@event4u/agent-config 16.5.0 → 17.0.0

package/src/config/gate-coverage.yml

@@ -52,15 +52,23 @@
 #                of them.
 #   canary       optional mutation recipe (Phase 7) — see CANARY below
 #
-# COVERAGE — THE HONEST DENOMINATOR (re-measured 2026-10-01)
+# COVERAGE — THE HONEST DENOMINATOR (re-measured 2026-10-22)
 # ----------------------------------------------------------
-# 347 `lint_*` / `check_*` / `audit_*` scripts live under `src/scripts/`
+# 361 `lint_*` / `check_*` / `audit_*` scripts live under `src/scripts/`
 # (the count is computed, not asserted — `check_gate_coverage.count_gate_scripts`,
 # re-run on the tree carrying these rows.)
-# **109 rows are listed below**, so the manifest covers 109/347 = 31.4 % of the
+# **118 rows are listed below**, so the manifest covers 118/361 = 32.7 % of the
 # population, and is complete for every gate that can carry a floor. The row
 # count is `grep -c '^  - id:'`, recomputed here rather than incremented.
 #
+# RECOUNTED 2026-10-12, on the change that adds `check_skill_body_budget`. The
+# header read 347 scripts against a measured 354 and 109 rows against 112
+# BEFORE this change — drift of 7 and 3, under the tolerance of 15 in
+# `check_gate_coverage.test.ts` § states an honest denominator, so nothing was
+# red and the next seven arrivals would not have redded it either. Both figures
+# are recomputed on this tree rather than incremented, which is the standing
+# instruction two blocks below and the reason this one exists at all.
+#
 # RECOUNTED 2026-09-29 ON A MERGE, and this is the hazard the two blocks below
 # predicted arriving in the form they predicted. Two branches each recounted
 # CORRECTLY for their own tree — 341/103 here, 341/102-then-103 there — and
@@ -655,6 +663,24 @@ gates:
         denominator by one while the numerator stands still, so the ratio falls
         below seed and the ratchet must red. A gate that stayed green here would
         be a count ratchet wearing a ratio's name.
+  - id: check_skill_body_budget
+    argv: []
+    ci_invocation: .github/workflows/rule-backstops.yml
+    min_scanned: 250
+    corpus: >
+      every `src/skills/<name>/SKILL.md` — 300 at baseline. The floor sits at
+      250 because the skill set moves by ones and a reading two fifths below
+      baseline means the walk lost `src/skills/`, not that a hundred skills
+      were deleted in one change.
+    status: enforced
+    no_canary_reason: >-
+      The only violation this gate can express is a body that GREW past a
+      ceiling already recorded for it, which is an in-place edit of a tracked
+      file. A create-only canary cannot plant it: a new skill has no recorded
+      ceiling and is deliberately not a finding, so the plant would run the
+      gate green and certify nothing. The red path is proved instead by
+      `tests/scripts/check_skill_body_budget.test.ts`, which grows a real body
+      from the live corpus by one byte and asserts exactly one finding.
   - id: check_skill_admissions
     argv: []
     min_scanned: 0
@@ -1523,6 +1549,11 @@ gates:
   - id: check_static_parity
     argv: ["--quiet"]
     min_scanned: 20
+    # Asserted in tests.yml collector-lifecycle from the run that job already
+    # makes; its paths filter covers src/** and tests/**, i.e. every change that
+    # can move this gate's scan count. The census re-runs it here only if that
+    # assertion disappears.
+    floor_asserted_in: .github/workflows/tests.yml
     corpus: "test files that spell the dispatcher module name, discovered by grep over tests/ and src/. The count is deliberately not written here — it moves with the suite, and four different figures for it appeared across the rounds of this change (R2 round-5 finding 12). The floor is 20, not 1: this gate's silent-failure shape is a discovery COLLAPSE — a grep that stops matching would run two suites of one file each and report them identical, which is the 'gate that scans nothing exits green' case in its most convincing disguise."
     status: enforced
     no_canary_reason: >-
@@ -2760,6 +2791,36 @@ gates:
       the count stays at 24 either side of this change. The self-test is
       therefore adopted on its merits rather than to clear a ratchet.
