0.24.3 — the budget stops being estimated, and two skills stop naming a file they do not ship

.claude-plugin/marketplace.json

@@ -12,7 +12,7 @@
       "displayName": "Agent Stack",
       "source": "./plugins/agent-stack",
       "description": "Four skills: agent-orchestrator — tool-calling loops, pipelines with checkpoints, provider routing with fallback, memory architecture, plus the wallet side of reselling LLM access; agent-evals — run/trace/thread evals, LLM judges, and fixtures grown from production; agent-interop — MCP servers and clients, A2A agent cards, the MCP Registry, and gateways; agent-harness — system prompts, tool shaping, workflow-vs-agent, and auditing an agent system.",
-      "version": "0.24.2",
+      "version": "0.24.3",
       "author": {
         "name": "ssheleg",
         "url": "https://x.com/sshlg93"

.github/workflows/validate.yml

@@ -31,6 +31,13 @@ jobs:
     steps:
       - uses: actions/checkout@v4
 
+      # The body budget is MEASURED, never estimated: `test/validate.py` refuses to issue
+      # a verdict from chars/3.9 and discloses instead, so without this step the budget
+      # check stops gating quietly. The skill-audit job installs the same package for the
+      # same reason (make-skill v0.28.0).
+      - name: Install a real tokenizer (the budget is MEASURED, never estimated)
+        run: python3 -m pip install --quiet tiktoken
+
       - name: Structural validator
         run: python3 test/validate.py
 

CHANGELOG.md

@@ -1,3 +1,26 @@
+## 0.24.3 — the budget stops being estimated, and two skills stop naming a file they do not ship
+
+Hygiene from the 2026-09-13 family audit (HK-11).
+
+- **The body budget is MEASURED.** This gate divided by a calibrated 3.9 chars/token; the
+  calibration drifted. Measured 2026-09-14 against the authority's tokenizer:
+  `agent-evals` estimated **~4961** and measured **4374**; `agent-orchestrator` estimated
+  **~4762** and measured **4309** — two files reported past the working limit with 300+
+  tokens of real headroom. make-skill v0.28.0 closed the same defect in the family
+  auditor by measuring. With no tokenizer installed the check now **discloses** rather
+  than issuing a verdict from the wrong instrument.
+- **`agent-harness` and `agent-interop` named `test/validate.py`** — a repository file
+  that does not ship inside a skill, so it resolves at the repo root and nowhere an
+  installed reader stands. Both say "this pack's repository validator" now.
+- **`agent-harness` declares `compatibility`**: it ships `scripts/audit_agent.py`, and a
+  skill that needs an interpreter says so in front matter.
+- **`plugin.json`'s homepage** matches `package.json` and the forge
+  (`skills.sshlg.me/skills/agent-stack/`) instead of the GitHub URL.
+- **The README names what loads on demand** — 27 reference files, counted per skill from
+  the tree, not stated — and both invocation forms (`/agent-stack:<skill>` and
+  `/<skill>`), because which one resolves depends on how the pack was installed. The
+  installer line said "both skills" over a pack of four.
+
 ## v0.24.2 — the ledger section names the version it ships
 
 The previous patch wrote the ledger section BEFORE bumping the version, so the

README.md

@@ -160,7 +160,7 @@ waterfall, and model-routing precedence.
 /plugin install agent-stack@agent-stack
 ```
 
-**npm installer** — copies both skills into `~/.claude/skills/`:
+**npm installer** — copies all four skills into `~/.claude/skills/`:
 
 ```bash
 npx @ssheleg/agent-stack
@@ -180,6 +180,26 @@ npx --yes sshlg-skills@latest update
 
 Restart your agent afterwards — skills load at session start.
 
