap",
+                "ref": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c0bf9f5ab49059fd46fbee3e58bc0fac2819bb9d727e921d2362e618e92f996c",
+                "revision": "sha256:425a1e6d432ff3aa077b6974b41c66fb0a88be4ae88b8ba7a827071dfd8f5f10"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "ref": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "resource_count": 343,
+                "resource_hashes_sha256": "3f8f5e717bf501c4d94bfb3ac40c15393b07dd519c9170343aeb54445fbe933c",
+                "revision": "sha256:9028b6d32dcbe068af6fc777e80162bc815835f63657dc536f93d9e8ce5599f3"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.npu_skill_factory",
+                "ref": "package:inferrsi.common.skills.npu_skill_factory",
+                "resource_count": 343,
+                "resource_hashes_sha256": "7cde43221171eca8e11527545216e348d8c464438bec7458aab055843bb440cf",
+                "revision": "sha256:c789873dd4195bee85724a481228f51852a09c870d630ad1e828901037f875b7"
+              }
+            ],
+            "selection": {
+              "cue": "How can I attribute disk traffic to the loader process rather than the whole device?",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 513297,
+            "sha256": "467b97551281ea312611214ece5f7200de1da656c9af15e5d2aa12084eb757df"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev06",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev07",
+          "mode": "search",
+          "output": {
+            "ordered_items": [
+              {
+                "id": "inferrsi.ascend.skills.ascendc_tiling_design"
+              }
+            ],
+            "serialized_bytes": 4033,
+            "sha256": "f87cc18112737157abb8fe1f4c27a0ceaf9656fca6ee8fbc84e760b14fd69a5d"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev07",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 429624,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_device_health",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascendc_tiling_design",
+                "ref": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "resource_count": 343,
+                "resource_hashes_sha256": "caa4787f895186508b8b6f863b2d03a830364585ffd65f0c098be319c7d54cad",
+                "revision": "sha256:9c90d2bc30d84cf0c99a0382d1e61b01c1210efe51d16ddb81c25afcb7c04e4b"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.assess_split_backward_schedule",
+                "ref": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "resource_count": 343,
+                "resource_hashes_sha256": "e3cbdc3297e17bf0af02a32b0fd231cb1f645fd7d21844b7764cd1b61cdc9243",
+                "revision": "sha256:cec60c01e6c992705c02cc68ab17a2a437ecf2cfbd03489b45fc98ac7692b044"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_operator_tuning",
+                "ref": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "resource_count": 343,
+                "resource_hashes_sha256": "1f4ba53716e4b333251713ef826742cb9596f086412d26e2715a3e71c75d6211",
+                "revision": "sha256:01c26a7a1203f0d8fea8e035e875c1c0e59bbbcc68a666d497a44071b05c7323"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "ref": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "resource_count": 343,
+                "resource_hashes_sha256": "7b4026d946b56b7fdafc6e086144d5ab04f6951f233afd5cfbcf003c04d05287",
+                "revision": "sha256:5de043dd66cc20bccdbb002ceee7efffef2136e4296e038406bf1e9884b1ec77"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sd_cpp_residency",
+                "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c83d7a76dffa5a3d97e932a3eac4602ad82e20babb09d9942504fff6b4db53d6",
+                "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde"
+              }
+            ],
+            "selection": {
+              "cue": "UB split buffer branch matrix",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 434145,
+            "sha256": "b176000866b7d392ad408fae0f6fef461f98a6c16a159c9955a0f87ee57a027e"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev07",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 64663,
+            "context_omissions": [],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "93aa1b775ced304cebfe6c4dec0503ddb2260ecb3cb01edc17ce329446c5ce2b",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md"
+                  }
+                ],
+                "content_bytes": 11860,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/ascend/ascendc-tiling-design/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md",
+                    "role": "core",
+                    "sha256": "93aa1b775ced304cebfe6c4dec0503ddb2260ecb3cb01edc17ce329446c5ce2b"
+                  }
+                ],
+                "revision": "sha256:9c90d2bc30d84cf0c99a0382d1e61b01c1210efe51d16ddb81c25afcb7c04e4b",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "93aa1b775ced304cebfe6c4dec0503ddb2260ecb3cb01edc17ce329446c5ce2b",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "1e64f28ea71672fb53c56f6108b87864738c6d03d6dcd32eedaef9e0625e96e3",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/SKILL.md"
+                  }
+                ],
+                "content_bytes": 13729,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/assess-split-backward-schedule/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/assess-split-backward-schedule/SKILL.md",
+                    "role": "core",
+                    "sha256": "1e64f28ea71672fb53c56f6108b87864738c6d03d6dcd32eedaef9e0625e96e3"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/assess-split-backward-schedule/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "61b1609f71c37b61d136b8c4d39f793de87cffb1cf31c8ec737a1c3fa6fb1b4e"
+                  }
+                ],
+                "revision": "sha256:cec60c01e6c992705c02cc68ab17a2a437ecf2cfbd03489b45fc98ac7692b044",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "1e64f28ea71672fb53c56f6108b87864738c6d03d6dcd32eedaef9e0625e96e3",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "38b151bc6b4e9734c6fa41efb7a30343559400c0c75bbdd26bd8a968b8ea4592",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/SKILL.md"
+                  }
+                ],
+                "content_bytes": 12236,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/ascend/ascend-operator-tuning/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/ascend-operator-tuning/SKILL.md",
+                    "role": "core",
+                    "sha256": "38b151bc6b4e9734c6fa41efb7a30343559400c0c75bbdd26bd8a968b8ea4592"
+                  }
+                ],
+                "revision": "sha256:01c26a7a1203f0d8fea8e035e875c1c0e59bbbcc68a666d497a44071b05c7323",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "38b151bc6b4e9734c6fa41efb7a30343559400c0c75bbdd26bd8a968b8ea4592",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "a95569bcc0da6e715fd74783a6a1e0489e92b680c582f15dafbf98df589d505a",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/SKILL.md",
+                      "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/metadata.json",
+                      "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/execution-lifetime.md",
+                      "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/source-map.md",
+                      "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/validation.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/SKILL.md"
+                  }
+                ],
+                "content_bytes": 13683,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local native-backend, model, service or efficacy validation.",
+                  "Static content checks do not promote this candidate or establish installed API conformance.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/SKILL.md",
+                    "role": "core",
+                    "sha256": "a95569bcc0da6e715fd74783a6a1e0489e92b680c582f15dafbf98df589d505a"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/source-map.md",
+                    "role": "reference",
+                    "sha256": "798b9cb89ed6b885eb47464ca9f137d2802a54571b9327e631d7742d3e5abd5d"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/execution-lifetime.md",
+                    "role": "reference",
+                    "sha256": "c4bc41e08aca68f661325f61e6f0db1760a079dc362f295c847bdc374fee3bf6"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/validation.md",
+                    "role": "reference",
+                    "sha256": "ebfc00bf7d6ab0a1dc415953ae84d56ccdc713d5614aff85e433c4a843793bff"
+                  }
+                ],
+                "revision": "sha256:5de043dd66cc20bccdbb002ceee7efffef2136e4296e038406bf1e9884b1ec77",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "a95569bcc0da6e715fd74783a6a1e0489e92b680c582f15dafbf98df589d505a",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/SKILL.md",
+                      "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/metadata.json",
+                      "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/execution-lifetime.md",
+                      "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/source-map.md",
+                      "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/validation.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md"
+                  }
+                ],
+                "content_bytes": 13155,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; native inference, allocator behavior, image quality and performance were not run.",
+                  "Eighteen bounded readset paths plus one separately identity-checked README match the pinned Git commit/tree blob IDs and sizes (19 blobs total); header coverage is limited to selected owner/state declarations, and this does not authenticate every archive member or submodule contents.",
+                  "GGML and other submodules are not expanded; matching private GGML headers are required even for a system library.",
+                  "CPU bypasses the inspected accelerator capacity gate and segmentation eligibility, so CPU checks cannot qualify accelerator memory budgets or prefetch.",
+                  "Video/audio pipelines, detailed model kernels, approximate denoising caches and full multi-device execution remain outside this readset.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json",
+                    "reason": "not_context_resource"
+                  },
+                  {
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md",
+                    "reason": "context_limit"
+                  },
+                  {
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md",
+                    "reason": "context_limit"
+                  },
+                  {
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json",
+                    "reason": "context_limit"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md",
+                    "role": "core",
+                    "sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "b67affbd7d237d82b7f51be49f1537e9820fe56ec70a71ee76809f9e599ca693"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE",
+                    "role": "reference",
+                    "sha256": "b53fa08f515cb5a6fff7b9fd8fcd0961a4b80df29d74372d25e1f9171aa042ee"
+                  }
+                ],
+                "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 454958,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_device_health",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascendc_tiling_design",
+                "ref": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "resource_count": 343,
+                "resource_hashes_sha256": "caa4787f895186508b8b6f863b2d03a830364585ffd65f0c098be319c7d54cad",
+                "revision": "sha256:9c90d2bc30d84cf0c99a0382d1e61b01c1210efe51d16ddb81c25afcb7c04e4b"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.assess_split_backward_schedule",
+                "ref": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "resource_count": 343,
+                "resource_hashes_sha256": "e3cbdc3297e17bf0af02a32b0fd231cb1f645fd7d21844b7764cd1b61cdc9243",
+                "revision": "sha256:cec60c01e6c992705c02cc68ab17a2a437ecf2cfbd03489b45fc98ac7692b044"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_operator_tuning",
+                "ref": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "resource_count": 343,
+                "resource_hashes_sha256": "1f4ba53716e4b333251713ef826742cb9596f086412d26e2715a3e71c75d6211",
+                "revision": "sha256:01c26a7a1203f0d8fea8e035e875c1c0e59bbbcc68a666d497a44071b05c7323"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "ref": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "resource_count": 343,
+                "resource_hashes_sha256": "7b4026d946b56b7fdafc6e086144d5ab04f6951f233afd5cfbcf003c04d05287",
+                "revision": "sha256:5de043dd66cc20bccdbb002ceee7efffef2136e4296e038406bf1e9884b1ec77"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sd_cpp_residency",
+                "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c83d7a76dffa5a3d97e932a3eac4602ad82e20babb09d9942504fff6b4db53d6",
+                "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde"
+              }
+            ],
+            "selection": {
+              "cue": "UB split buffer branch matrix",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 543313,
+            "sha256": "9176aff7fe535f60d6ba945a1c2f49d18d9699b455ae77e26cd5e921fbe676a8"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev07",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 85270,
+            "no_match": false,
+            "omissions": [],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascendc_tiling_design",
+                "ref": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "resource_count": 343,
+                "resource_hashes_sha256": "caa4787f895186508b8b6f863b2d03a830364585ffd65f0c098be319c7d54cad",
+                "revision": "sha256:9c90d2bc30d84cf0c99a0382d1e61b01c1210efe51d16ddb81c25afcb7c04e4b"
+              }
+            ],
+            "serialized_bytes": 85587,
+            "sha256": "ae41b0ffbd7acbffb95d3f5bfdc30e5102d9dce1864073f693da14604f50c774"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev08",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev08",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 85600,
+            "no_match": false,
+            "omissions": [],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "ref": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "resource_count": 343,
+                "resource_hashes_sha256": "3f8f5e717bf501c4d94bfb3ac40c15393b07dd519c9170343aeb54445fbe933c",
+                "revision": "sha256:9028b6d32dcbe068af6fc777e80162bc815835f63657dc536f93d9e8ce5599f3"
+              }
+            ],
+            "selection": {
+              "cue": "multi_stream_lazy_reclaim",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 86057,
+            "sha256": "8e4a81b12208ac5ede8768f9ae00f5ca91e9355d2aac306524106007c18e98c0"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev08",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 17459,
+            "context_omissions": [],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "8428df7157af024fa17cfe0b3818ed54bd9f468306dda46eb8d4f0edf4f3f230",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/metadata.json",
+                      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md"
+                  }
+                ],
+                "content_bytes": 17459,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md",
+                    "role": "core",
+                    "sha256": "8428df7157af024fa17cfe0b3818ed54bd9f468306dda46eb8d4f0edf4f3f230"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "80cba4d3a884e7ee2618326cf74b9f9c3a34e4d9cf443707f6bbe673f2d5ba65"
+                  }
+                ],
+                "revision": "sha256:9028b6d32dcbe068af6fc777e80162bc815835f63657dc536f93d9e8ce5599f3",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "8428df7157af024fa17cfe0b3818ed54bd9f468306dda46eb8d4f0edf4f3f230",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/metadata.json",
+                      "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 89687,
+            "no_match": false,
+            "omissions": [],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "ref": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "resource_count": 343,
+                "resource_hashes_sha256": "3f8f5e717bf501c4d94bfb3ac40c15393b07dd519c9170343aeb54445fbe933c",
+                "revision": "sha256:9028b6d32dcbe068af6fc777e80162bc815835f63657dc536f93d9e8ce5599f3"
+              }
+            ],
+            "selection": {
+              "cue": "multi_stream_lazy_reclaim",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 111101,
+            "sha256": "4fd1f7e989385f0ed48d163789ab382436741d2bfb78a7595f81d75b66950330"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev08",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev09",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev09",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 432011,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.kv_tier_management",
+                "ref": "package:inferrsi.common.skills.kv_tier_management",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ee42b55fa8a404149ecd1dccc32a892566ee8867710531bc0af553409758a019",
+                "revision": "sha256:5816c2e526a71152486af2230ace24d27b54baeb373bf5d27e353ccd44f640aa"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.bound_windowed_attention_replay",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb70e2e746f7ba833a696431c6a4deda830188acf5c4eafce6b0008818f8190a",
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sd_cpp_residency",
+                "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c83d7a76dffa5a3d97e932a3eac4602ad82e20babb09d9942504fff6b4db53d6",
+                "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.diffusers_invocation_state",
+                "ref": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "resource_count": 343,
+                "resource_hashes_sha256": "64b10a60e44276bc4a20b3441f65afd929341e8558fde32491ba27a20fde4020",
+                "revision": "sha256:a8529f2c6bf5f84ce48ba0d6426f1bf1bc6fce172cd0a599b6a149878e3b4be4"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.generation_cache_lifecycle",
+                "ref": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "resource_count": 343,
+                "resource_hashes_sha256": "052e200da2a160661cf44647685c70e52feec4b72fc0f2e7b34a10ce54710e03",
+                "revision": "sha256:22411ef79b30ddd4f8b01a7ed5d00fc4fc75a3c9f1e4d31f56f2e440d9b09d2f"
+              }
+            ],
+            "selection": {
+              "cue": "LMCache skip_l1 retain",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 435629,
+            "sha256": "4acef85df250bc9967208b838f534e75b4d9706be8e1dda8ec71bb7409eeca57"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev09",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 61835,
+            "context_omissions": [
+              {
+                "available_bytes": 3701,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "required_bytes": 10129
+              },
+              {
+                "available_bytes": 3701,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "required_bytes": 10146
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md"
+                  }
+                ],
+                "content_bytes": 10812,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/repositories/lmcache/kv-tier-management/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.kv_tier_management",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md",
+                    "role": "core",
+                    "sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27"
+                  }
+                ],
+                "revision": "sha256:5816c2e526a71152486af2230ace24d27b54baeb373bf5d27e353ccd44f640aa",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "4703beb6e5194ef2ff2c0b5ec69f64e35417a4bc0dfb50a10c80a920b77ef3b5",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/validation.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md"
+                  }
+                ],
+                "content_bytes": 29732,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no equivalence run was performed.",
+                  "Applies to architectures with a demonstrated boundary dependency, not to arbitrary sliding-window stacks.",
+                  "Window sizes, boundary-layer index and ring lengths are model-config specific and are not transferable.",
+                  "Query-row compaction does not imply dropping historical keys; the two decisions are separate.",
+                  "Prompt logprobs are unrecoverable for the clipped region; the reference forces them off with a warning.",
+                  "The truncation path is off by default in the reference source.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md",
+                    "role": "core",
+                    "sha256": "4703beb6e5194ef2ff2c0b5ec69f64e35417a4bc0dfb50a10c80a920b77ef3b5"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "fd917bd5ac80bee9c9f4525f8e0300f6d67dc144e2a16632af42a67f3d53d098"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/validation.md",
+                    "role": "reference",
+                    "sha256": "60e71e70c5519bcdf6c171eec2be2f342e8801bedd746d6005f11adc646dfaff"
+                  }
+                ],
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "4703beb6e5194ef2ff2c0b5ec69f64e35417a4bc0dfb50a10c80a920b77ef3b5",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/validation.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md"
+                  }
+                ],
+                "content_bytes": 21291,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; native inference, allocator behavior, image quality and performance were not run.",
+                  "Eighteen bounded readset paths plus one separately identity-checked README match the pinned Git commit/tree blob IDs and sizes (19 blobs total); header coverage is limited to selected owner/state declarations, and this does not authenticate every archive member or submodule contents.",
+                  "GGML and other submodules are not expanded; matching private GGML headers are required even for a system library.",
+                  "CPU bypasses the inspected accelerator capacity gate and segmentation eligibility, so CPU checks cannot qualify accelerator memory budgets or prefetch.",
+                  "Video/audio pipelines, detailed model kernels, approximate denoising caches and full multi-device execution remain outside this readset.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json",
+                    "reason": "not_context_resource"
+                  },
+                  {
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json",
+                    "reason": "not_relevant"
+                  },
+                  {
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md",
+                    "reason": "not_relevant"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md",
+                    "role": "core",
+                    "sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "b67affbd7d237d82b7f51be49f1537e9820fe56ec70a71ee76809f9e599ca693"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE",
+                    "role": "reference",
+                    "sha256": "b53fa08f515cb5a6fff7b9fd8fcd0961a4b80df29d74372d25e1f9171aa042ee"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md",
+                    "role": "reference",
+                    "sha256": "7b537cb972347dd68e303960f142892346d8b19b884e4d08016ea36a23bdaa7a"
+                  }
+                ],
+                "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 462705,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.kv_tier_management",
+                "ref": "package:inferrsi.common.skills.kv_tier_management",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ee42b55fa8a404149ecd1dccc32a892566ee8867710531bc0af553409758a019",
+                "revision": "sha256:5816c2e526a71152486af2230ace24d27b54baeb373bf5d27e353ccd44f640aa"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.bound_windowed_attention_replay",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb70e2e746f7ba833a696431c6a4deda830188acf5c4eafce6b0008818f8190a",
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sd_cpp_residency",
+                "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c83d7a76dffa5a3d97e932a3eac4602ad82e20babb09d9942504fff6b4db53d6",
+                "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.diffusers_invocation_state",
+                "ref": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "resource_count": 343,
+                "resource_hashes_sha256": "64b10a60e44276bc4a20b3441f65afd929341e8558fde32491ba27a20fde4020",
+                "revision": "sha256:a8529f2c6bf5f84ce48ba0d6426f1bf1bc6fce172cd0a599b6a149878e3b4be4"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.generation_cache_lifecycle",
+                "ref": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "resource_count": 343,
+                "resource_hashes_sha256": "052e200da2a160661cf44647685c70e52feec4b72fc0f2e7b34a10ce54710e03",
+                "revision": "sha256:22411ef79b30ddd4f8b01a7ed5d00fc4fc75a3c9f1e4d31f56f2e440d9b09d2f"
+              }
+            ],
+            "selection": {
+              "cue": "LMCache skip_l1 retain",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 541917,
+            "sha256": "1804ed714dd8eeb38a8f1d0fbcfa3aebac8202a23103b285a84c5a814f47194b"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev09",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev10",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev10",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 85740,
+            "no_match": false,
+            "omissions": [],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.checkpoint_recovery_consistency",
+                "ref": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "15c906f896eab2a681eb9bc9e42ee29d11b0594a1e1aeacbf6f414233209d1fa",
+                "revision": "sha256:73f77f89e298ce8724258cb3d74315438c036297bb38c63aac3d7dc95473ce53"
+              }
+            ],
+            "selection": {
+              "cue": "find_latest",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 86183,
+            "sha256": "5ba7d7ecbdb98e25ad2e25f1ce1c8764cef738f47d5657fc9e46a655b886b8a8"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev10",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 14400,
+            "context_omissions": [],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "6b7db1eb2780dec27858cb7aa0a3c093c5b5dfa4c30261aebfeaa255213ff688",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/references/dlrover-persistence-replicas.md",
+                      "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/SKILL.md"
+                  }
+                ],
+                "content_bytes": 14400,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/checkpoint-recovery-consistency/metadata.json",
+                    "reason": "not_context_resource"
+                  },
+                  {
+                    "path": "data/skill_library/methods/checkpoint-recovery-consistency/references/dlrover-persistence-replicas.md",
+                    "reason": "not_relevant"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/checkpoint-recovery-consistency/SKILL.md",
+                    "role": "core",
+                    "sha256": "6b7db1eb2780dec27858cb7aa0a3c093c5b5dfa4c30261aebfeaa255213ff688"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/checkpoint-recovery-consistency/references/sources.md",
+                    "role": "reference",
+                    "sha256": "9425b1aab963a2c8d35f3db897bd15ee69699bf5b0d0c8425c3c8bba3e9a0233"
+                  }
+                ],
+                "revision": "sha256:73f77f89e298ce8724258cb3d74315438c036297bb38c63aac3d7dc95473ce53",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "6b7db1eb2780dec27858cb7aa0a3c093c5b5dfa4c30261aebfeaa255213ff688",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/references/dlrover-persistence-replicas.md",
+                      "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 90211,
+            "no_match": false,
+            "omissions": [],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.checkpoint_recovery_consistency",
+                "ref": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "15c906f896eab2a681eb9bc9e42ee29d11b0594a1e1aeacbf6f414233209d1fa",
+                "revision": "sha256:73f77f89e298ce8724258cb3d74315438c036297bb38c63aac3d7dc95473ce53"
+              }
+            ],
+            "selection": {
+              "cue": "find_latest",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 108906,
+            "sha256": "38c065f634ce4b1297325fc9ccbfff2dac832b28624e31e464ee6995c2b89d10"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev10",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev11",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev11",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 436986,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "ref": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "resource_count": 343,
