@event4u/agent-config 14.14.0 → 14.15.0

package/.augment-plugin/marketplace.json

@@ -1,20 +1,20 @@
 {
   "name": "event4u-agent-config",
   "description": "Governed agent system — skills, rules, commands, and guidelines for AI-assisted development.",
-  "version": "14.14.0",
+  "version": "14.15.0",
   "owner": {
     "name": "event4u",
     "email": "dev@event4u.app"
   },
   "metadata": {
     "description": "A policy-driven execution system that enforces quality, consistency, and developer-like behavior for AI agents.",
-    "version": "14.14.0"
+    "version": "14.15.0"
   },
   "plugins": [
     {
       "name": "agent-config",
       "description": "Full governed agent system — rules, skills, commands, guidelines, and runtime tooling.",
-      "version": "14.14.0",
+      "version": "14.15.0",
       "source": ".",
       "category": "governance",
       "tags": [

package/.augment-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "name": "agent-config",
   "description": "A governed skill framework for AI agents — runtime execution, tooling, observability, feedback, and lifecycle management with policy-driven quality enforcement.",
-  "version": "14.14.0",
+  "version": "14.15.0",
   "author": {
     "name": "event4u",
     "email": "dev@event4u.app"

package/.claude-plugin/marketplace.json

@@ -6,7 +6,7 @@
   },
   "metadata": {
     "description": "Bootstrap shim for the event4u agent-config suite. This plugin ships ONLY the dispatcher hooks plus one install-pointer skill — ALL content (skills, rules, commands) is installed via `npx -y @event4u/agent-config init`, which writes the file projection AND registers the same hooks in ~/.claude/settings.json. DEPRECATED as a content channel; kept so the marketplace listing stays discoverable and existing installs keep working hooks.",
-    "version": "14.14.0",
+    "version": "14.15.0",
     "keywords": [
       "agent-config",
       "skills",

package/CHANGELOG.md

@@ -434,6 +434,69 @@ Entry-shape contract: [`docs/contracts/CHANGELOG-conventions.md`](docs/contracts
 > that forces a new era split (`# Era: 14.15.x`, etc.) — see
 > [`docs/contracts/CHANGELOG-conventions.md § Era splits`](docs/contracts/CHANGELOG-conventions.md).
 