+**Reaching a skill by name.** Claude Code routes on the description, and both
+invocation forms work: `/agent-stack:agent-orchestrator` names the plugin's copy
+explicitly, `/agent-orchestrator` the plain one. Which resolves depends on how this
+pack was installed, so both are written here rather than one promised.
+
+### What loads on demand
+
+Twenty-seven reference files ship with the four skills, each with a load trigger in
+its skill body. By skill:
+
+| Skill | References |
+|---|---|
+| `agent-orchestrator` | 12 — the tool-calling loop, memory and retrieval, context budgets, provider routing, the work graph, the resale wallet and the saga across database and provider API |
+| `agent-harness` | 7, plus `scripts/audit_agent.py` |
+| `agent-interop` | 6 — one per protocol surface, each pinning the revision it was read at |
+| `agent-evals` | 2 — `otel-genai.md` (the semantic conventions an eval reads) and `statistics.md` (when a difference is a difference) |
+
+A file with no load trigger in a body is a defect, not a bonus: the skill audit
+refuses one.
+
 ---
 
 ## When it triggers

SKILL-CARD.md

@@ -5,7 +5,7 @@
 | Field | Value |
 |---|---|
 | Pack | `agent-stack` |
-| Version | `0.24.2` |
+| Version | `0.24.3` |
 | Skills | `agent-orchestrator`, `agent-evals`, `agent-interop`, `agent-harness` |
 | License | MIT |
 | Source | https://github.com/ssheleg/agent-stack |

docs/evidence/verification.md

@@ -10,6 +10,18 @@ This file exists because its absence read as zero exposure. `sshlg-skills` board
 
 ---
 
+## Shipped state — v0.24.3 (2026-09-14)
+
+Hygiene from the 2026-09-13 family audit (HK-11).
+
+| REQ | What ships | How it was confirmed | Confirmed |
+|---|---|---|---|
+| HY-measure | The body budget is measured, never estimated | `_measure_body_tokens` uses tiktoken where installed and discloses where it is not. The drift it replaces: agent-evals ~4961 estimated / 4374 measured, agent-orchestrator ~4762 / 4309 | **observed** — two files were being reported past the working limit with 300+ tokens of headroom |
+| HY-resolve | No skill body names a file that does not ship inside it | `agent-harness/SKILL.md:122` and `agent-interop/SKILL.md:44` said `test/validate.py`, a repository file; both now say "this pack's repository validator" | **observed** |
+| HY-compat | A skill that needs an interpreter declares it | `agent-harness` ships `scripts/audit_agent.py` and now carries `compatibility` in front matter; `audit_skill.py --house` 0 GAP, 19 PASS | **observed** |
+| HY-shelf | The README names what loads on demand, counted from the tree | 27 reference files, per skill: orchestrator 12, harness 7, interop 6, evals 2 — the first draft of this table stated 12/4/7/4 from memory and was corrected against `ls` | **observed** — the counted number and the remembered one disagreed on three of four rows |
+| Gate | The whole suite on this tree | `npm test` EXIT=0; `OK: agent-stack structurally valid (15 checks …, v0.24.3)` | **observed** |
+
 ## Shipped state — v0.24.2 (2026-09-10)
 
 Sherlock external-v3 (24 findings), each carrying its own executable regression.

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ssheleg/agent-stack",
-  "version": "0.24.2",
+  "version": "0.24.3",
   "scripts": {
     "test": "python3 test/validate.py && python3 test/plant_guard_test.py && node test/installer_test.js && npm run test:audit",
     "test:audit": "for t in test/audit_regressions/*.py; do python3 \"$t\" || exit 1; done"

plugins/agent-stack/.claude-plugin/plugin.json

@@ -3,12 +3,12 @@
   "name": "agent-stack",
   "displayName": "Agent Stack",
   "description": "Four skills: agent-orchestrator — tool-calling loops, pipelines with checkpoints, provider routing with fallback, memory architecture, plus the wallet side of reselling LLM access; agent-evals — run/trace/thread evals, LLM judges, and fixtures grown from production; agent-interop — MCP servers and clients, A2A agent cards, the MCP Registry, and gateways; agent-harness — system prompts, tool shaping, workflow-vs-agent, and auditing an agent system.",
-  "version": "0.24.2",
+  "version": "0.24.3",
   "author": {
     "name": "ssheleg",
     "url": "https://x.com/sshlg93"
   },
-  "homepage": "https://github.com/ssheleg/agent-stack",
+  "homepage": "https://skills.sshlg.me/skills/agent-stack/",
   "repository": "https://github.com/ssheleg/agent-stack",
   "license": "MIT",
   "keywords": [

plugins/agent-stack/skills/agent-harness/SKILL.md

@@ -14,6 +14,9 @@ description: >-
   "аудит агента", "встроить агента". Not for the loop's plumbing, its evals, or its
   protocols — those are siblings.
 license: MIT
+compatibility: >-
+  scripts/audit_agent.py runs with python3 (standard library only); no network,
+  no package install. Everything else in this skill is prose and needs nothing.
 ---
 