+                "resource_hashes_sha256": "db84ef8ccf72ef81606670b384a3814297f9c999a5ffa25b05b4f0d10c500ebd",
+                "revision": "sha256:d8a8775c87f022125906e0ec25c0c57a083e272be5b339d5cdeb7a1704c0e4db"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.cuda.skills.reason_hopper_kernels",
+                "ref": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "resource_count": 343,
+                "resource_hashes_sha256": "4dfd3cbff82440ce22e8f16b14e6e75d25a34144661b077e21aed02a856de902",
+                "revision": "sha256:4d0628c3fd9cbcc647ec30d922166044f3544326f54f793779692d731e2a5b77"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.bound_windowed_attention_replay",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb70e2e746f7ba833a696431c6a4deda830188acf5c4eafce6b0008818f8190a",
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_llama_runtime_state",
+                "ref": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "resource_count": 343,
+                "resource_hashes_sha256": "03e86bd3af77f17d6c7ad2933534524ab7fe5993396a7fe9748a6014496a34b5",
+                "revision": "sha256:a3c3d3244cda589e8e7809e56f43b7bfff1852def8dd26fb03a9d29e39bca3dd"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.npu_skill_factory",
+                "ref": "package:inferrsi.common.skills.npu_skill_factory",
+                "resource_count": 343,
+                "resource_hashes_sha256": "7cde43221171eca8e11527545216e348d8c464438bec7458aab055843bb440cf",
+                "revision": "sha256:c789873dd4195bee85724a481228f51852a09c870d630ad1e828901037f875b7"
+              }
+            ],
+            "selection": {
+              "cue": "Why can larger attention tiles cause register spills and lower parallelism?",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 441258,
+            "sha256": "1fd164121d0001fd73173e443cca13d1fa3b40f69b4b39922e0d7c1459331cd3"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev11",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 56835,
+            "context_omissions": [
+              {
+                "available_bytes": 8701,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "required_bytes": 154821
+              },
+              {
+                "available_bytes": 8701,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "required_bytes": 29732
+              },
+              {
+                "available_bytes": 8701,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "required_bytes": 24572
+              },
+              {
+                "available_bytes": 8701,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.npu_skill_factory",
+                "required_bytes": 65054
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "a25738dbb3b9daa15c5827cb536a20fe13f7c8b9dd14aa3fb7cfa7e132879431",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/metadata.json",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/mechanisms.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/INTEGRATION_NOTES.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/NEXT_GATE_LEADS.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/README.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/io-reduction-can-plateau.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/overlap-and-fp8-counter-signals.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/partitioning-after-fusion.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/knowledge/attention-cost-model-boundaries.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/SKILL.md"
+                  }
+                ],
+                "content_bytes": 56835,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/SKILL.md",
+                    "role": "core",
+                    "sha256": "a25738dbb3b9daa15c5827cb536a20fe13f7c8b9dd14aa3fb7cfa7e132879431"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/mechanisms.md",
+                    "role": "reference",
+                    "sha256": "fe08b2292fa587be3874774ddf41063b84c579a0085df422ec6710d8aa90b8b7"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/io-reduction-can-plateau.md",
+                    "role": "reference",
+                    "sha256": "5cc35f514224ac6e6efd2da6db221bba468b2f4db66807295e67b240bcdfdba5"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/overlap-and-fp8-counter-signals.md",
+                    "role": "reference",
+                    "sha256": "80c8a60a6df55f3d84fab23e03567d8b38e23bcdbd63cc09881e46db1e2888ab"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/partitioning-after-fusion.md",
+                    "role": "reference",
+                    "sha256": "4d4cabe0991eab554dcfbd4fa9c8199e61fa4767709b0c4cd646e2588bb398a2"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/INTEGRATION_NOTES.md",
+                    "role": "reference",
+                    "sha256": "371b01ee201ef51fecbfd6da892d517aaba295930b9395025f03f34e21c030a1"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/knowledge/attention-cost-model-boundaries.md",
+                    "role": "reference",
+                    "sha256": "5ed8f888443e38ccb72545b6025447aab7b92936f78c151fed680e582126239b"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/NEXT_GATE_LEADS.md",
+                    "role": "reference",
+                    "sha256": "35b8b58ae5db02b2df51ccbe5cdf870a7383d17f345f9ee5eb7b0a4146dfc74b"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/README.md",
+                    "role": "reference",
+                    "sha256": "e0b1ec1b39c10067f5ddce58320851ad089568c4b2296a24777389097cb5b388"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/sources.md",
+                    "role": "reference",
+                    "sha256": "53436e0de4e6dda2d29426591bf61f8526378cdc6f9ec7967b88b1bcd4c61e6f"
+                  }
+                ],
+                "revision": "sha256:d8a8775c87f022125906e0ec25c0c57a083e272be5b339d5cdeb7a1704c0e4db",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "a25738dbb3b9daa15c5827cb536a20fe13f7c8b9dd14aa3fb7cfa7e132879431",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/metadata.json",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/mechanisms.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/INTEGRATION_NOTES.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/NEXT_GATE_LEADS.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/README.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/io-reduction-can-plateau.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/overlap-and-fp8-counter-signals.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/partitioning-after-fusion.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/knowledge/attention-cost-model-boundaries.md",
+                      "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 481490,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "ref": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "resource_count": 343,
+                "resource_hashes_sha256": "db84ef8ccf72ef81606670b384a3814297f9c999a5ffa25b05b4f0d10c500ebd",
+                "revision": "sha256:d8a8775c87f022125906e0ec25c0c57a083e272be5b339d5cdeb7a1704c0e4db"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.cuda.skills.reason_hopper_kernels",
+                "ref": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "resource_count": 343,
+                "resource_hashes_sha256": "4dfd3cbff82440ce22e8f16b14e6e75d25a34144661b077e21aed02a856de902",
+                "revision": "sha256:4d0628c3fd9cbcc647ec30d922166044f3544326f54f793779692d731e2a5b77"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.bound_windowed_attention_replay",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb70e2e746f7ba833a696431c6a4deda830188acf5c4eafce6b0008818f8190a",
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_llama_runtime_state",
+                "ref": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "resource_count": 343,
+                "resource_hashes_sha256": "03e86bd3af77f17d6c7ad2933534524ab7fe5993396a7fe9748a6014496a34b5",
+                "revision": "sha256:a3c3d3244cda589e8e7809e56f43b7bfff1852def8dd26fb03a9d29e39bca3dd"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.npu_skill_factory",
+                "ref": "package:inferrsi.common.skills.npu_skill_factory",
+                "resource_count": 343,
+                "resource_hashes_sha256": "7cde43221171eca8e11527545216e348d8c464438bec7458aab055843bb440cf",
+                "revision": "sha256:c789873dd4195bee85724a481228f51852a09c870d630ad1e828901037f875b7"
+              }
+            ],
+            "selection": {
+              "cue": "Why can larger attention tiles cause register spills and lower parallelism?",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 551101,
+            "sha256": "1d643e68fc4d9833f93a977651a8adccaac197560031521273be3068a034f603"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev11",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev12",
+          "mode": "search",
+          "output": {
+            "ordered_items": [
+              {
+                "id": "inferrsi.common.skills.os_process_placement"
+              }
+            ],
+            "serialized_bytes": 5621,
+            "sha256": "a9bee665f4da822ce6780d24bea786b256567b5e4b9ee6dfdae85de05c2e61e7"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev12",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 430402,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.os_process_placement",
+                "ref": "package:inferrsi.common.skills.os_process_placement",
+                "resource_count": 343,
+                "resource_hashes_sha256": "d0743478a2fb19183d4c1a8ab0cfcebf02a89512bbc35e09ade8f120a1e53afd",
+                "revision": "sha256:3ae1cc0f7d372b46a2e0a697a4485d63ca84d40afdba5414e9c6c1f2db15db7e"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.training_resource_budget",
+                "ref": "package:inferrsi.common.skills.training_resource_budget",
+                "resource_count": 343,
+                "resource_hashes_sha256": "846f754eb2d0d8f76837d8a532ebaed8f53bbbdb5f94695ebf7725de7a951497",
+                "revision": "sha256:0121883f97748a7cab0a53e5831988e4f34ca3d841987598f98d75506cad0734"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.host_memory_management",
+                "ref": "package:inferrsi.common.skills.host_memory_management",
+                "resource_count": 343,
+                "resource_hashes_sha256": "704c39f2041e99d4d66d9e4464c116274638e246db6a4b56e7c447dd711c198f",
+                "revision": "sha256:96c29b3da68c9493ce25421640f23cc2c15258005b6e006b6e9c2af285624fc3"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.cgroup_resource_controls",
+                "ref": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "resource_count": 343,
+                "resource_hashes_sha256": "16e727f435ae4d8b6f349404846e7106e958b83425751c86135d63770802114c",
+                "revision": "sha256:94ef2d4108261a8a911ec5ffc89edcdd6d7bda21f064f7c66701a8f663cdfea3"
+              }
+            ],
+            "selection": {
+              "cue": "NUMA CPU affinity placement",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 433035,
+            "sha256": "f0f0732feaf4882221188f4cbd4aca31c84af3f6f53bf6d86aad09f939315d1e"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev12",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 65309,
+            "context_omissions": [
+              {
+                "available_bytes": 48592,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.training_resource_budget",
+                "required_bytes": 74071
+              },
+              {
+                "available_bytes": 48592,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.host_memory_management",
+                "required_bytes": 49803
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "2003492e4d89fde0443b180d74933e764bd1979435e109c8d26e20eb8f6e5053",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/host/os-process-placement/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/host/os-process-placement/metadata.json",
+                      "src/inferrsi/data/skill_library/hardware/host/os-process-placement/references/llama-numa-thread-readset.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/host/os-process-placement/SKILL.md"
+                  }
+                ],
+                "content_bytes": 16944,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Only the existing llama.cpp NUMA/thread slice is normalized at 3e390af18de1131dda27913c00511725b4274c8a; other original sources remain unnormalized, with their identities and limits retained in the procedure/readsets.",
+                  "Linux NUMA source paths and separate configuration controls do not establish universally optimal thread counts, affinity combinations or measured placement gains.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/host/os-process-placement/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.os_process_placement",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/host/os-process-placement/SKILL.md",
+                    "role": "core",
+                    "sha256": "2003492e4d89fde0443b180d74933e764bd1979435e109c8d26e20eb8f6e5053"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/host/os-process-placement/references/llama-numa-thread-readset.md",
+                    "role": "reference",
+                    "sha256": "748b254b3b043a5ddb0f480eb4a566854d27eedd7b111499863e9bf0a5a6951e"
+                  }
+                ],
+                "revision": "sha256:3ae1cc0f7d372b46a2e0a697a4485d63ca84d40afdba5414e9c6c1f2db15db7e",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "2003492e4d89fde0443b180d74933e764bd1979435e109c8d26e20eb8f6e5053",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/host/os-process-placement/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/host/os-process-placement/metadata.json",
+                      "src/inferrsi/data/skill_library/hardware/host/os-process-placement/references/llama-numa-thread-readset.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/host/os-process-placement/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md"
+                  }
+                ],
+                "content_bytes": 41017,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Normalized provenance covers the existing llama.cpp loading slice at 3e390af18de1131dda27913c00511725b4274c8a and the fastsafetensors distributed-delivery slice at 219e2561a41ede2d5d8eba611c248bb0effbf580; other inherited claims remain unnormalized with original identities and limits retained.",
+                  "The readsets record static mechanisms and source recommendations, not validated startup, memory, placement or backend performance.",
+                  "The fastsafetensors delivery readset does not revalidate inherited I/O/save claims or qualify uneven scatter, accelerator/NPU compatibility, inter-node transport or agent efficacy.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/weight-loading-optimization/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                    "role": "core",
+                    "sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                    "role": "reference",
+                    "sha256": "6c640f28ee3f0253a17b81461c865cb9f947367002cb75bf72550c4572d02e97"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                    "role": "reference",
+                    "sha256": "5a39f14b102b0cee304541e011aa31e44e6b13527d1bee9987c3f3b66c064d3d"
+                  }
+                ],
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "fd1f5a7e6a55714730044155f5b38c1d700383d4e16c3e91963bbfbe1c773af5",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/SKILL.md"
+                  }
+                ],
+                "content_bytes": 7348,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/host/cgroup-resource-controls/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/host/cgroup-resource-controls/SKILL.md",
+                    "role": "core",
+                    "sha256": "fd1f5a7e6a55714730044155f5b38c1d700383d4e16c3e91963bbfbe1c773af5"
+                  }
+                ],
+                "revision": "sha256:94ef2d4108261a8a911ec5ffc89edcdd6d7bda21f064f7c66701a8f663cdfea3",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "fd1f5a7e6a55714730044155f5b38c1d700383d4e16c3e91963bbfbe1c773af5",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 460966,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.os_process_placement",
+                "ref": "package:inferrsi.common.skills.os_process_placement",
+                "resource_count": 343,
+                "resource_hashes_sha256": "d0743478a2fb19183d4c1a8ab0cfcebf02a89512bbc35e09ade8f120a1e53afd",
+                "revision": "sha256:3ae1cc0f7d372b46a2e0a697a4485d63ca84d40afdba5414e9c6c1f2db15db7e"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.training_resource_budget",
+                "ref": "package:inferrsi.common.skills.training_resource_budget",
+                "resource_count": 343,
+                "resource_hashes_sha256": "846f754eb2d0d8f76837d8a532ebaed8f53bbbdb5f94695ebf7725de7a951497",
+                "revision": "sha256:0121883f97748a7cab0a53e5831988e4f34ca3d841987598f98d75506cad0734"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.host_memory_management",
+                "ref": "package:inferrsi.common.skills.host_memory_management",
+                "resource_count": 343,
+                "resource_hashes_sha256": "704c39f2041e99d4d66d9e4464c116274638e246db6a4b56e7c447dd711c198f",
+                "revision": "sha256:96c29b3da68c9493ce25421640f23cc2c15258005b6e006b6e9c2af285624fc3"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.cgroup_resource_controls",
+                "ref": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "resource_count": 343,
+                "resource_hashes_sha256": "16e727f435ae4d8b6f349404846e7106e958b83425751c86135d63770802114c",
+                "revision": "sha256:94ef2d4108261a8a911ec5ffc89edcdd6d7bda21f064f7c66701a8f663cdfea3"
+              }
+            ],
+            "selection": {
+              "cue": "NUMA CPU affinity placement",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 540415,
+            "sha256": "fa83fe34560e480806391c460c276add522b96e93b1c95b26548e0cc0d81b5dc"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev12",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 85733,
+            "no_match": false,
+            "omissions": [],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.os_process_placement",
+                "ref": "package:inferrsi.common.skills.os_process_placement",
+                "resource_count": 343,
+                "resource_hashes_sha256": "d0743478a2fb19183d4c1a8ab0cfcebf02a89512bbc35e09ade8f120a1e53afd",
+                "revision": "sha256:3ae1cc0f7d372b46a2e0a697a4485d63ca84d40afdba5414e9c6c1f2db15db7e"
+              }
+            ],
+            "serialized_bytes": 86050,
+            "sha256": "936169ae69e7b168b2100db3e6913d2034ff12372bdfffbbcf3999b8d9e1e69e"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev13",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev13",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 433439,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.plan_activation_residency",
+                "ref": "package:inferrsi.common.skills.plan_activation_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "8141a002cba8cd3ffa905e29c6517fd935740140b897d2f96b65c724ac3d24b9",
+                "revision": "sha256:06df72aed0c6246c334dada5e335b85e49111753637f4fffa4ecb68da5ec3d43"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sd_cpp_residency",
+                "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c83d7a76dffa5a3d97e932a3eac4602ad82e20babb09d9942504fff6b4db53d6",
+                "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_h3_runtime_residency",
+                "ref": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "64ea29a64c41b0dac809d22e10aa650ac7d60cb008ab15d9e702161ce2ad63c2",
+                "revision": "sha256:601645ce381756514c931d83e0019b1901272e08c9fd4e0ae99318ef0d18405b"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.comfy_ui",
+                "ref": "package:inferrsi.common.skills.comfy_ui",
+                "resource_count": 343,
+                "resource_hashes_sha256": "66d8cb95bd1f2270900a3a81e6274cb2cc3c799229acc75d0676def6ef22f8a8",
+                "revision": "sha256:13e038ab45a19c1ceafc282b3504ddb63bba487b6a2669472f6f580d18ff6599"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "ref": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "resource_count": 343,
+                "resource_hashes_sha256": "8af3478b2f59c5fade35c614a9bfe4e4d53e8d5e9a5961870e8f4e62631cb937",
+                "revision": "sha256:0af6ac4cfd9d26bd90eeba094c3ceeed80e418dbf1872e6b1ffaa23b6b30e397"
+              }
+            ],
+            "selection": {
+              "cue": "gradient activation recomputation residency",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 435641,
+            "sha256": "44a1ed6da89de8f69fb33b0bb85f7aed356aaf4f2c796057dae9aebb715ad810"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev13",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 45716,
+            "context_omissions": [
+              {
+                "available_bytes": 19820,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "required_bytes": 36142
+              },
+              {
+                "available_bytes": 19820,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.comfy_ui",
+                "required_bytes": 45304
+              },
+              {
+                "available_bytes": 19820,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "required_bytes": 47942
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "542d8a1e6963f5019986e1fbd79d3a247a045a51487c09313026087ac457c15e",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/plan-activation-residency/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/plan-activation-residency/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/plan-activation-residency/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/plan-activation-residency/SKILL.md"
+                  }
+                ],
+                "content_bytes": 17414,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/plan-activation-residency/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.plan_activation_residency",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/plan-activation-residency/SKILL.md",
+                    "role": "core",
+                    "sha256": "542d8a1e6963f5019986e1fbd79d3a247a045a51487c09313026087ac457c15e"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/plan-activation-residency/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "5ff52655cfd06fcf970ed2a7d4d4fe1c31115b102af2d41497de0c74848ed2b3"
+                  }
+                ],
+                "revision": "sha256:06df72aed0c6246c334dada5e335b85e49111753637f4fffa4ecb68da5ec3d43",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "542d8a1e6963f5019986e1fbd79d3a247a045a51487c09313026087ac457c15e",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/plan-activation-residency/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/plan-activation-residency/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/plan-activation-residency/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/plan-activation-residency/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md"
+                  }
+                ],
+                "content_bytes": 28302,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; native inference, allocator behavior, image quality and performance were not run.",
+                  "Eighteen bounded readset paths plus one separately identity-checked README match the pinned Git commit/tree blob IDs and sizes (19 blobs total); header coverage is limited to selected owner/state declarations, and this does not authenticate every archive member or submodule contents.",
+                  "GGML and other submodules are not expanded; matching private GGML headers are required even for a system library.",
+                  "CPU bypasses the inspected accelerator capacity gate and segmentation eligibility, so CPU checks cannot qualify accelerator memory budgets or prefetch.",
+                  "Video/audio pipelines, detailed model kernels, approximate denoising caches and full multi-device execution remain outside this readset.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md",
+                    "role": "core",
+                    "sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "b67affbd7d237d82b7f51be49f1537e9820fe56ec70a71ee76809f9e599ca693"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE",
+                    "role": "reference",
+                    "sha256": "b53fa08f515cb5a6fff7b9fd8fcd0961a4b80df29d74372d25e1f9171aa042ee"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md",
+                    "role": "reference",
+                    "sha256": "7b537cb972347dd68e303960f142892346d8b19b884e4d08016ea36a23bdaa7a"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json",
+                    "role": "reference",
+                    "sha256": "af4830370b6eddc2d176a9ae14da4ab56a90b0fe048a01a89ee051df4365d8e9"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md",
+                    "role": "reference",
+                    "sha256": "5dcf5fa7bf734fbe13214f57d2dfb40ff8215d9cd34e072a9e615bf103b82fc2"
+                  }
+                ],
+                "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 469025,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.plan_activation_residency",
+                "ref": "package:inferrsi.common.skills.plan_activation_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "8141a002cba8cd3ffa905e29c6517fd935740140b897d2f96b65c724ac3d24b9",
+                "revision": "sha256:06df72aed0c6246c334dada5e335b85e49111753637f4fffa4ecb68da5ec3d43"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sd_cpp_residency",
+                "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c83d7a76dffa5a3d97e932a3eac4602ad82e20babb09d9942504fff6b4db53d6",
+                "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_h3_runtime_residency",
+                "ref": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "64ea29a64c41b0dac809d22e10aa650ac7d60cb008ab15d9e702161ce2ad63c2",
+                "revision": "sha256:601645ce381756514c931d83e0019b1901272e08c9fd4e0ae99318ef0d18405b"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.comfy_ui",
+                "ref": "package:inferrsi.common.skills.comfy_ui",
+                "resource_count": 343,
+                "resource_hashes_sha256": "66d8cb95bd1f2270900a3a81e6274cb2cc3c799229acc75d0676def6ef22f8a8",
+                "revision": "sha256:13e038ab45a19c1ceafc282b3504ddb63bba487b6a2669472f6f580d18ff6599"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "ref": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "resource_count": 343,
+                "resource_hashes_sha256": "8af3478b2f59c5fade35c614a9bfe4e4d53e8d5e9a5961870e8f4e62631cb937",
+                "revision": "sha256:0af6ac4cfd9d26bd90eeba094c3ceeed80e418dbf1872e6b1ffaa23b6b30e397"
+              }
+            ],
+            "selection": {
+              "cue": "gradient activation recomputation residency",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 527283,
+            "sha256": "a6c8fd92fd24a7304d4b207634068866b654d2e6d395240dae1bc74eccb9800d"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev13",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev14",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "dev14",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 432303,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_device_health",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_performance_analysis",
+                "ref": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "resource_count": 343,
+                "resource_hashes_sha256": "255c964b950c3e58c7a55001c35a22d48e386da437cea66ec2486a1f0412dd1f",
+                "revision": "sha256:f7ea7eb4d21d63a62a20b1a8b719277d4d2929907ff255af4acb1e30028b01dc"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.dense_performance_evidence",
+                "ref": "package:inferrsi.common.skills.dense_performance_evidence",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ee7e4bd7570b737eb14c98190545cee0fa11065e1b1a2d0cfe663928cfca0551",
+                "revision": "sha256:82e68013915ca052f96119da0a8d32d40d38d0aab792b04844c2e208d10706dc"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.performance_ceiling_analysis",
+                "ref": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c915247877787fcecc949d1d212ae1f18e3ba2a82f02b7c1cf343dedf0a7e885",
+                "revision": "sha256:6f67a5d347f3d05198f92cfce346d30051a544427ec8922b9de64432c3ffe316"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.claim_contract",
+                "ref": "package:inferrsi.common.skills.claim_contract",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c6481b673e503a8707f897b6e23c355888873b8818458e6a4b8e1fbfee9f9f8f",
+                "revision": "sha256:0f6955d51d86aa329bf643f37bf750084520c666360a64d6c10bfb678e64b8a2"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.inference_benchmark_design",
+                "ref": "package:inferrsi.common.skills.inference_benchmark_design",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb743a7fe0662feead981795dc60e2a5c9355ed876339f14ad7a47caab0bb670",