 
+  - id: check_step_receipts
+    argv: ["--base", "origin/main"]
+    min_scanned: 0
+    corpus: >
+      the roadmap files under `agents/roadmaps/` a change touches, measured
+      against the base ref — a DIFF, with renames followed. The floor is 0
+      deliberately: most changes tick no roadmap step, and git answers for its
+      own range, so there is no committed path list that can go stale. An
+      unreadable base exits 2 rather than reporting zero.
+
+      SELF-TEST ADOPTED: 6 cases, 3 rejecting, against the real CLI over a
+      throwaway repository — a step ticked without a receipt, a new roadmap
+      arriving ticked, and a hand-written receipt claiming `active` that the
+      diff does not allow must RED; a receipted tick, a reworded older tick and
+      a diff with no roadmap must stay GREEN. Sensitivity observed: dropping the
+      receipt condition turns the first two rejecting cases and two unit cases
+      red; allowing every lowered claim turns the third red.
+      The self-test passes `--enforce`; the CI call does not, so while the
+      depth table is in shadow mode a finding is reported and exits 0, and from
+      the flip on it exits 1 — blocking exactly when a lowered depth exists.
+    status: enforced
+    ci_invocation: .github/workflows/rule-backstops.yml
+    no_canary_reason: >-
+      A create-only canary cannot express this gate's violation. The corpus is a
+      DIFF against the base ref, so a planted file is in scope only once it is
+      committed, and the violation is a CHANGE of one line from unticked to
+      ticked, which an added file without a base version cannot represent
+      except as a new roadmap. Discrimination is proven by the gate's own
+      `--self-test` (6 cases, 3 rejecting), which builds these shapes.
+
   - id: check_agent_artifact_location
     argv: []
     min_scanned: 3000
@@ -3299,3 +3360,88 @@ gates:
       added a cross-pack link got no red check from anything. The step in
       consistency.yml runs it with this argv; `--self-test` proves the CLI still
       rejects an undeclared link, an unresolved pack and an empty corpus.
+
+  - id: lint_ticked_unmet_criteria
+    argv: []
+    min_scanned: 700
+    corpus: "every `*.md` under `agents/roadmaps/`, archive and `later/` included — 1072 on 2026-10-13. The floor is 700 because roadmaps are archived, not deleted, so the population only grows except when a sweep moves a tree; a resolver that stopped walking `archive/` would drop by hundreds at once. The count is files WALKED, not criteria read: a roadmap with no acceptance criteria is still inspected and still counted."
+    status: enforced
+    ci_invocation: .github/workflows/rule-backstops.yml
+    canary:
+      class: "a ticked acceptance criterion whose own text says the phrase"
+      path: agents/roadmaps/canary-ticked-unmet.md
+      content: |
+        Canary roadmap planted by check_gate_coverage --canary.
+
+        Acceptance Criteria
+
+        - [x] AC-1 — One half lands. The other half is NOT met.
+    note: >-
+      road-to-what-was-built-and-never-ran 1.3. Until this row the lint ran
+      nowhere: no task and no workflow named it, so the four hits it carried on
+      the trunk had never red anything. A real canary is possible here because
+      this gate is not a ratchet — any hit exits 1 at any headroom, so one
+      planted criterion reddens it whatever else the tree contains.
+
+  - id: lint_blocker_citations
+    argv: []
+    min_scanned: 100
+    corpus: "every `*.md` under `agents/roadmaps/` that the blocker reader yields — 159 on 2026-10-13, i.e. the roadmap files carrying a `## Blockers` section rather than the whole tree. The floor is 100 because an unresolved blocker is parked rather than deleted and the set turns over slowly, while a reader that stopped recognising the section heading would collapse to single digits."
+    status: enforced
+    ci_invocation: .github/workflows/rule-backstops.yml
+    no_canary_reason: >-
+      A create-only canary cannot discriminate on this gate, for the same reason
+      `lint_pack_boundaries` records: it is a ratchet, so it fails only when the
+      live violation count rises ABOVE the recorded baseline in
+      `src/config/gate-violation-baselines.json`, and any headroom there lets a
+      planted citation read green — the ledger would call a live gate dead. The
+      headroom this row first recorded (8 against 10) was closed on 2026-10-22,
+      when the baseline was lowered to 8 on the Rule backstops job's own
+      reading. A canary would still discriminate only while the baseline sits
+      exactly at the live count, which the next fixed citation changes.
+    note: >-
+      road-to-what-was-built-and-never-ran 1.3. Same history as the row above —
+      written, tested, and named by no caller until this change.