+## [14.15.0](https://github.com/event4u-app/agent-config/compare/14.14.1...14.15.0) (2026-09-03)
+
+### Release highlights
+
+- **Behaviour changes:** stage-2 impact scan, so a refused merge can be answered in four words (5a3b7c5).
+- **Default changes + migration:** _none_
+- **Security and correctness:** make the derived head publishable, so a release stops halting by construction (e6f6744); ask for the curated head before the release commit, not after (e31f07a).
+- **Honest nulls:** _none_
+- **Known limitations:** _none_
+
+### Features
+
+* **authz:** stage-2 impact scan, so a refused merge can be answered in four words ([5a3b7c5](https://github.com/event4u-app/agent-config/commit/5a3b7c5064b2e92523781530cbfe6ef09bd26f35))
+* **authz:** classify the 17 destructive operations the guard could not see ([a1327b2](https://github.com/event4u-app/agent-config/commit/a1327b2767751d4ce6e680af1b8f9732faf97b88))
+* **authz:** target-bound merge grants that outlive the clock ([c286570](https://github.com/event4u-app/agent-config/commit/c286570d1e9a421a51571bb27e3a2945a54e8f8c))
+
+### Bug Fixes
+
+* **docs:** cite the head label as code, not bold — the ratchet counts bold ([3f78ead](https://github.com/event4u-app/agent-config/commit/3f78ead4fd9023ed189ccd15aac2659397072ca8))
+* **release:** make the derived head publishable, so a release stops halting by construction ([e6f6744](https://github.com/event4u-app/agent-config/commit/e6f67444c0c11873f6231fd007b70b404f9e883a))
+* **ci:** drop code-comment-discipline from the rule-enforcement baseline ([731924a](https://github.com/event4u-app/agent-config/commit/731924ae778191bafefb3bbcfcb92c2acf76b558))
+* **release:** ask for the curated head before the release commit, not after ([e31f07a](https://github.com/event4u-app/agent-config/commit/e31f07ab8bb4b25e048f1b4c78ffa32b27fd59f5))
+
+### Documentation
+
+* **adr:** ADR-252 evidence section, and the census it moves ([19e99a4](https://github.com/event4u-app/agent-config/commit/19e99a4febfc6e9f8d092ecafc8c8e23c8f1699e))
+* **adr:** ADR-252 supersedes ADR-251 — specificity replaces recency, for merge only ([ef930e0](https://github.com/event4u-app/agent-config/commit/ef930e01d450a4b78730f545a4f04eab1812eccb))
+
+Tests: 20922 (+146 since 14.14.1)
+
+## [14.14.1](https://github.com/event4u-app/agent-config/compare/14.14.0...14.14.1) (2026-09-03)
+
+### Release highlights
+
+- **Behaviour changes:** four new read-only surfaces, all additive — `agent-config hooks:effect`, which dispatches each bound concern against a scratch root and reports whether it actually fired rather than whether the manifest binds it (b0a03a7); `doctor --strict`, carrying the failing exit as a separate mode so `doctor` itself keeps its always-zero exit and no existing caller changes (517afcb, the `doctor-exit-contract` choice); and the `dev:skill-scout` and `push-ready` Taskfile targets (576ff4f, 6735abf). No public artefact was removed in this span.
+- **Default changes + migration:** _none_
+- **Security and correctness:** `skill_scout` confines a human-supplied candidate name to a single path segment and asserts resolved-path containment BEFORE it reports a scan scope, so a traversing name is refused rather than normalised and the scan-scope line can never record a root it did not scan; it also re-checks file type and size at read time, closing a post-intake symlink that was followed out of the quarantine directory (576ff4f). Three runtime advisories that were failing the `npm audit --omit=dev --audit-level=high` gate are cleared (54f0242).
+- **Honest nulls:** the deferred council-topology carrier's two resumption triggers were re-measured live and neither had moved — Phase 2 needs five independent eligible council seats and `council:status` reports two of five; Phase 3 needs a two-day spend reservation that does not exist. The carrier stays parked with all 38 obligations in place, and no third disposition record was written (1fe6fa0).
+- **Known limitations:** the README daemon-claim repair is verified only one way. `check_claims` exiting 0 proves the retired phrasing is gone; it does NOT validate the replacement wording, because `README.md:30` is unmarkered prose (e200ba2). And `check_source_size_budget`'s excess is still five figures — this span lowered it 18,245 → 18,205 by moving prose and helpers under the 1500-line cap, not by removing debt.
+
+### Bug Fixes
+
+* **deps:** clear the three runtime advisories failing the audit gate (#1822) ([54f0242](https://github.com/event4u-app/agent-config/commit/54f024250c67781e34e6242450fdf97d9165bfee))
+* **release:** refuse an uncurated head before the first remote state (#1813) ([2b3d2b3](https://github.com/event4u-app/agent-config/commit/2b3d2b3474e6520ef696466aff583a76381f002c))
+
+### Documentation
+
+* **evidence:** record drain run 18 (#1823) ([76ed568](https://github.com/event4u-app/agent-config/commit/76ed5684367667000ed5b1bd2bc5af13bdef8943))
+
+### Other
+
+* complete road-to-wired-instruments (#1818) ([b0a03a7](https://github.com/event4u-app/agent-config/commit/b0a03a7af314cb231f4385005f7d73e8a6ca5dd4))
+* complete road-to-self-description-truth (#1817) ([e200ba2](https://github.com/event4u-app/agent-config/commit/e200ba2f97590d2686bd9a292cf54025fb47cee2))
+* complete road-to-governed-skill-scouting (#1815) ([576ff4f](https://github.com/event4u-app/agent-config/commit/576ff4fcc2568e7619ff489ddcf301f32f30da9f))
+* complete road-to-ship-control-coverage (#1816) ([5a0ca3a](https://github.com/event4u-app/agent-config/commit/5a0ca3a16fb8f0ab54df734de885abee8c8fedff))
+* complete road-to-artifact-location-and-doctor-reach (#1819) ([517afcb](https://github.com/event4u-app/agent-config/commit/517afcb524d201ad08b5d820638d84b2b3cdeb7b))
+* complete road-to-cascading-base-integration (#1821) ([6735abf](https://github.com/event4u-app/agent-config/commit/6735abf3f789bd2d1549aa02937189ba9fe2b952))
+* **topology:** repair three claims that went false, and record why the carrier stays parked (#1820) ([1fe6fa0](https://github.com/event4u-app/agent-config/commit/1fe6fa0fb3facfa7e735b68ee92e04616fbb3843))
+* six survivors of the 2026-09-b inbox round (#1814) ([0415101](https://github.com/event4u-app/agent-config/commit/04151017860e0066e20832bdee49221e047ee33e))
+* park governed-evidence-production to later/, and resolve metered-backend-park with (b) (#1811) ([c6b4f64](https://github.com/event4u-app/agent-config/commit/c6b4f640736c78554d7ce03f27257f44368ef1ae))
+
+Tests: 20776 (+215 since 14.14.0)
+
 ## [14.14.0](https://github.com/event4u-app/agent-config/compare/14.13.0...14.14.0) (2026-09-02)
 