+                "revision": "sha256:6b0cbeaa5d409c04228883884e192f275cad4874c7f6b3039abef9fd443fafb5"
+              }
+            ],
+            "selection": {
+              "cue": "When does missing validator freeze make a performance claim inconclusive?",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 437750,
+            "sha256": "e074b2d4a400f80ae868b08968631f451801aa928ec7aec0cfe700f250ba4071"
+          },
+          "reciprocal_rank_at_5": 0.25,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev14",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 40799,
+            "context_omissions": [
+              {
+                "available_bytes": 25812,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "required_bytes": 170782
+              },
+              {
+                "available_bytes": 24737,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.inference_benchmark_design",
+                "required_bytes": 47014
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "a057b7f86a24f34a8f84abbbb4910d58d4e50ff47060e1c404600017e3febe89",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/metadata.json",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-CANN-2.0-LICENSE.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-stage-evidence.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/SKILL.md"
+                  }
+                ],
+                "content_bytes": 26931,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Only the incremental canonical CannBot fixed-stage evidence boundary is normalized; inherited MindIE-SD guidance remains unchanged.",
+                  "CannBot callback predicates do not establish numerical accuracy or performance; dependent skills, agent-host integration and NPU execution are unverified.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/ascend/ascend-performance-analysis/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/ascend-performance-analysis/SKILL.md",
+                    "role": "core",
+                    "sha256": "a057b7f86a24f34a8f84abbbb4910d58d4e50ff47060e1c404600017e3febe89"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-stage-evidence.md",
+                    "role": "reference",
+                    "sha256": "f56e4125325c9c1b9235a1d64fd732861c7686444492422daff44ba8dd648be4"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-CANN-2.0-LICENSE.md",
+                    "role": "reference",
+                    "sha256": "89b1c114df62cfd183004f38d18d07f186436c3b37203dd69c3e84becf9d2282"
+                  }
+                ],
+                "revision": "sha256:f7ea7eb4d21d63a62a20b1a8b719277d4d2929907ff255af4acb1e30028b01dc",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "a057b7f86a24f34a8f84abbbb4910d58d4e50ff47060e1c404600017e3febe89",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/metadata.json",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-CANN-2.0-LICENSE.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-stage-evidence.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "53cc1867a13f54d1006d50517dea8848869ca50df3f0cd610b2424cc8d18c933",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/dense-performance-evidence/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/dense-performance-evidence/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/dense-performance-evidence/SKILL.md"
+                  }
+                ],
+                "content_bytes": 12793,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/dense-performance-evidence/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.dense_performance_evidence",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/dense-performance-evidence/SKILL.md",
+                    "role": "core",
+                    "sha256": "53cc1867a13f54d1006d50517dea8848869ca50df3f0cd610b2424cc8d18c933"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/inference-optimization-casebook/cases/admission-drain-occupancy.md",
+                    "role": "reference",
+                    "sha256": "e88bbac5d253a8e7e3a3d5e21047f23f2970c11ac23887d4303ab94fca52dd27"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/inference-optimization-casebook/cases/admission-drain-occupancy.json",
+                    "role": "reference",
+                    "sha256": "3319e23bbe2bd40e2d45fb89dc87bdc92d4df2b5ddd8086fcf84813b43885b7c"
+                  }
+                ],
+                "revision": "sha256:82e68013915ca052f96119da0a8d32d40d38d0aab792b04844c2e208d10706dc",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "53cc1867a13f54d1006d50517dea8848869ca50df3f0cd610b2424cc8d18c933",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/dense-performance-evidence/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/dense-performance-evidence/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/dense-performance-evidence/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "3af1dbbe81d91cb05fc2ef15a2ae25d7592a08558e1a5535f49c52f33a488c55",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/skill-index.json",
+                      "src/inferrsi/data/skill_library/workflows/claim-contract/SKILL.md"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/workflows/claim-contract/SKILL.md"
+                  }
+                ],
+                "content_bytes": 1075,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is unspecified.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [],
+                "ref": "package:inferrsi.common.skills.claim_contract",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/workflows/claim-contract/SKILL.md",
+                    "role": "core",
+                    "sha256": "3af1dbbe81d91cb05fc2ef15a2ae25d7592a08558e1a5535f49c52f33a488c55"
+                  }
+                ],
+                "revision": "sha256:0f6955d51d86aa329bf643f37bf750084520c666360a64d6c10bfb678e64b8a2",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "3af1dbbe81d91cb05fc2ef15a2ae25d7592a08558e1a5535f49c52f33a488c55",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/skill-index.json",
+                      "src/inferrsi/data/skill_library/workflows/claim-contract/SKILL.md"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/workflows/claim-contract/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 467584,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_device_health",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_performance_analysis",
+                "ref": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "resource_count": 343,
+                "resource_hashes_sha256": "255c964b950c3e58c7a55001c35a22d48e386da437cea66ec2486a1f0412dd1f",
+                "revision": "sha256:f7ea7eb4d21d63a62a20b1a8b719277d4d2929907ff255af4acb1e30028b01dc"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.dense_performance_evidence",
+                "ref": "package:inferrsi.common.skills.dense_performance_evidence",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ee7e4bd7570b737eb14c98190545cee0fa11065e1b1a2d0cfe663928cfca0551",
+                "revision": "sha256:82e68013915ca052f96119da0a8d32d40d38d0aab792b04844c2e208d10706dc"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.performance_ceiling_analysis",
+                "ref": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c915247877787fcecc949d1d212ae1f18e3ba2a82f02b7c1cf343dedf0a7e885",
+                "revision": "sha256:6f67a5d347f3d05198f92cfce346d30051a544427ec8922b9de64432c3ffe316"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.claim_contract",
+                "ref": "package:inferrsi.common.skills.claim_contract",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c6481b673e503a8707f897b6e23c355888873b8818458e6a4b8e1fbfee9f9f8f",
+                "revision": "sha256:0f6955d51d86aa329bf643f37bf750084520c666360a64d6c10bfb678e64b8a2"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.inference_benchmark_design",
+                "ref": "package:inferrsi.common.skills.inference_benchmark_design",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb743a7fe0662feead981795dc60e2a5c9355ed876339f14ad7a47caab0bb670",
+                "revision": "sha256:6b0cbeaa5d409c04228883884e192f275cad4874c7f6b3039abef9fd443fafb5"
+              }
+            ],
+            "selection": {
+              "cue": "When does missing validator freeze make a performance claim inconclusive?",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 524270,
+            "sha256": "a2df470f45879e6dadaa8c1af7f4f25bf456365e4948e6921276bb8983de4564"
+          },
+          "reciprocal_rank_at_5": 0.25,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "dev14",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": false,
+          "case": "dev15",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": false,
+          "case": "dev15",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "selection": {
+              "cue": "quasarbanjo xenofloral v7",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 452,
+            "sha256": "b4e79e7856cb502b1735da83ea5cc08996bdeb3eff8c380192ea6710457471a1"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": false,
+          "case": "dev15",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 0,
+            "context_omissions": [],
+            "contexts": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "selection": {
+              "cue": "quasarbanjo xenofloral v7",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 636,
+            "sha256": "a1344ff20112717b0a2fd180c8399638fee996e40c1fbc27236cd2508010dcf2"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": false,
+          "case": "dev15",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": false,
+          "case": "dev16",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": false,
+          "case": "dev16",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 430698,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.assess_serving_preemption",
+                "ref": "package:inferrsi.common.skills.assess_serving_preemption",
+                "resource_count": 343,
+                "resource_hashes_sha256": "1514c158f4d1bbe57e52a20bb872edbd6c628c9fe34cf5018e0d9c4abcab1a55",
+                "revision": "sha256:11751aa0d285e7fcf21c1c0f4564c46a09683b123895cc95bc22a3eccaae6221"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_h3_runtime_residency",
+                "ref": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "64ea29a64c41b0dac809d22e10aa650ac7d60cb008ab15d9e702161ce2ad63c2",
+                "revision": "sha256:601645ce381756514c931d83e0019b1901272e08c9fd4e0ae99318ef0d18405b"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "ref": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "resource_count": 343,
+                "resource_hashes_sha256": "8af3478b2f59c5fade35c614a9bfe4e4d53e8d5e9a5961870e8f4e62631cb937",
+                "revision": "sha256:0af6ac4cfd9d26bd90eeba094c3ceeed80e418dbf1872e6b1ffaa23b6b30e397"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.draft_acceptance_issue",
+                "ref": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "resource_count": 343,
+                "resource_hashes_sha256": "251e45e8af28e24055a9df480c2d541a1a0d3be419e40b212a5aae5cd61b4761",
+                "revision": "sha256:9f179f995ef669f59414e778a2064a4eece7fda9571ae0c191274b4f718b6b2d"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.host_memory_management",
+                "ref": "package:inferrsi.common.skills.host_memory_management",
+                "resource_count": 343,
+                "resource_hashes_sha256": "704c39f2041e99d4d66d9e4464c116274638e246db6a4b56e7c447dd711c198f",
+                "revision": "sha256:96c29b3da68c9493ce25421640f23cc2c15258005b6e006b6e9c2af285624fc3"
+              }
+            ],
+            "selection": {
+              "cue": "sourdough levain hydration recipe",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 431437,
+            "sha256": "4d442f904cbce4a4764030832d10d4e1a091b349bf15a12a4d347121054a54ef"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": false,
+          "case": "dev16",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 38037,
+            "context_omissions": [
+              {
+                "available_bytes": 28729,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "required_bytes": 36142
+              },
+              {
+                "available_bytes": 28729,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "required_bytes": 47942
+              },
+              {
+                "available_bytes": 27499,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.host_memory_management",
+                "required_bytes": 49803
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "0a5f48d9352c5bf460775ef47f9136829978811f4a16e84983e299782cae305d",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/assess-serving-preemption/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/assess-serving-preemption/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md"
+                  }
+                ],
+                "content_bytes": 36807,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/assess-serving-preemption/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.assess_serving_preemption",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/assess-serving-preemption/SKILL.md",
+                    "role": "core",
+                    "sha256": "0a5f48d9352c5bf460775ef47f9136829978811f4a16e84983e299782cae305d"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/assess-serving-preemption/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "f29fcecfba0490008a354c11b42403983fe165e56ff907ad61cc6040d2b6407f"
+                  }
+                ],
+                "revision": "sha256:11751aa0d285e7fcf21c1c0f4564c46a09683b123895cc95bc22a3eccaae6221",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "0a5f48d9352c5bf460775ef47f9136829978811f4a16e84983e299782cae305d",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/assess-serving-preemption/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/assess-serving-preemption/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "ce983acfb7d7d3272775e7cf00c92e4ddadd97511b602575a76e5ab7ade01b56",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/skill-index.json",
+                      "src/inferrsi/data/skill_library/workflows/draft-acceptance-issue/SKILL.md"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/workflows/draft-acceptance-issue/SKILL.md"
+                  }
+                ],
+                "content_bytes": 1230,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is unspecified.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [],
+                "ref": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/workflows/draft-acceptance-issue/SKILL.md",
+                    "role": "core",
+                    "sha256": "ce983acfb7d7d3272775e7cf00c92e4ddadd97511b602575a76e5ab7ade01b56"
+                  }
+                ],
+                "revision": "sha256:9f179f995ef669f59414e778a2064a4eece7fda9571ae0c191274b4f718b6b2d",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "ce983acfb7d7d3272775e7cf00c92e4ddadd97511b602575a76e5ab7ade01b56",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/skill-index.json",
+                      "src/inferrsi/data/skill_library/workflows/draft-acceptance-issue/SKILL.md"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/workflows/draft-acceptance-issue/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 460977,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.assess_serving_preemption",
+                "ref": "package:inferrsi.common.skills.assess_serving_preemption",
+                "resource_count": 343,
+                "resource_hashes_sha256": "1514c158f4d1bbe57e52a20bb872edbd6c628c9fe34cf5018e0d9c4abcab1a55",
+                "revision": "sha256:11751aa0d285e7fcf21c1c0f4564c46a09683b123895cc95bc22a3eccaae6221"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_h3_runtime_residency",
+                "ref": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "64ea29a64c41b0dac809d22e10aa650ac7d60cb008ab15d9e702161ce2ad63c2",
+                "revision": "sha256:601645ce381756514c931d83e0019b1901272e08c9fd4e0ae99318ef0d18405b"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "ref": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "resource_count": 343,
+                "resource_hashes_sha256": "8af3478b2f59c5fade35c614a9bfe4e4d53e8d5e9a5961870e8f4e62631cb937",
+                "revision": "sha256:0af6ac4cfd9d26bd90eeba094c3ceeed80e418dbf1872e6b1ffaa23b6b30e397"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.draft_acceptance_issue",
+                "ref": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "resource_count": 343,
+                "resource_hashes_sha256": "251e45e8af28e24055a9df480c2d541a1a0d3be419e40b212a5aae5cd61b4761",
+                "revision": "sha256:9f179f995ef669f59414e778a2064a4eece7fda9571ae0c191274b4f718b6b2d"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.host_memory_management",
+                "ref": "package:inferrsi.common.skills.host_memory_management",
+                "resource_count": 343,
+                "resource_hashes_sha256": "704c39f2041e99d4d66d9e4464c116274638e246db6a4b56e7c447dd711c198f",
+                "revision": "sha256:96c29b3da68c9493ce25421640f23cc2c15258005b6e006b6e9c2af285624fc3"
+              }
+            ],
+            "selection": {
+              "cue": "sourdough levain hydration recipe",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 505859,
+            "sha256": "504fb320c520276a163585366787f608153457c6eea9670efc3d394092945afd"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": false,
+          "case": "dev16",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "withheld01",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "withheld01",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 85410,
+            "no_match": false,
+            "omissions": [],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.training_input_pipeline",
+                "ref": "package:inferrsi.common.skills.training_input_pipeline",
+                "resource_count": 343,
+                "resource_hashes_sha256": "2316be47093b856fff9936ba91f578c2c42c12938ecedc7eae24e16a0fb6385a",
+                "revision": "sha256:d2e9d3993b49d3de2eac8ee492e8467c80bbb26787cee0687bb78a76c111f3d5"
+              }
+            ],
+            "selection": {
+              "cue": "prefetch_queue_depth executor_statistics",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 85882,
+            "sha256": "5e3948a70e07c223f3599b3b2b38f92377c00de65aa00efb732b41e6d4e3b383"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "withheld01",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 9502,
+            "context_omissions": [],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "64bd2c6b795d966045d1d3cca55976d3d49b08b06df7c7e06a70c2824cf55b7e",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/training-input-pipeline/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/training-input-pipeline/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md"
+                  }
+                ],
+                "content_bytes": 9502,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/training-input-pipeline/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.training_input_pipeline",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/training-input-pipeline/SKILL.md",
+                    "role": "core",
+                    "sha256": "64bd2c6b795d966045d1d3cca55976d3d49b08b06df7c7e06a70c2824cf55b7e"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/training-input-pipeline/references/sources.md",
+                    "role": "reference",
+                    "sha256": "60e0f3e88655d8e1d5eb4797f39ca4370deda01614f265b8d2ea4559d8ac020f"
+                  }
+                ],
+                "revision": "sha256:d2e9d3993b49d3de2eac8ee492e8467c80bbb26787cee0687bb78a76c111f3d5",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "64bd2c6b795d966045d1d3cca55976d3d49b08b06df7c7e06a70c2824cf55b7e",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/training-input-pipeline/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/training-input-pipeline/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 89438,
+            "no_match": false,
+            "omissions": [],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.training_input_pipeline",
+                "ref": "package:inferrsi.common.skills.training_input_pipeline",
+                "resource_count": 343,
+                "resource_hashes_sha256": "2316be47093b856fff9936ba91f578c2c42c12938ecedc7eae24e16a0fb6385a",
+                "revision": "sha256:d2e9d3993b49d3de2eac8ee492e8467c80bbb26787cee0687bb78a76c111f3d5"
+              }
+            ],
+            "selection": {
+              "cue": "prefetch_queue_depth executor_statistics",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 102738,
+            "sha256": "c9e3edc3cb701e76b16a4eda2257be592c37798f136788546aec03bb24c86665"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "withheld01",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "withheld02",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "withheld02",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 427591,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_device_health",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.calibrate_serving_simulation",
+                "ref": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "resource_count": 343,
+                "resource_hashes_sha256": "b4b818eb48639d4c8d599222a472bd4ec3ae2871b14e86e0c16b47689c03ecbe",
+                "revision": "sha256:4531196bea6c75c6fe11252c560708bbbcccf7b233bfd854e7fa39acfa70a9c0"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.assess_serving_preemption",
+                "ref": "package:inferrsi.common.skills.assess_serving_preemption",
+                "resource_count": 343,
+                "resource_hashes_sha256": "1514c158f4d1bbe57e52a20bb872edbd6c628c9fe34cf5018e0d9c4abcab1a55",
+                "revision": "sha256:11751aa0d285e7fcf21c1c0f4564c46a09683b123895cc95bc22a3eccaae6221"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.cann_attribution_overlap",
+                "ref": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c0bf9f5ab49059fd46fbee3e58bc0fac2819bb9d727e921d2362e618e92f996c",
+                "revision": "sha256:425a1e6d432ff3aa077b6974b41c66fb0a88be4ae88b8ba7a827071dfd8f5f10"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.storage_io_measurement",
+                "ref": "package:inferrsi.common.skills.storage_io_measurement",
+                "resource_count": 343,
+                "resource_hashes_sha256": "350d9c963ece5f1e1359ec238f4f804912b7bf0f08743b9cd4b6207583d52515",
+                "revision": "sha256:a57650b2bcaaac84bce2725a49eadb8c17533c82a1125b5ed5d840df5fca201c"
+              }
+            ],
+            "selection": {
+              "cue": "Can a serving simulator certify latency SLOs near queue capacity?",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 433025,
+            "sha256": "53a7e3db0c4873887c03aa8b14273d2153d214ca52be87912dab231e661915ac"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "withheld02",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 62288,
+            "context_omissions": [
+              {
+                "available_bytes": 12220,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.assess_serving_preemption",
+                "required_bytes": 36807
+              },
+              {
+                "available_bytes": 3248,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.storage_io_measurement",
+                "required_bytes": 8813
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "ac773c6f1c110a664dfca28a3ae9129b10c07614e60ca3640dea776220e7b352",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/SKILL.md"
+                  }
+                ],
+                "content_bytes": 12299,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/calibrate-serving-simulation/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/calibrate-serving-simulation/SKILL.md",
+                    "role": "core",
+                    "sha256": "ac773c6f1c110a664dfca28a3ae9129b10c07614e60ca3640dea776220e7b352"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/calibrate-serving-simulation/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "9696d3c0ef96b6611787446bd6f96d4d6255ca8d107105de8c1d9af83edb9fa5"
+                  }
+                ],
+                "revision": "sha256:4531196bea6c75c6fe11252c560708bbbcccf7b233bfd854e7fa39acfa70a9c0",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "ac773c6f1c110a664dfca28a3ae9129b10c07614e60ca3640dea776220e7b352",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md"
+                  }
+                ],
+                "content_bytes": 41017,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Normalized provenance covers the existing llama.cpp loading slice at 3e390af18de1131dda27913c00511725b4274c8a and the fastsafetensors distributed-delivery slice at 219e2561a41ede2d5d8eba611c248bb0effbf580; other inherited claims remain unnormalized with original identities and limits retained.",
+                  "The readsets record static mechanisms and source recommendations, not validated startup, memory, placement or backend performance.",
+                  "The fastsafetensors delivery readset does not revalidate inherited I/O/save claims or qualify uneven scatter, accelerator/NPU compatibility, inter-node transport or agent efficacy.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/weight-loading-optimization/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                    "role": "core",
+                    "sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                    "role": "reference",
+                    "sha256": "6c640f28ee3f0253a17b81461c865cb9f947367002cb75bf72550c4572d02e97"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                    "role": "reference",
+                    "sha256": "5a39f14b102b0cee304541e011aa31e44e6b13527d1bee9987c3f3b66c064d3d"
+                  }
+                ],
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "94197ceea1301d4138e9d712ef409f027ea71139fd357a5c8b6c077534063213",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md",
+                      "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "0e6f094dc554abea12e07def50f20b233b47127779a8aefad595b4ab1efd24c2",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md"
+                  }
+                ],
+                "content_bytes": 8972,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/ascend/cann-attribution-overlap/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md",
+                    "role": "core",
+                    "sha256": "0e6f094dc554abea12e07def50f20b233b47127779a8aefad595b4ab1efd24c2"
+                  }
+                ],
+                "revision": "sha256:425a1e6d432ff3aa077b6974b41c66fb0a88be4ae88b8ba7a827071dfd8f5f10",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "0e6f094dc554abea12e07def50f20b233b47127779a8aefad595b4ab1efd24c2",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 451987,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_device_health",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.calibrate_serving_simulation",
+                "ref": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "resource_count": 343,
+                "resource_hashes_sha256": "b4b818eb48639d4c8d599222a472bd4ec3ae2871b14e86e0c16b47689c03ecbe",
+                "revision": "sha256:4531196bea6c75c6fe11252c560708bbbcccf7b233bfd854e7fa39acfa70a9c0"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.weight_loading_optimization",
+                "ref": "package:inferrsi.common.skills.weight_loading_optimization",
+                "resource_count": 343,
+                "resource_hashes_sha256": "88aa83a7d7a6234dcaaf6209ae2a291021b84de18757717e7473a3f71be059f8",
+                "revision": "sha256:cc2ce422d68b98460135eba60c3a84cff07f9bbf95f560b2158f0ae717695881"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.assess_serving_preemption",
+                "ref": "package:inferrsi.common.skills.assess_serving_preemption",
+                "resource_count": 343,
+                "resource_hashes_sha256": "1514c158f4d1bbe57e52a20bb872edbd6c628c9fe34cf5018e0d9c4abcab1a55",
+                "revision": "sha256:11751aa0d285e7fcf21c1c0f4564c46a09683b123895cc95bc22a3eccaae6221"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.cann_attribution_overlap",
+                "ref": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c0bf9f5ab49059fd46fbee3e58bc0fac2819bb9d727e921d2362e618e92f996c",
+                "revision": "sha256:425a1e6d432ff3aa077b6974b41c66fb0a88be4ae88b8ba7a827071dfd8f5f10"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.storage_io_measurement",
+                "ref": "package:inferrsi.common.skills.storage_io_measurement",
+                "resource_count": 343,
+                "resource_hashes_sha256": "350d9c963ece5f1e1359ec238f4f804912b7bf0f08743b9cd4b6207583d52515",
+                "revision": "sha256:a57650b2bcaaac84bce2725a49eadb8c17533c82a1125b5ed5d840df5fca201c"
+              }
+            ],
+            "selection": {
+              "cue": "Can a serving simulator certify latency SLOs near queue capacity?",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 530723,