 # Agent harness — what the agent is told, and how to audit what someone else told theirs
@@ -119,7 +122,7 @@ because a decision table with two homes is one that will disagree with itself.
 ## References
 
 Each opens with its own **Load this when** line and a revision stamp — this material moves,
-and `test/validate.py` fails the build on a reference that does not say when it was read.
+and this pack's repository validator fails the build on a reference that does not say when it was read.
 
 | File | Read it when |
 |---|---|

plugins/agent-stack/skills/agent-interop/SKILL.md

@@ -41,7 +41,7 @@ against the live specification rather than recalled:
   still introduced `binds` as a core concept as of 2026-08-13.
 
 So: **every reference in this skill opens with a `**Spec pinned:**` line**, and
-`test/validate.py` fails the build without one. That is a mechanical check, not an
+This pack's repository validator fails the build without one. That is a mechanical check, not an
 aspiration — the class of error it prevents is the one where prose reads as current
 because nothing on the page says otherwise.
 

test/validate.py

@@ -93,6 +93,23 @@ def front_matter(path):
     return m.group(1), text
 
 
+def _measure_body_tokens(body):
+    """(count, how) — a real tokenizer where one is installed, the estimate otherwise.
+
+    The authority is the family auditor's `cl100k_base`. The divisor below was calibrated
+    against it once and drifted: on 2026-09-14 it read agent-evals at ~4961 and
+    agent-orchestrator at ~4762 where the tokenizer read 4374 and 4309.
+    """
+    try:
+        import tiktoken
+    except Exception:                                    # noqa: BLE001 - optional
+        return int(len(body) / 3.9), "estimate"
+    try:
+        return len(tiktoken.get_encoding("cl100k_base").encode(body)), "tiktoken:cl100k_base"
+    except Exception:                                    # noqa: BLE001 - a broken install is not a verdict
+        return int(len(body) / 3.9), "estimate"
+
+
 def scalar(block, key):
     """Read one front-matter scalar without a YAML dependency.
 
@@ -217,12 +234,22 @@ def scalar(block, key):
     # -> 3.9), so this tracks the authority closely and slightly high. Re-derive it if the
     # auditor's tokenizer changes; do not widen it to make a failing file pass.
     body = text.split("---", 2)[2] if text.count("---") >= 2 else text
-    body_tokens = int(len(body) / 3.9)
-    if body_tokens > BODY_BUDGET_TOKENS:
-        fail(f"{name}/SKILL.md: body ~{body_tokens} tokens, past the {BODY_BUDGET_TOKENS} "
-             "budget — the answer at this point is a split, not a trim")
+    body_tokens, how = _measure_body_tokens(body)
+    if how == "estimate":
+        # A calibrated divisor is still an estimate, and the calibration expires: measured
+        # 2026-09-14, this one read agent-evals at ~4961 and agent-orchestrator at ~4762
+        # while the authority's tokenizer read 4374 and 4309 — two files reported past the
+        # working limit with 300+ tokens of real headroom. make-skill v0.28.0 closed the
+        # same defect in the family auditor by MEASURING; a verdict from the wrong
+        # instrument gets quoted as if it were one, so with no tokenizer this discloses
+        # instead of judging.
+        notes.append(f"{name}/SKILL.md: body budget NOT MEASURED (~{body_tokens} tokens by "
+                     "chars/3.9) — `pip install tiktoken` to gate it")
+    elif body_tokens > BODY_BUDGET_TOKENS:
+        fail(f"{name}/SKILL.md: body {body_tokens} tokens ({how}), past the "
+             f"{BODY_BUDGET_TOKENS} budget — the answer at this point is a split, not a trim")
     elif body_tokens > BODY_WORKING_TOKENS:
-        notes.append(f"{name}/SKILL.md: body ~{body_tokens} tokens, past the "
+        notes.append(f"{name}/SKILL.md: body {body_tokens} tokens ({how}), past the "
                      f"{BODY_WORKING_TOKENS} working limit ({BODY_BUDGET_TOKENS} budget) — "
                      "displace before the next addition")