+
+  - id: check_credential_spelling
+    argv: []
+    min_scanned: 1200
+    corpus: "every non-test `*.ts` / `*.mts` file under `src/scripts` and `src/cli`, read from the filesystem so an unstaged file is seen — 1485 on 2026-10-18. The floor is 1200: the two roots lose files only to deliberate deletions, while a walk that stopped descending (a moved root, a renamed directory) collapses to a handful."
+    status: enforced
+    ci_invocation: .github/workflows/rule-backstops.yml
+    canary:
+      class: contract-violation
+      path: src/scripts/zz-canary-credential-spelling.ts
+      content: |
+        // Canary planted by check_gate_coverage --canary; never ships.
+        export const RE = /ghp_[A-Za-z0-9]{36}/;
+    note: >-
+      road-to-one-detector-and-the-senders-we-own 2.2. The rule table is the only
+      place a raw-key prefix is spelled; this gate fails a second spelling. Since
+      step 2.7 it also fails a prefix spelled through a character class or an
+      alternation, and a keyword-assignment shape (a table keyword beside the
+      `[:=]` operator class on one line); a shape assembled across lines is not
+      seen.
+
+  - id: check_egress_mutations
+    argv: []
+    min_scanned: 4
+    corpus: "every mutation recipe under `src/config/egress-mutations/` — 4 on 2026-10-22 (class precedence, a credential token that restores, a raw value in the count record, the IBAN checksum disabled). The floor is the four."
+    status: enforced
+    ci_invocation: .github/workflows/rule-backstops.yml
+    canary:
+      class: contract-violation
+      path: src/config/egress-mutations/zz-canary-survivor.json
+      content: |
+        {
+            "mutation": "canary-survivor",
+            "module": "src/scripts/_lib/egress_token.ts",
+            "find": "export const MAX_TOKEN_VARIANT = 99;",
+            "replace": "export const MAX_TOKEN_VARIANT = 99;",
+            "killed_by": "credential-token-never-restores"
+        }
+    note: >-
+      road-to-one-detector-and-the-senders-we-own 2.6. Each recipe is applied to
+      a scratch copy of the three engine modules, never in place; the invariant
+      it names must hold on the unmutated copy and fail on the mutated one. The
+      canary plants a recipe that changes nothing, which survives.

package/src/config/gate-reachability-exemptions.json

@@ -3,21 +3,19 @@
     "_registered_at": "2026-08-30",
     "_measured": "22 unreachable at the start of Phase 2, 14 after it. The 8 that left were wired one per commit, each run locally first.",
     "exempt": {
-        "ci-strict": "An ENTRY POINT, not a gate — the stricter sibling of `ci` itself, so wiring it into `ci` would be a cycle. Would run if strict became the default, which is a policy change rather than a wiring one.",
-        "check-augment-budget": "Emits a MEASUREMENT, not a verdict; a gate that cannot fail is a report. Would run once a pre-registered threshold exists to fail on.",
+        "ci-strict": "An ENTRY POINT, not a gate \u2014 the stricter sibling of `ci` itself, so wiring it into `ci` would be a cycle. Would run if strict became the default, which is a policy change rather than a wiring one. REMOVED 2026-10-14 and restored the same day: `check_gate_reachability` had read the script path out of this target's own comment (\"Guarded by `check-ci-strict-superset`\") and called it reachable once that guard was wired, so the row looked stale when the reading was wrong. The reading is fixed; the exemption stands.",
     "check-platform-anchor": "THE PERMISSION DOES NOT EXIST, which is a different reason from every other row here — the others are unwired by judgement, this one by platform capability. It reads `repos/{owner}/{repo}/rulesets` to check that the forge actually requires what the ratification mechanism claims, and that endpoint needs the repository `administration` scope. A workflow `GITHUB_TOKEN` cannot be granted it: `administration: read` was tried in `rule-backstops.yml` on 2026-09-10 and actionlint refused it as an \"unknown permission scope\", listing the sixteen that do exist. Wiring it with an unauthorised token is worse than leaving it manual, because the gate fails closed by design — an unreadable platform is UNVERIFIABLE, which is a failure — so it would red every kernel and governance PR for a reason unrelated to the platform's actual state. The three escapes were each rejected on the reasoning of the council that commissioned the gate: `continue-on-error` makes it advisory, skipping when unauthenticated makes it fail open, and `Sync + Generate Tools Consistency` would freeze merges repository-wide on any red since that job's context is the single required check. The script itself IS reachable pre-push via the `preflight` list in `taskfiles/ci-fast.yml`; this target is the on-demand `task check-platform-anchor` entry point beside it. WOULD RUN given a PAT in a repository secret carrying `administration: read` — a human action, recorded as the open half of blocker `ratification-platform-anchor`.",