+            "sha256": "eb926e0011d0a30484030b3c663e0b4bd8b4142f9facda0e3d8493e20d6e9262"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "withheld02",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "withheld03",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "withheld03",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 432101,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.assess_serving_preemption",
+                "ref": "package:inferrsi.common.skills.assess_serving_preemption",
+                "resource_count": 343,
+                "resource_hashes_sha256": "1514c158f4d1bbe57e52a20bb872edbd6c628c9fe34cf5018e0d9c4abcab1a55",
+                "revision": "sha256:11751aa0d285e7fcf21c1c0f4564c46a09683b123895cc95bc22a3eccaae6221"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "ref": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "resource_count": 343,
+                "resource_hashes_sha256": "b4f60b42dd217de69e7b619ed6d5bdb586ba15b86a0fbb045c905d36669c9fdc",
+                "revision": "sha256:7f70dc8cc627637a561a14f78df55f9b6522803d4652e66cc5be9b686cb64dc0"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "ref": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "resource_count": 343,
+                "resource_hashes_sha256": "75f66b30f6b01297bf08e3dd13e8cf2807c0b7f0fca17e448be531a7bddcd6a1",
+                "revision": "sha256:d2067aa28b25d0d7de079c19c303eaa9b30135737eccd69e492e93f7d53af23b"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.kv_tier_management",
+                "ref": "package:inferrsi.common.skills.kv_tier_management",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ee42b55fa8a404149ecd1dccc32a892566ee8867710531bc0af553409758a019",
+                "revision": "sha256:5816c2e526a71152486af2230ace24d27b54baeb373bf5d27e353ccd44f640aa"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.bound_windowed_attention_replay",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb70e2e746f7ba833a696431c6a4deda830188acf5c4eafce6b0008818f8190a",
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334"
+              }
+            ],
+            "selection": {
+              "cue": "FastServe suspended KV preemption progress",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 434835,
+            "sha256": "bdf2948a61a1f630f22db00b79aa3a7dd29374c2789ad302ac1d33b3f14a1cc9"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "withheld03",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 47619,
+            "context_omissions": [
+              {
+                "available_bytes": 28729,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "required_bytes": 31841
+              },
+              {
+                "available_bytes": 28729,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "required_bytes": 125365
+              },
+              {
+                "available_bytes": 17917,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "required_bytes": 29732
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "0a5f48d9352c5bf460775ef47f9136829978811f4a16e84983e299782cae305d",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/assess-serving-preemption/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/assess-serving-preemption/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md"
+                  }
+                ],
+                "content_bytes": 36807,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/assess-serving-preemption/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.assess_serving_preemption",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/assess-serving-preemption/SKILL.md",
+                    "role": "core",
+                    "sha256": "0a5f48d9352c5bf460775ef47f9136829978811f4a16e84983e299782cae305d"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/assess-serving-preemption/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "f29fcecfba0490008a354c11b42403983fe165e56ff907ad61cc6040d2b6407f"
+                  }
+                ],
+                "revision": "sha256:11751aa0d285e7fcf21c1c0f4564c46a09683b123895cc95bc22a3eccaae6221",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "0a5f48d9352c5bf460775ef47f9136829978811f4a16e84983e299782cae305d",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/assess-serving-preemption/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/assess-serving-preemption/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md"
+                  }
+                ],
+                "content_bytes": 10812,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/repositories/lmcache/kv-tier-management/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.kv_tier_management",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md",
+                    "role": "core",
+                    "sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27"
+                  }
+                ],
+                "revision": "sha256:5816c2e526a71152486af2230ace24d27b54baeb373bf5d27e353ccd44f640aa",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 461107,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.assess_serving_preemption",
+                "ref": "package:inferrsi.common.skills.assess_serving_preemption",
+                "resource_count": 343,
+                "resource_hashes_sha256": "1514c158f4d1bbe57e52a20bb872edbd6c628c9fe34cf5018e0d9c4abcab1a55",
+                "revision": "sha256:11751aa0d285e7fcf21c1c0f4564c46a09683b123895cc95bc22a3eccaae6221"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "ref": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "resource_count": 343,
+                "resource_hashes_sha256": "b4f60b42dd217de69e7b619ed6d5bdb586ba15b86a0fbb045c905d36669c9fdc",
+                "revision": "sha256:7f70dc8cc627637a561a14f78df55f9b6522803d4652e66cc5be9b686cb64dc0"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "ref": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "resource_count": 343,
+                "resource_hashes_sha256": "75f66b30f6b01297bf08e3dd13e8cf2807c0b7f0fca17e448be531a7bddcd6a1",
+                "revision": "sha256:d2067aa28b25d0d7de079c19c303eaa9b30135737eccd69e492e93f7d53af23b"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.kv_tier_management",
+                "ref": "package:inferrsi.common.skills.kv_tier_management",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ee42b55fa8a404149ecd1dccc32a892566ee8867710531bc0af553409758a019",
+                "revision": "sha256:5816c2e526a71152486af2230ace24d27b54baeb373bf5d27e353ccd44f640aa"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.bound_windowed_attention_replay",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb70e2e746f7ba833a696431c6a4deda830188acf5c4eafce6b0008818f8190a",
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334"
+              }
+            ],
+            "selection": {
+              "cue": "FastServe suspended KV preemption progress",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 518251,
+            "sha256": "7ff9648555ebc5513af1c8dc2eb6b8d09a72173071b9d60dfcc7f95341197a1e"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "withheld03",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "withheld04",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "withheld04",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 430471,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "ref": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "resource_count": 343,
+                "resource_hashes_sha256": "8dae7399288be873c56dd379c37268e885c186af8cecde28ec1d5678adb9f204",
+                "revision": "sha256:b2a7f738b1bf78369f99c7e754f293f0b0a94a041156be5831c7f71a4c473eb4"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_graph_mode",
+                "ref": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ab57a9a122977730364bef92b56851835a4a447dc1ccff65874750ef3bac00b1",
+                "revision": "sha256:2b5dc0b9fea661b7bcbe1652fdf8231e7aef22d57736317597d5935d00808663"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sd_cpp_residency",
+                "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c83d7a76dffa5a3d97e932a3eac4602ad82e20babb09d9942504fff6b4db53d6",
+                "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.bound_windowed_attention_replay",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb70e2e746f7ba833a696431c6a4deda830188acf5c4eafce6b0008818f8190a",
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.dcp_async_save_performance",
+                "ref": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "resource_count": 343,
+                "resource_hashes_sha256": "b59637d16a7076477f6011fcceeb3cb5b402aa95af17b841ac9b2ede396216e3",
+                "revision": "sha256:36243cf65b6eaa5f370fd4f7633b8fccb119ff75ed38ba652a6061469de0b5b3"
+              }
+            ],
+            "selection": {
+              "cue": "ACLGraph stable replay buffers",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 433376,
+            "sha256": "246d682026ae637d40def566a27fe95433ba9815666337b7aeed236d5b58fe13"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "withheld04",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 63132,
+            "context_omissions": [
+              {
+                "available_bytes": 15040,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "required_bytes": 29732
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "d494fc7e539cbb7388189bd7b3f04bb5c4983fdd7a82484348aad1cee7f741f2",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/metadata.json",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/SKILL.md"
+                  }
+                ],
+                "content_bytes": 19749,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/SKILL.md",
+                    "role": "core",
+                    "sha256": "d494fc7e539cbb7388189bd7b3f04bb5c4983fdd7a82484348aad1cee7f741f2"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "6ffbfb68cc7e04923b4fbca2a73a7555b7c3b45a6f40819bb5f3213ef05cf88d"
+                  }
+                ],
+                "revision": "sha256:b2a7f738b1bf78369f99c7e754f293f0b0a94a041156be5831c7f71a4c473eb4",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "d494fc7e539cbb7388189bd7b3f04bb5c4983fdd7a82484348aad1cee7f741f2",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/metadata.json",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "9203cc5a14a1d65df5fbf275734b93be023e06527459b2f79864f28a338cb44a",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/SKILL.md"
+                  }
+                ],
+                "content_bytes": 5839,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/hardware/ascend/ascend-graph-mode/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/hardware/ascend/ascend-graph-mode/SKILL.md",
+                    "role": "core",
+                    "sha256": "9203cc5a14a1d65df5fbf275734b93be023e06527459b2f79864f28a338cb44a"
+                  }
+                ],
+                "revision": "sha256:2b5dc0b9fea661b7bcbe1652fdf8231e7aef22d57736317597d5935d00808663",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "9203cc5a14a1d65df5fbf275734b93be023e06527459b2f79864f28a338cb44a",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/SKILL.md",
+                      "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/metadata.json",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md"
+                  }
+                ],
+                "content_bytes": 24908,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; native inference, allocator behavior, image quality and performance were not run.",
+                  "Eighteen bounded readset paths plus one separately identity-checked README match the pinned Git commit/tree blob IDs and sizes (19 blobs total); header coverage is limited to selected owner/state declarations, and this does not authenticate every archive member or submodule contents.",
+                  "GGML and other submodules are not expanded; matching private GGML headers are required even for a system library.",
+                  "CPU bypasses the inspected accelerator capacity gate and segmentation eligibility, so CPU checks cannot qualify accelerator memory budgets or prefetch.",
+                  "Video/audio pipelines, detailed model kernels, approximate denoising caches and full multi-device execution remain outside this readset.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json",
+                    "reason": "not_context_resource"
+                  },
+                  {
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md",
+                    "reason": "not_relevant"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md",
+                    "role": "core",
+                    "sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "b67affbd7d237d82b7f51be49f1537e9820fe56ec70a71ee76809f9e599ca693"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE",
+                    "role": "reference",
+                    "sha256": "b53fa08f515cb5a6fff7b9fd8fcd0961a4b80df29d74372d25e1f9171aa042ee"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md",
+                    "role": "reference",
+                    "sha256": "7b537cb972347dd68e303960f142892346d8b19b884e4d08016ea36a23bdaa7a"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json",
+                    "role": "reference",
+                    "sha256": "af4830370b6eddc2d176a9ae14da4ab56a90b0fe048a01a89ee051df4365d8e9"
+                  }
+                ],
+                "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md"
+                  }
+                ],
+                "content_bytes": 12636,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md",
+                    "role": "core",
+                    "sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md",
+                    "role": "reference",
+                    "sha256": "22afe4f4d77f441096cc657d55d763473bfcbea8fb531c8e493fe49ab84955f7"
+                  }
+                ],
+                "revision": "sha256:36243cf65b6eaa5f370fd4f7633b8fccb119ff75ed38ba652a6061469de0b5b3",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json",
+                      "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 457833,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "ref": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "resource_count": 343,
+                "resource_hashes_sha256": "8dae7399288be873c56dd379c37268e885c186af8cecde28ec1d5678adb9f204",
+                "revision": "sha256:b2a7f738b1bf78369f99c7e754f293f0b0a94a041156be5831c7f71a4c473eb4"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_graph_mode",
+                "ref": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ab57a9a122977730364bef92b56851835a4a447dc1ccff65874750ef3bac00b1",
+                "revision": "sha256:2b5dc0b9fea661b7bcbe1652fdf8231e7aef22d57736317597d5935d00808663"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sd_cpp_residency",
+                "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c83d7a76dffa5a3d97e932a3eac4602ad82e20babb09d9942504fff6b4db53d6",
+                "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.bound_windowed_attention_replay",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb70e2e746f7ba833a696431c6a4deda830188acf5c4eafce6b0008818f8190a",
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.dcp_async_save_performance",
+                "ref": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "resource_count": 343,
+                "resource_hashes_sha256": "b59637d16a7076477f6011fcceeb3cb5b402aa95af17b841ac9b2ede396216e3",
+                "revision": "sha256:36243cf65b6eaa5f370fd4f7633b8fccb119ff75ed38ba652a6061469de0b5b3"
+              }
+            ],
+            "selection": {
+              "cue": "ACLGraph stable replay buffers",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 540079,
+            "sha256": "5f59eb7e5e46575cdd98e9e9fe4418b371e520535d5527315b040710d075236e"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "withheld04",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 172955,
+            "no_match": false,
+            "omissions": [],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sd_cpp_residency",
+                "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c83d7a76dffa5a3d97e932a3eac4602ad82e20babb09d9942504fff6b4db53d6",
+                "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.dcp_async_save_performance",
+                "ref": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "resource_count": 343,
+                "resource_hashes_sha256": "b59637d16a7076477f6011fcceeb3cb5b402aa95af17b841ac9b2ede396216e3",
+                "revision": "sha256:36243cf65b6eaa5f370fd4f7633b8fccb119ff75ed38ba652a6061469de0b5b3"
+              }
+            ],
+            "serialized_bytes": 173274,
+            "sha256": "ad3941abe38ece8301a3ead812ee767ea65c7288e24f57afe0a471f8eac20e20"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "withheld05",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "withheld05",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 433368,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_device_health",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "ref": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "resource_count": 343,
+                "resource_hashes_sha256": "568becaa26f22d5fcb75a81fb68bb6b0f46a9345dbddd9518a5057a62e6c9e28",
+                "revision": "sha256:a2c3f95251ace650d22c0bc2804780c91236da923993a46cbb25a58962ae00d1"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_distributed_parallel",
+                "ref": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ecb6d382307b933beb053b3c380a5bb6ba87c13792c428ad4131c78f5a9bcd04",
+                "revision": "sha256:66dab07081fad44b9cfe69b9c0f7a8f1dbb2c3fa19157ac99e4eee9a6306f605"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.budget_speculative_verification",
+                "ref": "package:inferrsi.common.skills.budget_speculative_verification",
+                "resource_count": 343,
+                "resource_hashes_sha256": "6ef4bcbc8e2ee8a5fc04c5c970ce5ec223bc85904818d7e006fb6d67548d734b",
+                "revision": "sha256:2686550fa366a367d026447ac1f2bec6b9ad3a81d559b70dccb8098524a27e2d"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "ref": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "resource_count": 343,
+                "resource_hashes_sha256": "8af3478b2f59c5fade35c614a9bfe4e4d53e8d5e9a5961870e8f4e62631cb937",
+                "revision": "sha256:0af6ac4cfd9d26bd90eeba094c3ceeed80e418dbf1872e6b1ffaa23b6b30e397"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "ref": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "resource_count": 343,
+                "resource_hashes_sha256": "75f66b30f6b01297bf08e3dd13e8cf2807c0b7f0fca17e448be531a7bddcd6a1",
+                "revision": "sha256:d2067aa28b25d0d7de079c19c303eaa9b30135737eccd69e492e93f7d53af23b"
+              }
+            ],
+            "selection": {
+              "cue": "How is moving a request between replicas different from changing its sequence parallel group?",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 438800,
+            "sha256": "90023483fcd615d2ac31d9dea0f83b559c443293874331264bb7ed429e382a77"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "withheld05",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 51463,
+            "context_omissions": [
+              {
+                "available_bytes": 14073,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "required_bytes": 24091
+              },
+              {
+                "available_bytes": 14073,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.budget_speculative_verification",
+                "required_bytes": 30031
+              },
+              {
+                "available_bytes": 14073,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "required_bytes": 47942
+              },
+              {
+                "available_bytes": 14073,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "required_bytes": 125365
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "82d0099d3c33706f46c657d4fa7f76d3d486e019b36760d1de001921328a60cc",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/SKILL.md"
+                  }
+                ],
+                "content_bytes": 51463,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/evaluate-elastic-serving-placement/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/evaluate-elastic-serving-placement/SKILL.md",
+                    "role": "core",
+                    "sha256": "82d0099d3c33706f46c657d4fa7f76d3d486e019b36760d1de001921328a60cc"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/evaluate-elastic-serving-placement/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "ea416831fd7d0874e06d0a76fb0c7196df53072b717673a0cdf635db6bfd917b"
+                  }
+                ],
+                "revision": "sha256:a2c3f95251ace650d22c0bc2804780c91236da923993a46cbb25a58962ae00d1",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "82d0099d3c33706f46c657d4fa7f76d3d486e019b36760d1de001921328a60cc",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 465980,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_performance_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.cann_attribution_overlap",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_graph_mode",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_device_health",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.npu_skill_factory",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_tuning",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_model_adaptation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascendc_tiling_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_aclgraph_capture_replay",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.ascend_operator_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_device_memory_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "ref": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "resource_count": 343,
+                "resource_hashes_sha256": "568becaa26f22d5fcb75a81fb68bb6b0f46a9345dbddd9518a5057a62e6c9e28",
+                "revision": "sha256:a2c3f95251ace650d22c0bc2804780c91236da923993a46cbb25a58962ae00d1"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.ascend.skills.ascend_distributed_parallel",
+                "ref": "package:inferrsi.ascend.skills.ascend_distributed_parallel",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ecb6d382307b933beb053b3c380a5bb6ba87c13792c428ad4131c78f5a9bcd04",
+                "revision": "sha256:66dab07081fad44b9cfe69b9c0f7a8f1dbb2c3fa19157ac99e4eee9a6306f605"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.budget_speculative_verification",
+                "ref": "package:inferrsi.common.skills.budget_speculative_verification",
+                "resource_count": 343,
+                "resource_hashes_sha256": "6ef4bcbc8e2ee8a5fc04c5c970ce5ec223bc85904818d7e006fb6d67548d734b",
+                "revision": "sha256:2686550fa366a367d026447ac1f2bec6b9ad3a81d559b70dccb8098524a27e2d"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "ref": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "resource_count": 343,
+                "resource_hashes_sha256": "8af3478b2f59c5fade35c614a9bfe4e4d53e8d5e9a5961870e8f4e62631cb937",
+                "revision": "sha256:0af6ac4cfd9d26bd90eeba094c3ceeed80e418dbf1872e6b1ffaa23b6b30e397"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "ref": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "resource_count": 343,
+                "resource_hashes_sha256": "75f66b30f6b01297bf08e3dd13e8cf2807c0b7f0fca17e448be531a7bddcd6a1",
+                "revision": "sha256:d2067aa28b25d0d7de079c19c303eaa9b30135737eccd69e492e93f7d53af23b"
+              }
+            ],
+            "selection": {
+              "cue": "How is moving a request between replicas different from changing its sequence parallel group?",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 527088,
+            "sha256": "23051b189f8457ebd6a0782db69ab66791287b5bd1a232dcf63503fcef96de50"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "withheld05",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": true,
+          "case": "withheld06",
+          "mode": "search",
+          "output": {
+            "ordered_items": [
+              {
+                "id": "inferrsi.common.skills.performance_ceiling_analysis"
+              }
+            ],
+            "serialized_bytes": 12360,
+            "sha256": "878199c975f54b0a6256318c79d619e0d51a18a490ced61d43204c8e4ad2ec56"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "withheld06",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 437219,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.performance_ceiling_analysis",
+                "ref": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c915247877787fcecc949d1d212ae1f18e3ba2a82f02b7c1cf343dedf0a7e885",
+                "revision": "sha256:6f67a5d347f3d05198f92cfce346d30051a544427ec8922b9de64432c3ffe316"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.bound_windowed_attention_replay",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb70e2e746f7ba833a696431c6a4deda830188acf5c4eafce6b0008818f8190a",
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.kernel_writing_guide",
+                "ref": "package:inferrsi.common.skills.kernel_writing_guide",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ff2d799c7b84ff56e3dc76cdb1648283a621bd5c104c8317acc839969fc41208",
+                "revision": "sha256:adf2c9d4ec9e0157388dcb11ff0367773d2967717a8dd944ef92647b9307b9bc"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.npu_skill_factory",
+                "ref": "package:inferrsi.common.skills.npu_skill_factory",
+                "resource_count": 343,
+                "resource_hashes_sha256": "7cde43221171eca8e11527545216e348d8c464438bec7458aab055843bb440cf",
+                "revision": "sha256:c789873dd4195bee85724a481228f51852a09c870d630ad1e828901037f875b7"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.host_memory_management",
+                "ref": "package:inferrsi.common.skills.host_memory_management",
+                "resource_count": 343,
+                "resource_hashes_sha256": "704c39f2041e99d4d66d9e4464c116274638e246db6a4b56e7c447dd711c198f",
+                "revision": "sha256:96c29b3da68c9493ce25421640f23cc2c15258005b6e006b6e9c2af285624fc3"
+              }
+            ],
+            "selection": {
+              "cue": "roofline analytical empirical end-to-end ceiling",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 441466,
+            "sha256": "6c72e2c8df479a21536178c7f1ef1c210eb067b561fb5675ea8dba7c2fa718c4"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "withheld06",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 51702,
+            "context_omissions": [
+              {
+                "available_bytes": 65536,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "required_bytes": 170782
+              },
+              {
+                "available_bytes": 13834,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.npu_skill_factory",
+                "required_bytes": 65054
+              },
+              {
+                "available_bytes": 13834,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.host_memory_management",
+                "required_bytes": 49803
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "4703beb6e5194ef2ff2c0b5ec69f64e35417a4bc0dfb50a10c80a920b77ef3b5",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/validation.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md"
+                  }
+                ],
+                "content_bytes": 29732,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no equivalence run was performed.",
+                  "Applies to architectures with a demonstrated boundary dependency, not to arbitrary sliding-window stacks.",
+                  "Window sizes, boundary-layer index and ring lengths are model-config specific and are not transferable.",
+                  "Query-row compaction does not imply dropping historical keys; the two decisions are separate.",
+                  "Prompt logprobs are unrecoverable for the clipped region; the reference forces them off with a warning.",
+                  "The truncation path is off by default in the reference source.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md",
+                    "role": "core",
+                    "sha256": "4703beb6e5194ef2ff2c0b5ec69f64e35417a4bc0dfb50a10c80a920b77ef3b5"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "fd917bd5ac80bee9c9f4525f8e0300f6d67dc144e2a16632af42a67f3d53d098"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/validation.md",
+                    "role": "reference",
+                    "sha256": "60e71e70c5519bcdf6c171eec2be2f342e8801bedd746d6005f11adc646dfaff"
+                  }
+                ],
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "4703beb6e5194ef2ff2c0b5ec69f64e35417a4bc0dfb50a10c80a920b77ef3b5",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/validation.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "01fb1712d7eb988c5e412f02340d2e48b4f8564fcf4ab232dc5897869d0d50a0",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/kernel-writing-guide/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/kernel-writing-guide/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/kernel-writing-guide/references/akg-affine-tail-guards.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/kernel-writing-guide/SKILL.md"
