Repository navigation
knowledge: add Husky document-only research candidate - #191
Merged
Merged
Conversation
Signed-off-by: WeiQing Chen <530634352@qq.com> Agent-assisted-by: Codex / approved_arex_author (gpt-6.1-sol, xhigh) Agent-assisted-by: Codex / runtime_memory_author (gpt-6.1-sol, xhigh)
Signed-off-by: WeiQing Chen <530634352@qq.com> Agent-assisted-by: Codex / approved_arex_author (gpt-6.1-sol, xhigh)
Skill evaluationPLANNED: execution is not authorized by knowing this hash. Selected comparisons: 0 · Estimated task calls: 0. Applicable only while these revisions and inputs match the current PR. Evidence and logs · Retention: 30 days. Pipeline evidence only; offline fixed outputs do not establish skill efficacy. Plan bindings, budget and frozen examPlan SHA-256: Model/settings:
Uncovered / unsupported paths (106)
|
Signed-off-by: WeiQing Chen <530634352@qq.com> Agent-assisted-by: Codex / approved_arex_author (gpt-6.1-sol, xhigh)
david6666666
marked this pull request as ready for review
October 27, 2026 11:14
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
Preserve existing main and admitted runtime graphs. Resolve only discovery/navigation/cardinalities; document-only source and all evidence boundaries unchanged. Signed-off-by: WeiQing Chen <530634352@qq.com> Agent-assisted-by: Codex / approved_arex_author (gpt-6.1-sol, xhigh)
Signed-off-by: WeiQing Chen <530634352@qq.com> Agent-assisted-by: Codex / approved_arex_author (gpt-6.1-sol, xhigh)
Keep canonical main removals without resurrecting audit/evidence assets. Candidate runtime, registry, catalog and source snapshot remain byte-identical. Signed-off-by: WeiQing Chen <530634352@qq.com> Agent-assisted-by: Codex / approved_arex_author (gpt-6.1-sol, xhigh)
5 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Governing issue
Refs #189
User explicitly accepted a document-only Husky adaptation. Final informed document-only candidate admission, sequential merge and issue closure were explicitly authorized by the user on2026-10-27T12:08:19Z; candidate/backend-unverified, unknown source/licensing and no-reproduction limits remain. Preserve merge order DwarfStar → Splash → Lily → Husky; Strata skipped. Refresh this independent branch against resulting main/catalog before integration.
Delivered
One coherent document-research procedure: bounded attributed official-page summary, original falsifiable verification/ablation questions, availability/provenance ledger, one original offline schema auditor and five assertion-bearing cases. Registry/catalog, maintained matrices and exact packaging counts update together.
Source: https://husky.underdog.ai/ (displayed2026-09-20, read2026-10-27), unversioned. No canonical source repository, implementation revision, source legal license or retained raw receipts identified. Direct raw HTTP403 was not bypassed; no HTML hash, repository ID, install/API or native behavior is invented. No publisher source code or full benchmark table copied.
AREX portable instructions at ac3fe1afa80fb9a09775ecfb2b6cc3ba850a2db6 informed this authorized document adaptation. Not complete repo-to-skill or native verification, not a DisCo CLI/model API run or automatic import. Missing repository/environment/native stages remain unavailable.
Actual verification and limits
-Ruff,3 focused document/helper tests and20 combined graph/index/library tests PASS. Actual AREX recursive license validation:1 SKILL.md valid with NO_LICENSE unavailable warning; repository resolver query not attempted because canonical identity is unavailable. Publisher license remains UNKNOWN.
-Helper:2 positive cards,8 negative schema mutations, valid/invalid CLI and help checks. Even a fully populated or forged runtime_run card remains qualification=not_adjudicated; no artifact bytes or execution is authenticated.
-Initial clean wheel and CI installed checks exposed an unrelated synthetic roofline expectation mistakenly changed60→61 during cardinality maintenance. Restored exactly60 without changing registry61/records109. Original failure remains recorded. Corrected noneditable installed smoke outside checkout passes109 records; complete pytest exited0 (only non-collected installed-script delta during run, verified separately) and full ruff PASS. Final ready-for-review CI run40494676535 is terminal SUCCESS (all4 matrix jobs); exact-head knowledge-model-policy SUCCESS.
-Independent read-only review bound to b66732f/tree0f8e970 found no unresolved blocker for the limited candidate.20 focused tests, index61, license gate and installed -I smoke109 PASS; all8 runtime files match wheel/installed bytes. Fresh forged-card case correctly refuses schema→PASS, but omitted some ancillary license/admission boundary restatements; this is not blanket efficacy evidence. Attributable external report: husky-independent-review.md, not a GitHub APPROVE.
-No implementation/model/native output/state/quality/performance/service or behavioral skill efficacy was reproduced. Author reports, proposed mechanisms, negative observations, excluded repeats and native evidence remain distinct.
-No GPU, weights, paid provider, host/toolchain change or app/account action.
Artifacts:
tests/fixtures/repo_to_skill/husky/. Final candidate admission is limited to the explicitly approved document-only scope; no campaign PASS, native promotion or GitHub approval is claimed.Knowledge authorship
Sequential integration
DwarfStar, Splash and Lily merged first. Current assessed main5c6a8f6af7ecd8fa71cedd31889820eb1546aa50 independently removes the retrieval audit/evidence campaign; those removals and its inventory adjustments are incorporated without resurrection. Husky runtime/fixtures/test and all prior admitted graphs remain unchanged. Registry, catalog and every indexed input are byte-identical to reviewed84e581bd:64 roots/94 skill documents/112 records. The existing real sourcebb81ef2fec1b79d71eba1026ab1466117c3262cc remains valid; generator --check confirms64 current records, with no unnecessary catalog rewrite.29 prior integration-focused tests and27 refreshed evidence-inventory/graph/index tests pass, as does ruff. New exact-head Ready CI and tiny base-removal delta review are required before merge; no backend/efficacy promotion.