         "check-corpus": "Requires council configuration, which is user-global (ADR-104) and absent in CI by construction. Would run against a fixture corpus that needs no configured council.",
         "check-media-deps": "Requires external media tooling not installed on runners. Would run with those dependencies in the runner image, or behind a skip-with-reason path.",
         "check-trunk-drift": "ADVISORY BY ITS OWN DOCSTRING — 'run it in the /create-pr pre-flight or wire into CI per your branch-protection policy'. Its red state is 'this branch is behind trunk', which is the normal state of an open PR, so wiring it would fail correct work. RECLASSIFIED from should-run during Phase 2 after reading what it asserts. Would run under a branch-protection policy that requires an up-to-date base.",
         "dev:standing-rule-delivery": "`dev:` namespace — an inspection aid for a maintainer mid-change. Would run if promoted out of `dev:`, which is a scope decision.",
         "lint-agents-layout-strict": "`--strict` over a corpus whose non-strict form is already reachable; strict carries a known backlog. Would run once that backlog is cleared and strict becomes the default.",
-        "lint-bench-ab": "Bench namespace, spend-bearing. Would run given a cached or dry arm that costs nothing per invocation.",
         "lint-explain-trace": "NOT A GATE — a validator that takes a JSON document, and a bare invocation exits non-zero with 'pass a JSON file path or --stdin'. RECLASSIFIED from should-run during Phase 2, which is exactly what 2.2's run-it-locally-first rule exists to catch: wiring it would have made CI permanently red. Would run wired to a specific trace fixture.",
         "lint-originality-shingles": "`--top 10` is a ranked REPORT, not a pass/fail. Would run given a ratcheted threshold, like the other originality gates carry.",
-        "mcp:glama-drift-check": "Queries an external registry — network-dependent and flaky by construction. Would run against a recorded snapshot instead of a live query.",
         "migration-verify": "Verifies a physical move during a migration; there is no migration in flight. Would run during one.",
         "value": "Renders a dashboard; the render is the product, not a check. Nothing would make it a gate.",
         "push-ready": "NOT A GATE — a developer-invoked helper that MUTATES the tree: it merges the base set in, regenerates derived files, and hands the push back to the contributor. It is gate-shaped only because step 5 calls `check_branch_freshness`, which IS wired to CI on its own. Wiring this target would make CI perform merges, which is the exact thing `road-to-cascading-base-integration` 2.2 forbids the pre-push hook from doing one layer down. Nothing would make it a gate; the checks inside it are wired separately.",
-        "value:lint": "Lints that dashboard, which is generated and regenerated on demand. Would run if the dashboard became a committed artefact a drift could break."
+        "value:lint": "Lints that dashboard, which is generated and regenerated on demand. Would run if the dashboard became a committed artefact a drift could break.",
+        "voice-lexicon": "MANUAL BY THE DESIGN THAT CREATED IT, not by omission. `road-to-a-brand-the-tree-can-check` 5.6 specifies \u201cnot in default CI\u201d and the design it inherits says why: a voice/style gate in a team's default pipeline is a social bug \u2014 it makes one person's wording preferences a merge condition for everyone. Two further reasons the wiring would be theatre rather than enforcement: its brand findings are `warning` and it exits 0, so a CI step could never fail on them; and its lexicon is a per-project `agents/context-spine/brand-voice.md` slot, which this repository does not carry, so in this repository's own CI it would report SKIPPED on every run \u2014 a gate that scans nothing and exits green, which is the shape the reachability work exists to stop rather than to create. WOULD RUN when BOTH halves hold: this repository registers its own brand-voice slot (so the lexicon is non-empty and `assertScanned` has something to assert), AND the project decides its own persisted prose is in scope \u2014 at which point the honest wiring is a ratcheted finding count, not a bare pass/fail, because the findings are advisory by construction."
     }
 }