+                  }
+                ],
+                "content_bytes": 21970,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no local efficacy validation.",
+                  "Original typed source inventory is not normalized; retain source identities and limits in the procedure/readsets.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/methods/kernel-writing-guide/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.kernel_writing_guide",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/kernel-writing-guide/SKILL.md",
+                    "role": "core",
+                    "sha256": "01fb1712d7eb988c5e412f02340d2e48b4f8564fcf4ab232dc5897869d0d50a0"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/methods/kernel-writing-guide/references/akg-affine-tail-guards.md",
+                    "role": "reference",
+                    "sha256": "113deb2db10b171db3425bb3f1792bee763706b794c8cc2e222b4cc0d0114118"
+                  }
+                ],
+                "revision": "sha256:adf2c9d4ec9e0157388dcb11ff0367773d2967717a8dd944ef92647b9307b9bc",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "01fb1712d7eb988c5e412f02340d2e48b4f8564fcf4ab232dc5897869d0d50a0",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/methods/kernel-writing-guide/SKILL.md",
+                      "src/inferrsi/data/skill_library/methods/kernel-writing-guide/metadata.json",
+                      "src/inferrsi/data/skill_library/methods/kernel-writing-guide/references/akg-affine-tail-guards.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/methods/kernel-writing-guide/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 478766,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.repository_evolution_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dcp_async_save_performance",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.storage_io_measurement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.budget_speculative_verification",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.calibrate_serving_simulation",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_serving_preemption",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.generation_cache_lifecycle",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.audit_afd_p2p_exchanges",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_vllm_kv_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cgroup_resource_controls",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.dense_performance_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.evaluate_elastic_serving_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_optimization_casebook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.reason_hopper_kernels",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.choose_parallel_training_plan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.comfy_ui",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.competitor_scan",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.os_process_placement",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.rsi_evidence",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.skill_evolution_playbook",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.audit_h3_runtime_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.lazy_evaluation_stream_lifetimes",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.diagnose_attention_bottlenecks",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.host_overhead_optimization",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.inference_benchmark_design",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.assess_split_backward_schedule",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.checkpoint_recovery_consistency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.claim_contract",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.plan_activation_residency",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.cuda.skills.host_overhead_profiling",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.cite_check_downstream_pr",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.draft_acceptance_issue",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.performance_ceiling_analysis",
+                "ref": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c915247877787fcecc949d1d212ae1f18e3ba2a82f02b7c1cf343dedf0a7e885",
+                "revision": "sha256:6f67a5d347f3d05198f92cfce346d30051a544427ec8922b9de64432c3ffe316"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.bound_windowed_attention_replay",
+                "ref": "package:inferrsi.common.skills.bound_windowed_attention_replay",
+                "resource_count": 343,
+                "resource_hashes_sha256": "fb70e2e746f7ba833a696431c6a4deda830188acf5c4eafce6b0008818f8190a",
+                "revision": "sha256:0a4b1c3323f63771684eaf577ede6f4359ccf196ce596f9f7c2f86b596098334"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.kernel_writing_guide",
+                "ref": "package:inferrsi.common.skills.kernel_writing_guide",
+                "resource_count": 343,
+                "resource_hashes_sha256": "ff2d799c7b84ff56e3dc76cdb1648283a621bd5c104c8317acc839969fc41208",
+                "revision": "sha256:adf2c9d4ec9e0157388dcb11ff0367773d2967717a8dd944ef92647b9307b9bc"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.npu_skill_factory",
+                "ref": "package:inferrsi.common.skills.npu_skill_factory",
+                "resource_count": 343,
+                "resource_hashes_sha256": "7cde43221171eca8e11527545216e348d8c464438bec7458aab055843bb440cf",
+                "revision": "sha256:c789873dd4195bee85724a481228f51852a09c870d630ad1e828901037f875b7"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.host_memory_management",
+                "ref": "package:inferrsi.common.skills.host_memory_management",
+                "resource_count": 343,
+                "resource_hashes_sha256": "704c39f2041e99d4d66d9e4464c116274638e246db6a4b56e7c447dd711c198f",
+                "revision": "sha256:96c29b3da68c9493ce25421640f23cc2c15258005b6e006b6e9c2af285624fc3"
+              }
+            ],
+            "selection": {
+              "cue": "roofline analytical empirical end-to-end ceiling",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 543170,
+            "sha256": "0c9097a61b83e59a924ee2be136aae957cd16a393c3533a8b0d76758d1e12097"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": true,
+          "case": "withheld06",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 90114,
+            "no_match": false,
+            "omissions": [],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.performance_ceiling_analysis",
+                "ref": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c915247877787fcecc949d1d212ae1f18e3ba2a82f02b7c1cf343dedf0a7e885",
+                "revision": "sha256:6f67a5d347f3d05198f92cfce346d30051a544427ec8922b9de64432c3ffe316"
+              }
+            ],
+            "serialized_bytes": 90431,
+            "sha256": "70cb18c9f4f5c7f3ab58acab24d362b94f743c2488d270976d84ade1458688ed"
+          },
+          "reciprocal_rank_at_5": 1.0,
+          "success_at_5": 1
+        },
+        {
+          "answerable": false,
+          "case": "withheld07",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": false,
+          "case": "withheld07",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "selection": {
+              "cue": "zephyrcodex latticeotter q83",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 455,
+            "sha256": "a20a7f3222447ce17be8178a15f84a96cd11f61b85861f826c9a22c51d49d87b"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": false,
+          "case": "withheld07",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 0,
+            "context_omissions": [],
+            "contexts": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "selection": {
+              "cue": "zephyrcodex latticeotter q83",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 639,
+            "sha256": "121d4fffb7abe4d79ecb33694570a620007712feeb0062b3602eebb7f1d902a7"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": false,
+          "case": "withheld07",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": false,
+          "case": "withheld08",
+          "mode": "search",
+          "output": {
+            "ordered_items": [],
+            "serialized_bytes": 2,
+            "sha256": "4f53cda18c2baa0c0354bb5f9a3ecbe5ed12ab4d8e11ba873c2f11161202b945"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": false,
+          "case": "withheld08",
+          "mode": "relevance",
+          "output": {
+            "collisions": [],
+            "material_bytes": 434773,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_llama_runtime_state",
+                "ref": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "resource_count": 343,
+                "resource_hashes_sha256": "03e86bd3af77f17d6c7ad2933534524ab7fe5993396a7fe9748a6014496a34b5",
+                "revision": "sha256:a3c3d3244cda589e8e7809e56f43b7bfff1852def8dd26fb03a9d29e39bca3dd"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sd_cpp_residency",
+                "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c83d7a76dffa5a3d97e932a3eac4602ad82e20babb09d9942504fff6b4db53d6",
+                "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "ref": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "resource_count": 343,
+                "resource_hashes_sha256": "8af3478b2f59c5fade35c614a9bfe4e4d53e8d5e9a5961870e8f4e62631cb937",
+                "revision": "sha256:0af6ac4cfd9d26bd90eeba094c3ceeed80e418dbf1872e6b1ffaa23b6b30e397"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "ref": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "resource_count": 343,
+                "resource_hashes_sha256": "75f66b30f6b01297bf08e3dd13e8cf2807c0b7f0fca17e448be531a7bddcd6a1",
+                "revision": "sha256:d2067aa28b25d0d7de079c19c303eaa9b30135737eccd69e492e93f7d53af23b"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.diffusers_invocation_state",
+                "ref": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "resource_count": 343,
+                "resource_hashes_sha256": "64b10a60e44276bc4a20b3441f65afd929341e8558fde32491ba27a20fde4020",
+                "revision": "sha256:a8529f2c6bf5f84ce48ba0d6426f1bf1bc6fce172cd0a599b6a149878e3b4be4"
+              }
+            ],
+            "selection": {
+              "cue": "baroque harpsichord tuning temperament",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 435955,
+            "sha256": "0b35e7fc7320ef46c35a3eecfff7896a76cb5bc9c78d60b6b65435a9dadb3423"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": false,
+          "case": "withheld08",
+          "mode": "context_material",
+          "output": {
+            "collisions": [],
+            "context_budget": {
+              "accounting": "utf8-resource-content",
+              "excludes": "provenance, omission and control metadata",
+              "max_bytes": 65536
+            },
+            "context_bytes": 49333,
+            "context_omissions": [
+              {
+                "available_bytes": 27809,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "required_bytes": 47942
+              },
+              {
+                "available_bytes": 27809,
+                "reason": "context_limit",
+                "ref": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "required_bytes": 125365
+              }
+            ],
+            "contexts": [
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "95e81b6bc096a954fa73139ea5bc001d1efa6c99747161da62f012b494e5c013",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/SKILL.md"
+                  }
+                ],
+                "content_bytes": 24572,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; native runtime and efficacy validation not run.",
+                  "No cross-backend preemption, model-quality or performance claim.",
+                  "KV, recurrent and hybrid state contracts must be checked independently.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/metadata.json",
+                    "reason": "not_context_resource"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/SKILL.md",
+                    "role": "core",
+                    "sha256": "95e81b6bc096a954fa73139ea5bc001d1efa6c99747161da62f012b494e5c013"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "3d12a401e9decb55a2f4f3b2bfa0d360ef5110f63a0469dfe285654e78054190"
+                  }
+                ],
+                "revision": "sha256:a3c3d3244cda589e8e7809e56f43b7bfff1852def8dd26fb03a9d29e39bca3dd",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "95e81b6bc096a954fa73139ea5bc001d1efa6c99747161da62f012b494e5c013",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md"
+                  }
+                ],
+                "content_bytes": 13155,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; native inference, allocator behavior, image quality and performance were not run.",
+                  "Eighteen bounded readset paths plus one separately identity-checked README match the pinned Git commit/tree blob IDs and sizes (19 blobs total); header coverage is limited to selected owner/state declarations, and this does not authenticate every archive member or submodule contents.",
+                  "GGML and other submodules are not expanded; matching private GGML headers are required even for a system library.",
+                  "CPU bypasses the inspected accelerator capacity gate and segmentation eligibility, so CPU checks cannot qualify accelerator memory budgets or prefetch.",
+                  "Video/audio pipelines, detailed model kernels, approximate denoising caches and full multi-device execution remain outside this readset.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json",
+                    "reason": "not_context_resource"
+                  },
+                  {
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md",
+                    "reason": "not_relevant"
+                  },
+                  {
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json",
+                    "reason": "not_relevant"
+                  },
+                  {
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md",
+                    "reason": "not_relevant"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md",
+                    "role": "core",
+                    "sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md",
+                    "role": "reference",
+                    "sha256": "b67affbd7d237d82b7f51be49f1537e9820fe56ec70a71ee76809f9e599ca693"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE",
+                    "role": "reference",
+                    "sha256": "b53fa08f515cb5a6fff7b9fd8fcd0961a4b80df29d74372d25e1f9171aa042ee"
+                  }
+                ],
+                "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "3e93b2b4764f61ea00ec7d1a96aa5b9e2c5f2e746cb6e65287cdce659c056cf6",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md",
+                      "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md"
+                  }
+                ]
+              },
+              {
+                "attribution": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "7581b56336d07213a87d01477793ae1a8acf0b7cf876fce36c9a761ee32c1ff4",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/SKILL.md",
+                      "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/metadata.json",
+                      "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/ownership.md",
+                      "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/source-map.md",
+                      "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/validation.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/SKILL.md"
+                  }
+                ],
+                "content_bytes": 11606,
+                "limitations": [
+                  "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+                  "The imported status describes this index, not the original skill; its declared status is candidate.",
+                  "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+                  "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+                  "Source-derived candidate; no installed API, native scheduler, model, concurrency or service validation.",
+                  "Supplied manifest verifies a selected pipeline_utils Git blob, not the whole source tree.",
+                  "Other pipelines and schedulers require separate contracts; common indexing is not backend qualification.",
+                  "Imported guidance; reading does not validate or execute it.",
+                  "Historical revisions require the matching installed snapshot or a retained store.",
+                  "Selected resources are complete files; this view is not a standalone Bundle.",
+                  "Follow omitted or other linked resources through the exact owner revision."
+                ],
+                "omissions": [
+                  {
+                    "path": "data/skill_library/frameworks/diffusers/diffusers-invocation-state/metadata.json",
+                    "reason": "not_context_resource"
+                  },
+                  {
+                    "path": "data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/ownership.md",
+                    "reason": "not_relevant"
+                  }
+                ],
+                "ref": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "resources": [
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/diffusers/diffusers-invocation-state/SKILL.md",
+                    "role": "core",
+                    "sha256": "7581b56336d07213a87d01477793ae1a8acf0b7cf876fce36c9a761ee32c1ff4"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/source-map.md",
+                    "role": "reference",
+                    "sha256": "10562a75683bd9bd07ec2e37a1cbd0f93f2534d6f85b55a566adb2d97485215f"
+                  },
+                  {
+                    "locator": "whole-resource",
+                    "path": "data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/validation.md",
+                    "role": "reference",
+                    "sha256": "5a0b854b87d091d7d505b0bdd5c03330052507a851ab6edf35b93a6675f2d4f7"
+                  }
+                ],
+                "revision": "sha256:a8529f2c6bf5f84ce48ba0d6426f1bf1bc6fce172cd0a599b6a149878e3b4be4",
+                "sources": [
+                  {
+                    "commit": "f321699b42a490276ffa4ea52350591683b575cc",
+                    "content_sha256": "7581b56336d07213a87d01477793ae1a8acf0b7cf876fce36c9a761ee32c1ff4",
+                    "paths": [
+                      "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/SKILL.md",
+                      "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/metadata.json",
+                      "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/ownership.md",
+                      "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/source-map.md",
+                      "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/validation.md",
+                      "src/inferrsi/data/skill_library/skill-index.json"
+                    ],
+                    "revision_precision": "full",
+                    "type": "git",
+                    "url": "https://github.com/InferMatrixLab/InferRSI/blob/f321699b42a490276ffa4ea52350591683b575cc/src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/SKILL.md"
+                  }
+                ]
+              }
+            ],
+            "material_bytes": 469565,
+            "no_match": false,
+            "omissions": [
+              {
+                "id": "package:inferrsi.common.skills.host_memory_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kernel_writing_guide",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.kv_tier_management",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.optimization_search_policy",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.performance_ceiling_analysis",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_input_pipeline",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.training_resource_budget",
+                "reason": "output_limit"
+              },
+              {
+                "id": "package:inferrsi.common.skills.weight_loading_optimization",
+                "reason": "output_limit"
+              }
+            ],
+            "ordered_items": [
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_llama_runtime_state",
+                "ref": "package:inferrsi.common.skills.audit_llama_runtime_state",
+                "resource_count": 343,
+                "resource_hashes_sha256": "03e86bd3af77f17d6c7ad2933534524ab7fe5993396a7fe9748a6014496a34b5",
+                "revision": "sha256:a3c3d3244cda589e8e7809e56f43b7bfff1852def8dd26fb03a9d29e39bca3dd"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sd_cpp_residency",
+                "ref": "package:inferrsi.common.skills.audit_sd_cpp_residency",
+                "resource_count": 343,
+                "resource_hashes_sha256": "c83d7a76dffa5a3d97e932a3eac4602ad82e20babb09d9942504fff6b4db53d6",
+                "revision": "sha256:9a0d9397fbac8a5f480ef012d13c9edd180cbc66cf2f0891fc14b83843313cde"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "ref": "package:inferrsi.common.skills.audit_sglang_diffusion_lifetimes",
+                "resource_count": 343,
+                "resource_hashes_sha256": "8af3478b2f59c5fade35c614a9bfe4e4d53e8d5e9a5961870e8f4e62631cb937",
+                "revision": "sha256:0af6ac4cfd9d26bd90eeba094c3ceeed80e418dbf1872e6b1ffaa23b6b30e397"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "ref": "package:inferrsi.common.skills.diagnose_vllm_admission_pressure",
+                "resource_count": 343,
+                "resource_hashes_sha256": "75f66b30f6b01297bf08e3dd13e8cf2807c0b7f0fca17e448be531a7bddcd6a1",
+                "revision": "sha256:d2067aa28b25d0d7de079c19c303eaa9b30135737eccd69e492e93f7d53af23b"
+              },
+              {
+                "dependencies_sha256": "809334a5ef2698c430e2ba00b4f8809d924c21645eb93bd561639d042baa537f",
+                "id": "inferrsi.common.skills.diffusers_invocation_state",
+                "ref": "package:inferrsi.common.skills.diffusers_invocation_state",
+                "resource_count": 343,
+                "resource_hashes_sha256": "64b10a60e44276bc4a20b3441f65afd929341e8558fde32491ba27a20fde4020",
+                "revision": "sha256:a8529f2c6bf5f84ce48ba0d6426f1bf1bc6fce172cd0a599b6a149878e3b4be4"
+              }
+            ],
+            "selection": {
+              "cue": "baroque harpsichord tuning temperament",
+              "method": "weighted-lexical-v1",
+              "ranking_applied": true,
+              "scope": "generated-packaged-skills"
+            },
+            "serialized_bytes": 533888,
+            "sha256": "bc2e924a1158532fc45063706c569706e559f80e841ded5494168ad894ff1e47"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        },
+        {
+          "answerable": false,
+          "case": "withheld08",
+          "mode": "default_diagnostic",
+          "output": {
+            "collisions": [],
+            "material_bytes": 0,
+            "no_match": true,
+            "omissions": [],
+            "ordered_items": [],
+            "serialized_bytes": 312,
+            "sha256": "2ac3a9ad1e6fdd68f013bdb3cf9b6836e6d00219872d5a92696fdfe6a3b25859"
+          },
+          "reciprocal_rank_at_5": 0,
+          "success_at_5": 0
+        }
+      ]
+    }
+  },
+  "metrics": {
+    "A": {
+      "development/search": {
+        "answerable_count": 14,
+        "success_at_5": 0.21428571428571427,
+        "mrr_at_5": 0.21428571428571427,
+        "negative_count": 2,
+        "negative_empty_count": 2
+      },
+      "development/relevance": {
+        "answerable_count": 14,
+        "success_at_5": 1,
+        "mrr_at_5": 0.9464285714285714,
+        "negative_count": 2,
+        "negative_empty_count": 1
+      },
+      "development/context_material": {
+        "answerable_count": 14,
+        "success_at_5": 1,
+        "mrr_at_5": 0.9464285714285714,
+        "negative_count": 2,
+        "negative_empty_count": 1
+      },
+      "development/default_diagnostic": {
+        "answerable_count": 14,
+        "success_at_5": 0.21428571428571427,
+        "mrr_at_5": 0.21428571428571427,
+        "negative_count": 2,
+        "negative_empty_count": 2
+      },
+      "withheld/search": {
+        "answerable_count": 6,
+        "success_at_5": 0.16666666666666666,
+        "mrr_at_5": 0.16666666666666666,
+        "negative_count": 2,
+        "negative_empty_count": 2
+      },
+      "withheld/relevance": {
+        "answerable_count": 6,
+        "success_at_5": 1,
+        "mrr_at_5": 1.0,
+        "negative_count": 2,
+        "negative_empty_count": 1
+      },
+      "withheld/context_material": {
+        "answerable_count": 6,
+        "success_at_5": 1,
+        "mrr_at_5": 1.0,
+        "negative_count": 2,
+        "negative_empty_count": 1
+      },
+      "withheld/default_diagnostic": {
+        "answerable_count": 6,
+        "success_at_5": 0.16666666666666666,
+        "mrr_at_5": 0.16666666666666666,
+        "negative_count": 2,
+        "negative_empty_count": 2
+      },
+      "all/search": {
+        "answerable_count": 20,
+        "success_at_5": 0.2,
+        "mrr_at_5": 0.2,
+        "negative_count": 4,
+        "negative_empty_count": 4
+      },
+      "all/relevance": {
+        "answerable_count": 20,
+        "success_at_5": 1,
+        "mrr_at_5": 0.9625,
+        "negative_count": 4,
+        "negative_empty_count": 2
+      },
+      "all/context_material": {
+        "answerable_count": 20,
+        "success_at_5": 1,
+        "mrr_at_5": 0.9625,
+        "negative_count": 4,
+        "negative_empty_count": 2
+      },
+      "all/default_diagnostic": {
+        "answerable_count": 20,
+        "success_at_5": 0.2,
+        "mrr_at_5": 0.2,
+        "negative_count": 4,
+        "negative_empty_count": 4
+      }
+    },
+    "B": {
+      "development/search": {
+        "answerable_count": 14,
+        "success_at_5": 0.21428571428571427,
+        "mrr_at_5": 0.21428571428571427,
+        "negative_count": 2,
+        "negative_empty_count": 2
+      },
+      "development/relevance": {
+        "answerable_count": 14,
+        "success_at_5": 1,
+        "mrr_at_5": 0.9464285714285714,
+        "negative_count": 2,
+        "negative_empty_count": 1
+      },
+      "development/context_material": {
+        "answerable_count": 14,
+        "success_at_5": 1,
+        "mrr_at_5": 0.9464285714285714,
+        "negative_count": 2,
+        "negative_empty_count": 1
+      },
+      "development/default_diagnostic": {
+        "answerable_count": 14,
+        "success_at_5": 0.21428571428571427,
+        "mrr_at_5": 0.21428571428571427,
+        "negative_count": 2,
+        "negative_empty_count": 2
+      },
+      "withheld/search": {
+        "answerable_count": 6,
+        "success_at_5": 0.16666666666666666,
+        "mrr_at_5": 0.16666666666666666,
+        "negative_count": 2,
+        "negative_empty_count": 2
+      },
+      "withheld/relevance": {
+        "answerable_count": 6,
+        "success_at_5": 1,
+        "mrr_at_5": 1.0,
+        "negative_count": 2,
+        "negative_empty_count": 1
+      },
+      "withheld/context_material": {
+        "answerable_count": 6,
+        "success_at_5": 1,
+        "mrr_at_5": 1.0,
+        "negative_count": 2,
+        "negative_empty_count": 1
+      },
+      "withheld/default_diagnostic": {
+        "answerable_count": 6,
+        "success_at_5": 0.16666666666666666,
+        "mrr_at_5": 0.16666666666666666,
+        "negative_count": 2,
+        "negative_empty_count": 2
+      },
+      "all/search": {
+        "answerable_count": 20,
+        "success_at_5": 0.2,
+        "mrr_at_5": 0.2,
+        "negative_count": 4,
+        "negative_empty_count": 4
+      },
+      "all/relevance": {
+        "answerable_count": 20,
+        "success_at_5": 1,
+        "mrr_at_5": 0.9625,
+        "negative_count": 4,
+        "negative_empty_count": 2
+      },
+      "all/context_material": {
+        "answerable_count": 20,
+        "success_at_5": 1,
+        "mrr_at_5": 0.9625,
+        "negative_count": 4,
+        "negative_empty_count": 2
+      },
+      "all/default_diagnostic": {
+        "answerable_count": 20,
+        "success_at_5": 0.2,
+        "mrr_at_5": 0.2,
+        "negative_count": 4,
+        "negative_empty_count": 4
+      }
+    }
+  },
+  "measurement_structure_errors": {},
+  "parity_failures": [],
+  "source_and_corpus_unchanged_at_end": {
+    "A": true,
+    "B": true
+  },
+  "completed_at_utc": "2026-10-27T11:29:09.151662+00:00"
+}
diff --git a/evidence/knowledge-retrieval-20261027/results/validation-summary.md b/evidence/knowledge-retrieval-20261027/results/validation-summary.md
new file mode 100644
index 00000000..4b7265ad
--- /dev/null
+++ b/evidence/knowledge-retrieval-20261027/results/validation-summary.md
@@ -0,0 +1,46 @@
+# Local software and package validation
+
+These are software checks reported by the accountable orchestrator for the
+fixed runtime candidate, separate from the frozen CPU retrieval experiment.
+This summary deliberately excludes private local paths and raw command logs.
+
+| Check | Actual result and qualification |
+| --- | --- |
+| Whole local pytest run | **10 failed, 1,340 passed, 6 skipped, 2 errors, 135 subtests passed**, 953.11 seconds. This is a failed whole-suite run, not a green suite. |
+| Focused failure recheck | The same eleven selected failing/erroring test functions, including parameter variants, were rerun separately on baseline and candidate. Each reported **9 failed, 4 passed, 3 errors**. The CRLF case was then isolated from earlier fixture pollution: both sides reported **1 failed, 1 passed**, with `ValueError: Skill metadata must remain unchanged`. All initial **10 failures and 2 errors** were thereby reproduced on the baseline; neither run becomes a passing whole suite. |
+| Manifest regressions | All 20 newly added manifest tests passed in the local suite. The runtime author separately reported **80 targeted tests passed**; a separate selected baseline run reported **104 passed**. These scopes are distinct from whole-suite status. |
+| Wheel and sdist | Default isolated builds succeeded. An earlier `--no-isolation` attempt failed because setuptools was unavailable in that environment; later isolated build success supersedes that build attempt, not the failed pytest run. |
+| Installed artifact contents | All **469 packaged data files** matched the source byte-for-byte, and packaged runtime source matched the evaluated source. Wheel SHA-256: `4da1447422e5690573028373bda9baf8ab54b177031d7ba3d17586cb0533822d`. |
+| Evaluation validator | Normal and optimized-Python synthetic selfchecks passed. Explicit source-label checks rejected an absent support marker under `-O`. Exact measurement-set checks reject missing/duplicate rows and pairs. Original two-pair baseline-v-baseline mechanics data passed the revised structural/parity recheck; these remain dry-run evidence. |
+
+The orchestrator classified the local failures by their observed boundaries:
+oversized parametrized pytest names exceeding Windows' 32,767-character
+environment-variable limit; GBK decoding; four cases needing unavailable Bash;
+two Windows short/long temporary-path cases; one asset-manifest snapshot case;
+and a process-control case using unavailable `os.getpgrp`. The CRLF delimiter
+harness case failed in the initial whole run, but produced a fixture teardown
+error in the focused reruns after earlier fixture pollution. Preserve that
+status difference rather than treating the focused counts as an identical
+repetition of the initial run. Its subsequent isolated two-case rerun restored
+the original failing behavior on both revisions. Those isolated test durations
+are software-test execution time and are excluded from retrieval performance.
+
+The candidate and baseline focused rechecks establish reproducibility of those
+initial failures in this local environment. They do not establish Linux CI,
+all-platform portability or untested application behavior. A clean supported
+Linux environment should rerun the full suite, especially the shell/process,
+path and fixture-sensitive tests. CI status and revision-scoped independent
+review must be reported separately when available.
+
+After evidence integration, **23 inventory tests passed** in 1.10 seconds; the
+read-only inventory checker reported **no errors, 9 bundles and 120 hashes**.
+`ruff check src tests tools` passed. A wheel-only, site-disabled smoke read
+108 records and reproduced all four `DefaultStager` output hashes from the
+formal candidate result. Its first harness attempt failed an unresolved
+temporary short-path import-origin check before retrieval. Resolving the
+expected path allowed the smoke to pass; that retry established no package
+defect. Remote CI is pending at PR creation.
+
+The complete 24-case and 20-pair formal results are retained separately in
+`quality.json` and `benchmark.json`; both frozen gates passed. Software-check
+counts, package hashes and dry-run timings remain distinct from those results.
diff --git a/evidence/knowledge-retrieval-20261027/results/validator-selfcheck.json b/evidence/knowledge-retrieval-20261027/results/validator-selfcheck.json
new file mode 100644
index 00000000..92828a17
--- /dev/null
+++ b/evidence/knowledge-retrieval-20261027/results/validator-selfcheck.json
@@ -0,0 +1,4 @@
+{
+  "selfcheck": "passed",
+  "entry_python_optimization": 1
+}
diff --git a/evidence/knowledge-retrieval-20261027/validator/README.md b/evidence/knowledge-retrieval-20261027/validator/README.md
new file mode 100644
index 00000000..ead53dc6
--- /dev/null
+++ b/evidence/knowledge-retrieval-20261027/validator/README.md
@@ -0,0 +1,82 @@
+# Reproduce the bounded offline evaluation
+
+Only Python's standard library is required. Run the script using the same
+interpreter for both source checkouts. Commands below use operator-supplied
+paths rather than relying on a live main branch or an editable installation.
+Do not install or modify either evaluated checkout during a run.
+
+```text
+python -S validator/evaluate.py selfcheck
+python -S validator/evaluate.py validate --source <baseline-checkout> --dry-run
+python -S validator/evaluate.py identity --source <baseline-checkout> --dry-run --output results/baseline-identity.json
+```
+
+Identity JSON records the complete `python_sources.sha256`, the actual imported
+module paths relative to the checkout, their byte hashes, exact Git revision,
+complete corpus inventory and count, library bytes and Python/OS fields without
+personal absolute paths or host names. It uses `-S` and prepends the explicit
+checkout's `src` to exclude editable-install hooks. CLI subprocesses run the
+supported `python -S -m inferrsi` entrypoint under that same source selection.
+
+Before formal work, freeze the packet once, then obtain candidate identity:
+
+```text
+python -S validator/evaluate.py freeze
+python -S validator/evaluate.py check-freeze
+python -S validator/evaluate.py identity --source <candidate-checkout> --output results/candidate-identity.json
+```
+
+Copy `git_head` and `python_sources.sha256` from that identity. Candidate source
+and revision, equal complete corpus and all frozen files are checked before
+candidate retrieval/timing. Uncommitted tracked runtime sources are rejected.
+Paths are accepted locally but are not saved in shared results. The optional
+`--python <interpreter>` selects the worker/CLI interpreter explicitly.
+
+```text
+python -S validator/evaluate.py profile --source <baseline-checkout> --output results/baseline-profile.json
+python -S validator/evaluate.py trace --source <baseline-checkout> --output results/baseline-trace-129.json
+python -S validator/evaluate.py quality --baseline <baseline-checkout> --candidate <candidate-checkout> --candidate-sha <fixed-candidate-sha> --candidate-source-sha <fixed-source-digest> --split all --output results/quality.json
+python -S validator/evaluate.py bench --baseline <baseline-checkout> --candidate <candidate-checkout> --candidate-sha <fixed-candidate-sha> --candidate-source-sha <fixed-source-digest> --output results/benchmark.json
+```
+
+The withheld fixture must stay sealed from runtime author and orchestrator until
+the runtime candidate revision/source identity is fixed. Only `quality --split
+all` opens it, after the source preflight. Profile and runtime trace use fixed
+development cases. Trace covers the distinct catalog/default/relevance/context
+code paths introduced or retained by merged #129; it makes no claim about that
+earlier campaign's agent-task result.
+
+For a mechanics check only, use baseline for both roots with its revision and
+source digest, `quality --dry-run --case-limit 2` and `bench --dry-run --pairs 2`.
+Dry runs cannot open the withheld split and always remain INCONCLUSIVE. Keep
+their files separate from formal results. There is no formal sample-count
+override: changing it requires a new prospective protocol and freeze record.
+
+Benchmark writes every completed side to the output file immediately, retaining
+partial samples with INCONCLUSIVE status if interrupted. Do not pool retries or
+partial runs. A complete run stores 20 pairs, ten AB and ten BA, raw API/CLI
+nanoseconds, byte-equivalence digests, p50/p95 and the frozen bootstrap result.
+Actual duration is recorded; the protocol's duration estimate is not a cutoff.
+The script's benchmark PASS means only that its complete timing/parity gate
+passed. Overall campaign acceptance also requires full 24-case quality parity,
+runtime integrity tests and independent review.
+
+Completeness is checked by exact case/mode sets and pair IDs/orders, including
+duplicates and extras, rather than positional truncation. Missing or malformed
+measurement sets remain INCONCLUSIVE. Critical checks use explicit exceptions,
+so optimized Python cannot remove them; workers/CLI clear PYTHONOPTIMIZE and
+record optimization flags. See `plan/amendment-1.md` for the prospective review
+correction and preserved, unmeasured first freeze.
+
+Canonical digests cover full serializations including every resource hash and
+omission. Readable projections retain ordered IDs, exact revisions, inventory
+hashes/counts, context resource roles/hashes, omissions and material/context
+budgets without duplicating whole library bodies per query. Full CLI captures
+also retain the exact stdout byte count/hash and separate UTF-8 decode/JSON parse
+time. Catalog search is token-AND; relevance collect is weighted lexical. Their
+quality scores are reported independently, not as interchangeable algorithms.
+
+`freeze.json` commits exact bytes, so CRLF normalization, BOM insertion or any
+script/fixture edit invalidates the packet. Preserve bytes when copying it into
+the evidence tree. An intentional revision needs a new freeze and documented
+supersession rather than replacing prior measured evidence.
diff --git a/evidence/knowledge-retrieval-20261027/validator/development.json b/evidence/knowledge-retrieval-20261027/validator/development.json
new file mode 100644
index 00000000..2097991c
--- /dev/null
+++ b/evidence/knowledge-retrieval-20261027/validator/development.json
@@ -0,0 +1,23 @@
+{
+  "schema_version": 1,
+  "split": "development",
+  "construction": "Source-content-first synthetic questions. Labels were chosen from original shipped procedures before running either retrieval entrypoint. Labels are partial relevance judgements, not an exhaustive corpus annotation.",
+  "cases": [
+    {"id": "dev01", "category": "natural-language", "cue": "Why must checkpoint upload finish before cached staging buffers can be reused?", "backend": null, "relevant": [{"id": "inferrsi.common.skills.dcp_async_save_performance", "source": "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md", "support": "upload\ncompletion", "rationale": "The lifetime contract distinguishes staging completion from the upload completion needed to reuse cached CPU storage."}]},
+    {"id": "dev02", "category": "symbol", "cue": "DefaultStager", "backend": null, "relevant": [{"id": "inferrsi.common.skills.dcp_async_save_performance", "source": "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md", "support": "DefaultStager", "rationale": "The procedure states DefaultStager version gates, one-worker behavior and cached-storage reuse constraints."}]},
+    {"id": "dev03", "category": "multi-term", "cue": "safetensors index sharded downloads", "backend": null, "relevant": [{"id": "inferrsi.common.skills.weight_loading_optimization", "source": "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md", "support": "weight_map", "rationale": "The weight acquisition section verifies downloaded shards against index weight_map keys and distinguishes incomplete files."}]},
+    {"id": "dev04", "category": "natural-language", "cue": "How should model loading overhead be separated from per request latency?", "backend": null, "relevant": [{"id": "inferrsi.common.skills.weight_loading_optimization", "source": "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md", "support": "Fixed vs per-request split", "rationale": "Its startup accounting procedure separates once-per-process loading/warmup from per-request execution."}]},
+    {"id": "dev05", "category": "multi-term", "cue": "fio iodepth achieved depth", "backend": null, "relevant": [{"id": "inferrsi.common.skills.storage_io_measurement", "source": "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md", "support": "Requested versus achieved concurrency", "rationale": "The storage method requires validating achieved I/O depth and explains synchronous/buffered-engine limitations."}]},
+    {"id": "dev06", "category": "natural-language", "cue": "How can I attribute disk traffic to the loader process rather than the whole device?", "backend": null, "relevant": [{"id": "inferrsi.common.skills.storage_io_measurement", "source": "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md", "support": "pidstat", "rationale": "The skill separates per-device iostat measurements from per-process pidstat attribution and application outcomes."}]},
+    {"id": "dev07", "category": "backend-filter", "cue": "UB split buffer branch matrix", "backend": "ascend", "relevant": [{"id": "inferrsi.ascend.skills.ascendc_tiling_design", "source": "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md", "support": "Universal four-element template", "rationale": "The AscendC design procedure explicitly enumerates multi-core split, UB split, buffer plan and branch matrix."}]},
+    {"id": "dev08", "category": "backend-symbol", "cue": "multi_stream_lazy_reclaim", "backend": "ascend", "relevant": [{"id": "inferrsi.ascend.skills.torchnpu_device_memory_lifecycle", "source": "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md", "support": "multi_stream_lazy_reclaim", "rationale": "The source-derived allocator procedure explains the option's event-query tradeoff and retained-HBM verification."}]},
+    {"id": "dev09", "category": "multi-term", "cue": "LMCache skip_l1 retain", "backend": null, "relevant": [{"id": "inferrsi.common.skills.kv_tier_management", "source": "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md", "support": "skip_l1", "rationale": "The MP-mode procedure contrasts L1 store deletion with retained prefetched keys and describes their RAM costs."}]},
+    {"id": "dev10", "category": "symbol", "cue": "find_latest", "backend": null, "relevant": [{"id": "inferrsi.common.skills.checkpoint_recovery_consistency", "source": "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/SKILL.md", "support": "find_latest()", "rationale": "The recovery procedure describes selecting a complete global iteration using find_latest and complete rank coverage."}]},
+    {"id": "dev11", "category": "backend-natural", "cue": "Why can larger attention tiles cause register spills and lower parallelism?", "backend": "cuda", "relevant": [{"id": "inferrsi.cuda.skills.diagnose_attention_bottlenecks", "source": "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/SKILL.md", "support": "Tune tile size against resource ceilings", "rationale": "The attention diagnosis weighs reduced IO against register/shared-memory demand, occupancy and spills."}]},
+    {"id": "dev12", "category": "backend-filter", "cue": "NUMA CPU affinity placement", "backend": "cpu", "relevant": [{"id": "inferrsi.common.skills.os_process_placement", "source": "src/inferrsi/data/skill_library/hardware/host/os-process-placement/SKILL.md", "support": "CPU affinity: check, then pin", "rationale": "The common host procedure explains NUMA-local process placement, affinity checking and constant placement during comparisons."}]},
+    {"id": "dev13", "category": "multi-term", "cue": "gradient activation recomputation residency", "backend": null, "relevant": [{"id": "inferrsi.common.skills.plan_activation_residency", "source": "src/inferrsi/data/skill_library/methods/plan-activation-residency/SKILL.md", "support": "selective retention/recomputation", "rationale": "This procedure compares lifetime reordering, address packing and selective activation recomputation under a legal training DAG."}]},
+    {"id": "dev14", "category": "natural-language", "cue": "When does missing validator freeze make a performance claim inconclusive?", "backend": null, "relevant": [{"id": "inferrsi.common.skills.claim_contract", "source": "src/inferrsi/data/skill_library/workflows/claim-contract/SKILL.md", "support": "missing validator freeze", "rationale": "The claim contract explicitly assigns INCONCLUSIVE when a comparison lacks the validator freeze or other essential evidence."}]},
+    {"id": "dev15", "category": "no-answer-nonce", "cue": "quasarbanjo xenofloral v7", "backend": null, "relevant": [], "no_answer_rationale": "Constructed nonce identifiers have no intended systems-library answer; report actual returned items separately, not as recall."},
+    {"id": "dev16", "category": "no-answer-domain", "cue": "sourdough levain hydration recipe", "backend": "cpu", "relevant": [], "no_answer_rationale": "A constructed cooking request outside this inference/training procedures corpus; this is a labelled negative probe, not proof of exhaustive semantic absence."}
+  ]
+}
diff --git a/evidence/knowledge-retrieval-20261027/validator/evaluate.py b/evidence/knowledge-retrieval-20261027/validator/evaluate.py
new file mode 100644
index 00000000..d9f02c02
--- /dev/null
+++ b/evidence/knowledge-retrieval-20261027/validator/evaluate.py
@@ -0,0 +1,547 @@
+"""Issue #190's frozen, standard-library-only retrieval evaluation packet."""
+
+from __future__ import annotations
+
+import argparse
+import cProfile
+import hashlib
+import importlib
+import json
+import os
+import platform
+import pstats
+import random
+import statistics
+import subprocess
+import sys
+import time
+from collections import Counter
+from datetime import datetime, timezone
+from pathlib import Path
+
+PACKET = Path(__file__).resolve().parents[1]
+PROTOCOL = PACKET / "plan/protocol.json"
+FREEZE_FILES = (
+    "plan/protocol.json", "plan/README.md", "validator/development.json",
+    "validator/withheld.json", "validator/evaluate.py", "validator/README.md",
+    "plan/amendment-1.md", "plan/freeze-v1.json",
+)
+MODES = ("search", "relevance", "context_material", "default_diagnostic")
+
+
+def require(condition, message):
+    if not condition:
+        raise ValueError(message)
+
+
+def indexed_rows(rows, expected):
+    counts = Counter((r.get("mode"), r.get("case")) for r in rows)
+    errors = {"missing": sorted(expected - counts.keys(), key=repr),
+              "unexpected": sorted(counts.keys() - expected, key=repr),
+              "duplicates": sorted((key for key, n in counts.items() if n != 1), key=repr)}
+    return {(r.get("mode"), r.get("case")): r for r in rows}, {k: v for k, v in errors.items() if v}
+
+
+def compare_quality(left, right, selected):
+    expected = {(mode, case["id"]) for case in selected for mode in MODES}
+    a, a_errors = indexed_rows(left, expected)
+    b, b_errors = indexed_rows(right, expected)
+    errors = {side: error for side, error in (("A", a_errors), ("B", b_errors)) if error}
+    failures = [{"mode": mode, "case": case} for mode, case in sorted(expected)
+                if (mode, case) in a and (mode, case) in b and a[mode, case] != b[mode, case]]
+    return errors, failures
+
+
+def canonical(value):
+    return json.dumps(value, sort_keys=True, ensure_ascii=False,
+                      separators=(",", ":"), allow_nan=False).encode("utf-8")
+
+
+def sha(data):
+    return hashlib.sha256(data).hexdigest()
+
+
+def read(path):
+    return json.loads(Path(path).read_text(encoding="utf-8"))
+
+
+def write(path, value):
+    path = Path(path)
+    path.parent.mkdir(parents=True, exist_ok=True)
+    path.write_text(json.dumps(value, ensure_ascii=False, indent=2,
+                               allow_nan=False) + "\n", encoding="utf-8")
+
+
+def stamp():
+    return datetime.now(timezone.utc).isoformat()
+
+
+def freeze(make=False, dry=False):
+    hashes = {p: sha((PACKET / p).read_bytes()) for p in FREEZE_FILES}
+    path = PACKET / "plan/freeze.json"
+    if make:
+        if path.exists():
+            raise ValueError("freeze_exists: never replace an existing freeze silently")
+        value = {"schema_version": 1, "revision": 2,
+                 "supersedes_content_sha256": read(PACKET / "plan/freeze-v1.json")["content_sha256"],
+                 "frozen_at_utc": stamp(), "files": hashes,
+                 "content_sha256": sha(canonical(hashes))}
+        write(path, value)
+        return value
+    if not path.exists() and dry:
+        return {"state": "unfrozen_dry_run", "content_sha256": sha(canonical(hashes))}
+    value = read(path)
+    if value["files"] != hashes or value["content_sha256"] != sha(canonical(hashes)):
+        raise ValueError("freeze_mismatch: stop before importing or timing a candidate")
+    return value
+
+
+def cases(split="development", limit=None):
+    result = read(PACKET / "validator/development.json")["cases"]
+    if split == "all":
+        result += read(PACKET / "validator/withheld.json")["cases"]
+    return result[:limit] if limit is not None else result
+
+
+def inventory(root, paths):
+    files = {p.relative_to(root).as_posix(): {"sha256": sha(p.read_bytes()),
+                                           "bytes": p.stat().st_size}
+             for p in sorted(set(paths)) if p.is_file()
+             and "__pycache__" not in p.parts and p.suffix != ".pyc"}
+    return {"files": files, "sha256": sha(canonical(files)), "file_count": len(files),
+            "bytes": sum(v["bytes"] for v in files.values())}
+
+
+def git(root, *args):
+    run = subprocess.run(["git", "-C", str(root), *args], capture_output=True, check=True)
+    return run.stdout.decode("utf-8").strip()
+
+
+def source_identity(root):
+    data = root / "src/inferrsi/data"
+    source = inventory(root, (root / "src/inferrsi").rglob("*.py"))
+    corpus = inventory(root, [*(data / "repo_packs").rglob("*.json"),
+                              *(data / "knowledge_packs").rglob("*"),
+                              *(data / "skill_library").rglob("*")])
+    library = inventory(root, (data / "skill_library").rglob("*"))
+    records = [p for p in (data / "repo_packs").rglob("*.json") if "knowledge" in p.parts]
+    records += list((data / "knowledge_packs").rglob("*.json"))
+    record_values = [read(p) for p in records]
+    return {"git_head": git(root, "rev-parse", "HEAD"),
+            "tracked_source_dirty": bool(git(root, "status", "--porcelain",
+                                              "--untracked-files=no", "--", "src/inferrsi")),
+            "python_sources": source, "corpus": corpus,
+            "library": {k: v for k, v in library.items() if k != "files"},
+            "knowledge_record_count": len(record_values),
+            "packaged_skill_count": sum(r.get("index_type") == "packaged-skill"
+                                        for r in record_values)}
+
+
+def fixture_check(root, selected):
+    data = root / "src/inferrsi/data"
+    paths = [p for p in (data / "repo_packs").rglob("*.json") if "knowledge" in p.parts]
+    paths += list((data / "knowledge_packs").rglob("*.json"))
+    records = {r["id"]: r for p in paths if "id" in (r := read(p))}
+    require(len({c["id"] for c in selected}) == len(selected), "duplicate_fixture_case")
+    for case in selected:
+        require(case["cue"].strip() and case["category"], "invalid_fixture_case: " + case["id"])
+        if not case["relevant"]:
+            require(case["no_answer_rationale"], "missing_negative_rationale: " + case["id"])
+        for label in case["relevant"]:
+            path = (root / label["source"]).resolve()
+            require(path.is_relative_to(root) and path.is_file(), "invalid_label_source: " + case["id"])
+            require(label["support"] in path.read_text(encoding="utf-8"), "label_support_absent: " + case["id"])
+            record = records[label["id"]]
+            require(label["source"] in record["source_files"], "unbound_label_source: " + case["id"])
+            require(label["rationale"] and record["repo"] == "inferrsi", "invalid_label_scope: " + case["id"])
+            require(case["backend"] is None or record["backend"] in (case["backend"], "common"), "invalid_label_backend: " + case["id"])
+    return {"cases": len(selected), "answerable": sum(bool(c["relevant"]) for c in selected),
+            "negative": sum(not c["relevant"] for c in selected),
+            "fixture_sha256": sha(canonical(selected))}
+
+
+def load_source(root):
+    sys.path.insert(0, str(root / "src"))
+    modules = {}
+    for name in ("inferrsi", "inferrsi.catalog", "inferrsi.knowledge_operations", "inferrsi.cli"):
+        module = importlib.import_module(name)
+        path = Path(module.__file__).resolve()
+        if not path.is_relative_to(root / "src/inferrsi"):
+            raise ValueError("module_origin_mismatch: " + name)
+        modules[name] = {"source": path.relative_to(root).as_posix(),
+                         "sha256": sha(path.read_bytes())}
+    return modules
+
+
+def environment():
+    clock = time.get_clock_info("perf_counter")
+    return {"python_version": platform.python_version(),
+            "implementation": platform.python_implementation(),
+            "python_executable_sha256": sha(Path(sys.executable).read_bytes()),
+            "system": platform.system(), "release": platform.release(),
+            "machine": platform.machine(), "logical_cpu_count": os.cpu_count(),
+            "perf_counter_resolution_seconds": clock.resolution,
+            "site_initializers_disabled": bool(sys.flags.no_site),
+            "python_optimization": sys.flags.optimize,
+            "filesystem_cache": "not controlled; process cold is not OS-cache cold"}
+
+
+def invoke(mode, case, reader=None):
+    from inferrsi.catalog import search_knowledge
+    from inferrsi.knowledge_operations import KnowledgeReader
+    common = {"repo": "inferrsi", "backend": case["backend"], "limit": 5}
+    if mode == "search":
+        return search_knowledge(case["cue"], **common)
+    reader = reader if reader is not None else KnowledgeReader()
+    if mode == "default_diagnostic":
+        return reader.collect(case["cue"], **common)
+    options = {"material": True, "max_bytes": 1048576, "context_bytes": 65536} if mode == "context_material" else {}
+    return reader.collect("", relevance=case["cue"], **common, **options)
+
+
+def snapshot(value, compact=False):
+    encoded = canonical(value)
+    result = {"sha256": sha(encoded), "serialized_bytes": len(encoded)}
+    if compact:
+        return result
+    items = value if isinstance(value, list) else value["items"]
+    result["ordered_items"] = [
+        {"id": item["id"], **{k: item[k] for k in ("ref", "revision") if k in item},
+         **({"resource_count": len(item["resources"]),
+             "resource_hashes_sha256": sha(canonical(item["resources"])),
+             "dependencies_sha256": sha(canonical(item["dependencies"]))}
+            if "resources" in item else {})} for item in items]
+    if isinstance(value, dict):
+        for key in ("no_match", "omissions", "collisions", "material_bytes", "selection",
+                    "context_omissions", "context_bytes", "context_budget"):
+            if key in value:
+                result[key] = value[key]
+        if "contexts" in value:
+            result["contexts"] = [
+                {k: v for k, v in context.items() if k != "resources"}
+                | {"resources": [{k: v for k, v in resource.items() if k != "content"}
+                                  for resource in context["resources"]]}
+                for context in value["contexts"]]
+    return result
+
+
+def worker(root, request):
+    modules = load_source(root)
+    if "expected_source_sha256" in request:
+        actual = inventory(root, (root / "src/inferrsi").rglob("*.py"))["sha256"]
+        if actual != request["expected_source_sha256"]:
+            raise ValueError("worker_source_digest_changed_before_measurement")
+    from inferrsi.knowledge_operations import KnowledgeReader
+    action = request["action"]
+    if action == "identity":
+        return {"modules": modules, "environment": environment(), **source_identity(root)}
+    selected = request["cases"]
+    if action in ("quality", "trace"):
+        outputs = []
+        for case in selected:
+            for mode in MODES:
+                value = invoke(mode, case)
+                ids = [v["id"] for v in (value if isinstance(value, list) else value["items"])]
+                ranks = [i + 1 for i, identity in enumerate(ids[:5])
+                         if identity in {label["id"] for label in case["relevant"]}]
+                outputs.append({"case": case["id"], "mode": mode,
+                                "answerable": bool(case["relevant"]),
+                                "success_at_5": int(bool(ranks)),
+                                "reciprocal_rank_at_5": 1 / ranks[0] if ranks else 0,
+                                "output": snapshot(value)})
+        return {"modules": modules, "outputs": outputs}
+    if action == "profile":
+        profiler = cProfile.Profile()
+        profiler.enable()
+        invoke("relevance", selected[0])
+        profiler.disable()
+        rows = []
+        for (filename, line, name), (primitive, calls, total, cumulative, _) in pstats.Stats(profiler).stats.items():
+            path = Path(filename)
+            label = path.relative_to(root).as_posix() if path.is_relative_to(root) else "<external>/" + path.name
+            rows.append({"source": label, "line": line, "function": name,
+                         "primitive_calls": primitive, "calls": calls,
+                         "total_seconds": total, "cumulative_seconds": cumulative})
+        return {"modules": modules, "case": selected[0]["id"],
+                "mode": "fresh-reader relevance, cProfile diagnostic only",
+                "profile": sorted(rows, key=lambda r: -r["cumulative_seconds"])[:80]}
+    reader = KnowledgeReader()
+    warmups = [("relevance", case) for case in selected]
+    warmups += [("default_diagnostic", selected[1]), ("context_material", selected[2])]
+    for mode, case in warmups:
+        invoke(mode, case, reader)
+    cache = {"records_populated": reader._records is not None, "library_populated": reader._library is not None}
+    require(all(cache.values()), "warmup_cache_not_populated")
+    measurements = []
+    workloads = [(mode, case) for case in selected for mode in ("hot_relevance", "cold_relevance", "search")]
+    workloads += [(mode, selected[index]) for index, modes in
+                  ((1, ("hot_default_diagnostic", "cold_default_diagnostic")),
+                   (2, ("hot_context_material", "cold_context_material"))) for mode in modes]
+    for mode, case in workloads:
+        operation = mode.removeprefix("hot_").removeprefix("cold_")
+        start = time.perf_counter_ns()
+        value = invoke(operation, case, reader if mode.startswith("hot_") else None)
+        elapsed = time.perf_counter_ns() - start
+        measurements.append({"mode": mode, "case": case["id"], "ns": elapsed,
+                             "output": snapshot(value, compact=True)})
+    return {"modules": modules, "warmup_cache_state": cache, "api": measurements}
+
+
+def child(python, root, request):
+    env = {**os.environ, "PYTHONPATH": str(root / "src"), "PYTHONDONTWRITEBYTECODE": "1",
+           "PYTHONIOENCODING": "utf-8", "PYTHONOPTIMIZE": "0"}
+    run = subprocess.run([python, "-S", str(Path(__file__).resolve()), "_worker", "--source", str(root)],
+                         input=canonical(request), capture_output=True, env=env, cwd=PACKET)
+    if run.returncode:
+        raise ValueError("isolated_worker_failed: " + run.stderr.decode("utf-8", "replace")[-400:].replace(str(root), "<source>"))
+    return json.loads(run.stdout)
+
+
+def cli(python, root, mode, case, expected_source):
+    if inventory(root, (root / "src/inferrsi").rglob("*.py"))["sha256"] != expected_source:
+        raise ValueError("CLI_source_digest_changed_before_measurement")
+    args = ["knowledge", "search", case["cue"], "--repo-type", "inferrsi"] if mode == "search" else ["collect"]
+    if mode == "default_diagnostic":
+        args += [case["cue"]]
+    elif mode != "search":
+        args += ["--relevance", case["cue"]]
+    if mode == "context_material":
+        args += ["--material", "--max-bytes", "1048576", "--context-bytes", "65536"]
+    args += ["--limit", "5", "--json"]
+    env = {**os.environ, "PYTHONPATH": str(root / "src"), "PYTHONDONTWRITEBYTECODE": "1", "PYTHONIOENCODING": "utf-8", "PYTHONOPTIMIZE": "0"}
+    start = time.perf_counter_ns()
+    run = subprocess.run([python, "-S", "-m", "inferrsi", *args], capture_output=True, env=env, cwd=PACKET)
+    elapsed = time.perf_counter_ns() - start
+    if run.returncode:
+        raise ValueError("CLI_invocation_failed: mode=" + mode)
+    start = time.perf_counter_ns()
+    value = json.loads(run.stdout.decode("utf-8"))
+    decode_ns = time.perf_counter_ns() - start
+    return {"mode": mode, "case": case["id"], "process_ns": elapsed, "decode_ns": decode_ns,
+            "scope": "repo=inferrsi" if mode == "search" else "unfiltered packaged CLI diagnostic",
+            "stdout_sha256": sha(run.stdout), "stdout_bytes": len(run.stdout),
+            "output": snapshot(value, compact=True)}
+
+
+def percentile(values, p):
+    ordered = sorted(values)
+    position = (len(ordered) - 1) * p
+    low = int(position)
+    return ordered[low] + (ordered[min(low + 1, len(ordered) - 1)] - ordered[low]) * (position - low)
+
+
+def bootstrap(reductions, protocol):
+    rule = protocol["performance"]["bootstrap"]
+    rng = random.Random(rule["seed"])
+    draws = [statistics.median(rng.choices(reductions, k=len(reductions))) for _ in range(rule["iterations"])]
+    return {"paired_median_reduction": statistics.median(reductions),
+            "confidence_95": [percentile(draws, p) for p in rule["quantiles"]],
+            "paired_reductions": reductions, "bootstrap": rule}
+
+
+def quality_summary(rows):
+    result = {}
+    for split in ("development", "withheld", "all"):
+        subset = [r for r in rows if split == "all" or r["case"].startswith("dev" if split == "development" else "withheld")]
+        for mode in MODES:
+            positive = [r for r in subset if r["mode"] == mode and r["answerable"]]
+            negative = [r for r in subset if r["mode"] == mode and not r["answerable"]]
+            if not positive and not negative:
+                continue
+            result[split + "/" + mode] = {
+                "answerable_count": len(positive),
+                "success_at_5": statistics.mean(r["success_at_5"] for r in positive) if positive else None,
+                "mrr_at_5": statistics.mean(r["reciprocal_rank_at_5"] for r in positive) if positive else None,
+                "negative_count": len(negative),
+                "negative_empty_count": sum(not r["output"]["ordered_items"] for r in negative)}
+    return result
+
+
+def preflight(args, protocol, frozen):
+    roots = {"A": Path(args.baseline).resolve(), "B": Path(args.candidate).resolve()}
+    identities = {side: child(args.python, root, {"action": "identity"}) for side, root in roots.items()}
+    if identities["A"]["git_head"] != protocol["baseline_sha"]:
+        raise ValueError("baseline_revision_mismatch")
+    if identities["B"]["git_head"] != args.candidate_sha:
+        raise ValueError("candidate_revision_mismatch")
+    if identities["B"]["python_sources"]["sha256"] != args.candidate_source_sha:
+        raise ValueError("candidate_source_digest_mismatch")
+    if any(v["tracked_source_dirty"] for v in identities.values()):
+        raise ValueError("tracked_source_dirty")
+    if identities["A"]["corpus"]["sha256"] != identities["B"]["corpus"]["sha256"]:
+        raise ValueError("corpus_mismatch")
+    if identities["A"]["environment"] != identities["B"]["environment"]:
+        raise ValueError("interpreter_or_environment_mismatch")
+    return roots, {"started_at_utc": stamp(), "dry_run": args.dry_run,
+                   "entry_python_optimization": sys.flags.optimize,
+                   "verdict": "INCONCLUSIVE", "freeze_content_sha256": frozen["content_sha256"],
+                   "identities": identities}
+
+
+def benchmark_summary(pairs, protocol, expected_pairs=None):
+    expected_pairs = protocol["performance"]["pairs"] if expected_pairs is None else expected_pairs
+    hot = protocol["performance"]["hot_cases"]
+    api_keys = {(mode, case) for case in hot for mode in ("hot_relevance", "cold_relevance", "search")}
+    api_keys |= {(mode, hot[index]) for index, modes in
+                 ((1, ("hot_default_diagnostic", "cold_default_diagnostic")),
+                  (2, ("hot_context_material", "cold_context_material"))) for mode in modes}
+    cli_keys = {(mode, hot[2] if mode == "context_material" else hot[1]) for mode in MODES}
+    errors = []
+    ids = [pair.get("pair") for pair in pairs]
+    if any(type(i) is not int for i in ids) or Counter(ids) != Counter(range(expected_pairs)):
+        errors.append({"pair_ids": ids, "expected_pair_ids": list(range(expected_pairs))})
+    for pair in pairs:
+        index = pair.get("pair")
+        if type(index) is not int or pair.get("order") != ("AB" if index % 2 == 0 else "BA"):
+            errors.append({"pair": index, "invalid_order": pair.get("order")})
+        for side in ("A", "B"):
+            for family, keys in (("api", api_keys), ("cli", cli_keys)):
+                _, error = indexed_rows(pair.get(side, {}).get(family, []), keys)
+                if error:
+                    errors.append({"pair": index, "side": side, "family": family, **error})
+    if errors:
+        return {"measurement_structure_errors": errors, "primary": None,
+                "performance_gate_passed": False, "output_parity_assessed": False,
+                "output_parity_failures": [], "latency_ns": {}}
+    grouped = {}
+    reductions = []
+    parity_failures = []
+    for pair in pairs:
+        for side in ("A", "B"):
+            for row in pair[side]["api"]:
+                grouped.setdefault("api/" + row["mode"] + "/" + row["case"] + "/" + side, []).append(row["ns"])
+            for row in pair[side]["cli"]:
+                for metric in ("process_ns", "decode_ns"):
+                    grouped.setdefault("cli/" + row["mode"] + "/" + row["case"] + "/" + metric + "/" + side, []).append(row[metric])
+        medians = {side: statistics.median(r["ns"] for r in pair[side]["api"] if r["mode"] == "hot_relevance") for side in ("A", "B")}
+        reductions.append(1 - medians["B"] / medians["A"])
+        for family in ("api", "cli"):
+            keys = api_keys if family == "api" else cli_keys
+            left_rows, _ = indexed_rows(pair["A"][family], keys)
+            right_rows, _ = indexed_rows(pair["B"][family], keys)
+            for key in sorted(keys):
+                left, right = left_rows[key], right_rows[key]
+                if left["output"] != right["output"] or (family == "cli" and left["stdout_sha256"] != right["stdout_sha256"]):
+                    parity_failures.append({"pair": pair["pair"], "family": family, "mode": left["mode"], "case": left["case"]})
+    estimate = bootstrap(reductions, protocol)
+    gate = protocol["performance"]["gate"]
+    passed = (estimate["paired_median_reduction"] >= gate["minimum_paired_median_reduction"]
+              and estimate["confidence_95"][0] > gate["lower_95_percent_bound_must_exceed"])
+    return {"primary": estimate, "performance_gate_passed": passed,
+            "measurement_structure_errors": [], "output_parity_assessed": True,
+            "output_parity_failures": parity_failures,
+            "latency_ns": {key: {"n": len(values), "p50": percentile(values, .5), "p95": percentile(values, .95)}
+                           for key, values in sorted(grouped.items())}}
+
+
+def main():
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("action", choices=("freeze", "check-freeze", "validate", "identity", "profile", "trace", "quality", "bench", "selfcheck", "_worker"))
+    parser.add_argument("--source")
+    parser.add_argument("--baseline")
+    parser.add_argument("--candidate")
+    parser.add_argument("--candidate-sha")
+    parser.add_argument("--candidate-source-sha")
+    parser.add_argument("--python", default=sys.executable)
+    parser.add_argument("--output")
+    parser.add_argument("--split", choices=("development", "all"), default="development")
+    parser.add_argument("--dry-run", action="store_true")
+    parser.add_argument("--case-limit", type=int)
+    parser.add_argument("--pairs", type=int)
+    args = parser.parse_args()
+    if args.action == "_worker":
+        print(canonical(worker(Path(args.source).resolve(), json.load(sys.stdin))).decode("utf-8"))
+        return
+    protocol = read(PROTOCOL)
+    if args.action == "freeze":
+        value = freeze(make=True)
+    elif args.action == "selfcheck":
+        require(percentile([1, 3], .95) == 2.9, "percentile_selfcheck")
+        require(bootstrap([.3] * 20, protocol)["confidence_95"] == [.3, .3], "bootstrap_selfcheck")
+        require(snapshot({"items": [], "no_match": True})["ordered_items"] == [], "snapshot_selfcheck")
+        selected = [{"id": f"synthetic{i:02}"} for i in range(24)]
+        quality = [{"case": c["id"], "mode": mode} for c in selected for mode in MODES]
+        require(compare_quality(quality, quality[:-1], selected)[0], "missing_quality_row_accepted")
+        require(compare_quality(quality, quality + [quality[0]], selected)[0], "duplicate_quality_row_accepted")
+        hot = protocol["performance"]["hot_cases"]
+        api = [{"mode": mode, "case": case, "ns": 100, "output": {}} for case in hot for mode in ("hot_relevance", "cold_relevance", "search")]
+        api += [{"mode": mode, "case": hot[index], "ns": 100, "output": {}} for index, modes in ((1, ("hot_default_diagnostic", "cold_default_diagnostic")), (2, ("hot_context_material", "cold_context_material"))) for mode in modes]
+        cli_rows = [{"mode": mode, "case": hot[2] if mode == "context_material" else hot[1], "process_ns": 100, "decode_ns": 10, "output": {}, "stdout_sha256": "synthetic"} for mode in MODES]
+        pairs = [{"pair": i, "order": "AB" if i % 2 == 0 else "BA", "A": {"api": api, "cli": cli_rows}, "B": {"api": [{**r, "ns": 50} for r in api], "cli": cli_rows}} for i in range(20)]
+        require(benchmark_summary(pairs, protocol)["performance_gate_passed"], "complete_synthetic_pairs_rejected")
+        for pair in pairs:
+            pair["B"] = {"api": api[:-1], "cli": []}
+        require(benchmark_summary(pairs, protocol)["measurement_structure_errors"], "missing_benchmark_rows_accepted")
+        pairs[0]["pair"] = 1
+        require(benchmark_summary(pairs, protocol)["measurement_structure_errors"], "duplicate_pair_accepted")
+        value = {"selfcheck": "passed", "entry_python_optimization": sys.flags.optimize}
+    else:
+        frozen = freeze(dry=args.dry_run)
+        if not args.dry_run and (args.case_limit is not None or args.pairs not in (None, protocol["performance"]["pairs"])):
+            raise ValueError("formal_sample_count_change_requires_new_protocol_freeze")
+        if args.dry_run and args.split == "all":
+            raise ValueError("dry_run_must_not_open_withheld")
+        if args.action == "check-freeze":
+            value = frozen
+        elif args.action == "identity":
+            value = child(args.python, Path(args.source).resolve(), {"action": "identity"})
+        elif args.action in ("validate", "profile", "trace"):
+            if args.split == "all":
+                raise ValueError("withheld_requires_quality_preflight_and_candidate_freeze")
+            selected = cases(limit=args.case_limit)
+            value = fixture_check(Path(args.source).resolve(), selected)
+            if args.action != "validate":
+                cue = next(c for c in selected if c["id"] == ("dev05" if args.action == "profile" else "dev09"))
+                value = child(args.python, Path(args.source).resolve(), {"action": args.action, "cases": [cue]})
+        else:
+            if not all((args.baseline, args.candidate, args.candidate_sha, args.candidate_source_sha)):
+                raise ValueError("explicit_baseline_candidate_revision_and_source_digest_required")
+            roots, value = preflight(args, protocol, frozen)
+            if args.action == "quality":
+                selected = cases(args.split, args.case_limit)
+                value["fixture"] = fixture_check(roots["A"], selected)
+                value["variants"] = {side: child(args.python, root, {"action": "quality", "cases": selected, "expected_source_sha256": value["identities"][side]["python_sources"]["sha256"]}) for side, root in roots.items()}
+                value["metrics"] = {side: quality_summary(v["outputs"]) for side, v in value["variants"].items()}
+                left, right = (value["variants"][side]["outputs"] for side in ("A", "B"))
+                value["measurement_structure_errors"], value["parity_failures"] = compare_quality(left, right, selected)
+                value["verdict"] = "INCONCLUSIVE" if value["measurement_structure_errors"] else "FAIL" if value["parity_failures"] else "PASS" if args.split == "all" and not args.dry_run else "INCONCLUSIVE"
+            else:
+                if not args.output:
+                    raise ValueError("bench_requires_output_for_incremental_raw_sample_retention")
+                selected = [c for c in cases() if c["id"] in protocol["performance"]["hot_cases"]]
+                value["pairs"] = []
+                start = time.perf_counter_ns()
+                for index in range(args.pairs or protocol["performance"]["pairs"]):
+                    pair = {"pair": index, "order": "AB" if index % 2 == 0 else "BA"}
+                    value["pairs"].append(pair)
+                    for side in pair["order"]:
+                        source_sha = value["identities"][side]["python_sources"]["sha256"]
+                        pair[side] = child(args.python, roots[side], {"action": "bench", "cases": selected, "expected_source_sha256": source_sha})
+                        pair[side]["cli"] = [cli(args.python, roots[side], mode, selected[2] if mode == "context_material" else selected[1], source_sha) for mode in ("search", "default_diagnostic", "relevance", "context_material")]
+                        value["elapsed_ns"] = time.perf_counter_ns() - start
+                        write(args.output, value)
+                    print("completed paired group " + str(index + 1), file=sys.stderr, flush=True)
+                value["summary"] = benchmark_summary(value["pairs"], protocol, args.pairs or protocol["performance"]["pairs"])
+                sufficient = len(value["pairs"]) == protocol["performance"]["pairs"] and not args.dry_run
+                value["verdict"] = "INCONCLUSIVE" if not sufficient or value["summary"]["measurement_structure_errors"] else "PASS" if value["summary"]["performance_gate_passed"] and not value["summary"]["output_parity_failures"] else "FAIL"
+            value["source_and_corpus_unchanged_at_end"] = {}
+            for side, root in roots.items():
+                final = child(args.python, root, {"action": "identity"})
+                same = all(final[key]["sha256"] == value["identities"][side][key]["sha256"] for key in ("python_sources", "corpus"))
+                value["source_and_corpus_unchanged_at_end"][side] = same
+                if not same:
+                    value["verdict"] = "INCONCLUSIVE"
+            value["completed_at_utc"] = stamp()
+    if args.output:
+        write(args.output, value)
+    # Keep stdout free of fixture text, personal paths and full corpus inventories.
+    print(json.dumps({k: value[k] for k in ("verdict", "dry_run", "cases", "answerable", "negative", "content_sha256", "selfcheck") if k in value}))
+
+
+if __name__ == "__main__":
+    try:
+        main()
+    except (ValueError, AssertionError, OSError, subprocess.SubprocessError) as exc:
+        # Paths supplied by the operator stay local; do not leak full tracebacks.
+        print("evaluation stopped: " + type(exc).__name__ + ": " + str(exc).replace(str(PACKET), "<packet>"), file=sys.stderr)
+        raise SystemExit(2)
diff --git a/evidence/knowledge-retrieval-20261027/validator/withheld.json b/evidence/knowledge-retrieval-20261027/validator/withheld.json
new file mode 100644
index 00000000..17232fa4
--- /dev/null
+++ b/evidence/knowledge-retrieval-20261027/validator/withheld.json
@@ -0,0 +1,15 @@
+{
+  "schema_version": 1,
+  "split": "withheld",
+  "construction": "Sealed source-content-first synthetic questions and partial relevance judgements, authored without retrieval rankings. Release only after runtime candidate revision/source digest freeze.",
+  "cases": [
+    {"id": "withheld01", "category": "symbol", "cue": "prefetch_queue_depth executor_statistics", "backend": null, "relevant": [{"id": "inferrsi.common.skills.training_input_pipeline", "source": "src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md", "support": "executor_statistics()", "rationale": "The DALI section explains queue buffering, executor statistics and execution-mode gates together."}]},
+    {"id": "withheld02", "category": "natural-language", "cue": "Can a serving simulator certify latency SLOs near queue capacity?", "backend": null, "relevant": [{"id": "inferrsi.common.skills.calibrate_serving_simulation", "source": "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/SKILL.md", "support": "near capacity", "rationale": "The calibration procedure requires dynamic queue-delay validation at the intended load and rejects near-capacity predictive certification without evidence."}]},
+    {"id": "withheld03", "category": "multi-term", "cue": "FastServe suspended KV preemption progress", "backend": null, "relevant": [{"id": "inferrsi.common.skills.assess_serving_preemption", "source": "src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md", "support": "Protect progress", "rationale": "The preemption procedure counts suspended state and requires progress under pressure, rather than treating priority alone as a guarantee."}]},
+    {"id": "withheld04", "category": "backend-symbol", "cue": "ACLGraph stable replay buffers", "backend": "ascend", "relevant": [{"id": "inferrsi.ascend.skills.ascend_aclgraph_capture_replay", "source": "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/SKILL.md", "support": "caller-managed stable replay buffers", "rationale": "The ACLGraph procedure states the caller's stable-buffer responsibility and the limited coverage of address assertions."}]},
+    {"id": "withheld05", "category": "natural-language", "cue": "How is moving a request between replicas different from changing its sequence parallel group?", "backend": null, "relevant": [{"id": "inferrsi.common.skills.evaluate_elastic_serving_placement", "source": "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/SKILL.md", "support": "Choose the actual decision", "rationale": "This procedure directly separates Llumnix replica migration from LoongServe sequence-parallel group changes and their state costs."}]},
+    {"id": "withheld06", "category": "multi-term", "cue": "roofline analytical empirical end-to-end ceiling", "backend": "cuda", "relevant": [{"id": "inferrsi.common.skills.performance_ceiling_analysis", "source": "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/SKILL.md", "support": "Three-level ceiling model", "rationale": "The headroom procedure distinguishes analytical, attainable component and end-to-end ceilings and supplies a bounded roofline arithmetic helper."}]},
+    {"id": "withheld07", "category": "no-answer-nonce", "cue": "zephyrcodex latticeotter q83", "backend": null, "relevant": [], "no_answer_rationale": "Constructed nonce terms with no intended library answer; any returned candidates remain visible as negative-query behavior."},
+    {"id": "withheld08", "category": "no-answer-domain", "cue": "baroque harpsichord tuning temperament", "backend": "rocm", "relevant": [], "no_answer_rationale": "A synthetic music-history request outside the shipped systems procedures; not an exhaustive semantic absence assertion."}
+  ]
+}
diff --git a/src/inferrsi/knowledge_operations.py b/src/inferrsi/knowledge_operations.py
index 58121f0a..f9e9761b 100644
--- a/src/inferrsi/knowledge_operations.py
+++ b/src/inferrsi/knowledge_operations.py
@@ -118,18 +118,22 @@ def manifest(self) -> dict:
         if record["entrypoint"] not in self.content:
             raise ValueError("Missing entrypoint in complete bundle")
         for required in record.get("required_resources", []):
+            # Present string keys already passed the content path validation.
+            if isinstance(required, str) and required in self.content:
+                continue
             if safe_path(required) not in self.content:
                 raise ValueError(f"Missing required resource: {required}")
+        resources = {path: _hash(data) for path, data in sorted(self.content.items())}
         for dependency, revision in self.dependencies.items():
             safe_path(dependency)
             if not re.fullmatch(r"sha256:[0-9a-f]{64}", revision):
                 raise ValueError("Dependency requires a resolved content revision")
             if dependency in self.content:
-                actual = "sha256:" + _hash(self.content[dependency])
+                actual = "sha256:" + resources[dependency]
             else:
                 members = {
-                    path: _hash(data)
-                    for path, data in sorted(self.content.items())
+                    path: digest
+                    for path, digest in resources.items()
                     if path.startswith(dependency + "/")
                 }
                 if not members:
@@ -139,7 +143,7 @@ def manifest(self) -> dict:
                 raise RevisionConflict(f"Dependency bytes disagree with revision: {dependency}")
         return {
             "metadata": record,
-            "resources": {path: _hash(data) for path, data in sorted(self.content.items())},
+            "resources": resources,
             "dependencies": dict(sorted(self.dependencies.items())),
         }
 
diff --git a/tests/test_bundle_manifest.py b/tests/test_bundle_manifest.py
new file mode 100644
index 00000000..1b0b8d8d
--- /dev/null
+++ b/tests/test_bundle_manifest.py
@@ -0,0 +1,146 @@
+import hashlib
+import json
+from collections import Counter
+
+import pytest
+
+from inferrsi import knowledge_operations
+from inferrsi.knowledge_operations import Bundle, RevisionConflict
+
+
+def _hash(data):
+    return hashlib.sha256(data).hexdigest()
+
+
+def _json(value):
+    return json.dumps(value, sort_keys=True, ensure_ascii=False, separators=(",", ":")).encode()
+
+
+@pytest.fixture
+def bundle():
+    content = {
+        "support/z.txt": b"last reference",
+        "SKILL.md": "Core procedure: café".encode(),
+        "support/a.txt": b"first reference",
+    }
+    members = {path: _hash(data) for path, data in content.items() if path.startswith("support/")}
+    return Bundle(
+        {
+            "id": "fixture",
+            "kind": "skill",
+            "sources": [{"type": "unknown"}],
+            "applicability": {},
+            "state": "candidate",
+            "limitations": [],
+            "attribution": "fixture",
+            "entrypoint": "SKILL.md",
+            "required_resources": ["SKILL.md", "support/a.txt"],
+        },
+        content,
+        {
+            "support/a.txt": "sha256:" + _hash(content["support/a.txt"]),
+            "support": "sha256:" + _hash(_json(members)),
+        },
+    )
+
+
+def test_manifest_and_revision_preserve_complete_digest_formula(bundle):
+    expected = {
+        "metadata": bundle.metadata,
+        "resources": {path: _hash(data) for path, data in sorted(bundle.content.items())},
+        "dependencies": dict(sorted(bundle.dependencies.items())),
+    }
+    manifest = bundle.manifest()
+    assert manifest == expected
+    assert list(manifest["resources"]) == sorted(bundle.content)
+    assert bundle.revision == "sha256:" + _hash(_json(expected))
+    assert bundle.view(material=True) == {
+        **bundle.metadata,
+        "revision": bundle.revision,
+        "resources": expected["resources"],
+        "dependencies": expected["dependencies"],
+        "content": bundle.content["SKILL.md"].decode(),
+    }
+
+
+def test_manifest_hashes_each_resource_once_per_call(bundle, monkeypatch):
+    calls = Counter()
+
+    def counted(data):
+        calls[data] += 1
+        return _hash(data)
+
+    monkeypatch.setattr(knowledge_operations, "_hash", counted)
+    for count in (1, 2):
+        bundle.manifest()
+        assert all(calls[data] == count for data in bundle.content.values())
+        assert sum(calls.values()) == count * (len(bundle.content) + 1)
+
+
+def test_manifest_rechecks_changed_bytes_and_metadata(bundle):
+    original = bundle.revision
+    bundle.content["SKILL.md"] = b"Changed core procedure"
+    changed = bundle.revision
+    assert changed != original
+    assert bundle.manifest()["resources"]["SKILL.md"] == _hash(bundle.content["SKILL.md"])
+    bundle.metadata["state"] = "changed"
+    assert bundle.revision != changed
+
+
+@pytest.mark.parametrize("path", ["support/a.txt", "support/z.txt"])
+def test_manifest_rechecks_replaced_dependency_bytes(bundle, path):
+    bundle.manifest()
+    bundle.content[path] = b"Changed reference"
+    with pytest.raises(RevisionConflict, match="Dependency bytes disagree"):
+        bundle.manifest()
+
+
+@pytest.mark.parametrize(
+    ("path", "message"),
+    [("SKILL.md", "Missing entrypoint"), ("support/a.txt", "Missing required resource")],
+)
+def test_manifest_rechecks_deleted_resources(bundle, path, message):
+    bundle.manifest()
+    del bundle.content[path]
+    with pytest.raises(ValueError, match=message):
+        bundle.manifest()
+
+
+@pytest.mark.parametrize("path", ["../outside", "/absolute", "a\\b", "C:/outside", "a/../b"])
+def test_manifest_rechecks_new_unsafe_content_paths(bundle, path):
+    bundle.manifest()
+    bundle.content[path] = b"Unsafe resource"
+    with pytest.raises(ValueError, match="Unsafe bundle path"):
+        bundle.manifest()
+
+
+@pytest.mark.parametrize(
+    ("required", "error", "message"),
+    [
+        ("missing.txt", ValueError, "Missing required resource"),
+        ("../outside", ValueError, "Unsafe bundle path"),
+        (1, TypeError, None),
+        (["SKILL.md"], TypeError, None),
+    ],
+)
+def test_manifest_rechecks_changed_required_resources(bundle, required, error, message):
+    bundle.manifest()
+    bundle.metadata["required_resources"].append(required)
+    with pytest.raises(error, match=message):
+        bundle.manifest()
+
+
+@pytest.mark.parametrize(
+    ("path", "revision", "error", "message"),
+    [
+        ("support/a.txt", "sha256:" + "0" * 64, RevisionConflict, "bytes disagree"),
+        ("support/a.txt", "unresolved", ValueError, "resolved content revision"),
+        ("absent", "sha256:" + "0" * 64, ValueError, "Missing dependency"),
+        ("../outside", "sha256:" + "0" * 64, ValueError, "Unsafe bundle path"),
+    ],
+)
+def test_manifest_rechecks_changed_dependencies(bundle, path, revision, error, message):
+    bundle.manifest()
+    bundle.dependencies[path] = revision
+    with pytest.raises(error, match=message):
+        bundle.manifest()
diff --git a/tests/test_evidence_inventory.py b/tests/test_evidence_inventory.py
index 4e87fcb2..f726313c 100644
--- a/tests/test_evidence_inventory.py
+++ b/tests/test_evidence_inventory.py
@@ -67,6 +67,30 @@ def test_changed_hashed_file_fails(bundle):
     assert "SHA-256 mismatch" in str(validate_inventory(repo)["errors"])
 
 
+def test_nested_bundle_hashes_are_verified(bundle):
+    repo, root, inventory = bundle
+    hashes = json.loads((root / "plan/freeze.json").read_text(encoding="utf-8"))
+    write_json(root / "plan/freeze.json", {"schema_version": 1, "files": hashes})
+    inventory["bundles"]["synthetic"]["manifests"][0]["field"] = "files"
+    write_json(repo / "tools/evidence_inventory.json", inventory)
+    result = validate_inventory(repo)
+    assert result["errors"] == []
+    assert result["checked_hashes"] == 1
+    (root / "validator/check.py").write_bytes(b"# changed nested payload\n")
+    assert "SHA-256 mismatch" in str(validate_inventory(repo)["errors"])
+
+
+@pytest.mark.parametrize("manifest", [
+    {"other": {}}, {"files": []}, {"files": None}, {"files": "not a mapping"}, {"files": {}},
+])
+def test_nested_bundle_field_must_exist_and_be_nonempty_mapping(bundle, manifest):
+    repo, root, inventory = bundle
+    write_json(root / "plan/freeze.json", manifest)
+    inventory["bundles"]["synthetic"]["manifests"][0]["field"] = "files"
+    write_json(repo / "tools/evidence_inventory.json", inventory)
+    assert validate_inventory(repo)["errors"]
+
+
 @pytest.mark.parametrize("path", ["../outside", "/outside", "C:/outside", "..\\outside", "a/../b"])
 def test_unsafe_manifest_path_fails(bundle, path):
     repo, root, _ = bundle
@@ -145,6 +169,6 @@ def test_checked_in_inventory():
     repo = Path(__file__).resolve().parents[1]
     result = validate_inventory(repo)
     assert result["errors"] == []
-    assert len(result["bundles"]) == 8
-    assert result["checked_hashes"] == 112
+    assert len(result["bundles"]) == 9
+    assert result["checked_hashes"] == 120
     assert result["unavailable"]  # Historical raw artifacts are not in this checkout.
diff --git a/tools/evidence_inventory.json b/tools/evidence_inventory.json
index d1218689..60e1617d 100644
--- a/tools/evidence_inventory.json
+++ b/tools/evidence_inventory.json
@@ -7,6 +7,13 @@
       "manifests": [{"path": "plan/freeze.json", "schema": "bundle-sha256"}],
       "external": ["Runtime evidence referenced by results/status.json is stored separately."]
     },
+    "knowledge-retrieval-20261027": {
+      "kind": "campaign",
+      "manifests": [{
+        "path": "plan/freeze.json", "schema": "bundle-sha256", "field": "files"
+      }],
+      "external": ["Private local pytest logs are not copied; results/validation-summary.md retains the failed suite, baseline failure reproductions and package-check qualification."]
+    },
     "modular-refactor-20261013": {
       "kind": "campaign",
       "manifests": [{
diff --git a/tools/validate_evidence.py b/tools/validate_evidence.py
index 067a9300..3af5a70d 100644
--- a/tools/validate_evidence.py
+++ b/tools/validate_evidence.py
@@ -57,7 +57,9 @@ def validate_manifest(repo, bundle, entry, unavailable):
     schema = entry["schema"]
     checked = 0
     if schema in ("bundle-sha256", "historical-sha256"):
-        if not data:
+        if schema == "bundle-sha256" and "field" in entry:
+            data = data[entry["field"]]
+        if not isinstance(data, dict) or not data:
             raise ValueError("empty hash manifest")
         mapping = entry.get("files", {})
         if schema == "historical-sha256" and set(mapping) != set(data):