 ### Release highlights

package/README.md

@@ -27,7 +27,7 @@ It is both deep **and** disciplined — and honest about what it deliberately is
 - **Surgical uninstall** — removes only its own keys from a shared host config (matched by JSON-pointer + SHA-256), never a neighbour tool's entries.<!-- claim:surgical-uninstall -->
 - **Pack-scoped install** — writes the active pack only, not a 500-artefact dump.
 
-**What it deliberately is *not*** — the core is a governance layer with optional, individually opt-in embedded engines (code intelligence, gated reach, the setup GUI, the bench lab — [ADR-124](docs/decisions/ADR-124-embedded-engine-doctrine.md)): **no background daemon, no separate state database, no self-rewriting memory, no auto-build pipeline.** Engines are never mandatory, never default-on without measured lift, and terminate with the command that invoked them. The host agent runs the loop; every learned change is human-reviewed; the same layer stays portable across tools. Capability without a process to babysit.
+**What it deliberately is *not*** — the core is a governance layer with optional, individually opt-in embedded engines (code intelligence, gated reach, the setup GUI, the bench lab — [ADR-124](docs/decisions/ADR-124-embedded-engine-doctrine.md)): **no mandatory or always-on daemon, no separate state database, no self-rewriting memory, no auto-build pipeline.** Engines are never mandatory, never default-on without measured lift, and terminate with the command that invoked them. The host agent runs the loop; every learned change is human-reviewed; the same layer stays portable across tools. Capability without a process to babysit.
 
 > **Where this comes from (honest provenance).** The skills, rules and personas
 > are distilled from real production work on TypeScript and PHP codebases. The

package/dist/agent-src/commands/pr/create.md

@@ -149,10 +149,25 @@ than less: the gate can now read the PR's real `baseRefName` from the forge, so
 a stacked or release-line PR is measured against the branch it will actually
 merge into instead of against the repo default.
 
+**The whole SEQUENCE is executable now, not just this step.** All six steps —
+fetch, integrate the base set, regenerate, verify, re-check freshness, then hand
+the push back to you — sit behind one runner, and a dry mode prints each step
+with its outcome without touching the tree. In this package that runner is the
+`push-ready` target in `taskfiles/dev.yml`; a consumer project wires the same
+six steps behind whatever its own task runner is, with steps 3 and 4 being its
+own generator and verify commands. The pre-push hook refuses and points at the
+runner; the hook never merges, because a merge inside `pre-push` rewrites the
+tree at the moment you believe your work is finished.
+
 **The resolution is executable now, not just described.**
 `./scripts-run src/scripts/sync_pr_branch` resolves the base from the open PR
 (so a stacked or release-line PR is measured against what it actually merges
-into), fetches, and merges it in when the branch is behind. On a conflict it
+into), fetches, and merges it in when the branch is behind. The base is a
+**set**: when the PR targets something other than the default branch, whether
+the default branch joins the set is a per-target policy read from the *target's
+own commit*, never from this branch — see `src/scripts/_lib/branch_convergence.ts`.
+A non-default target with no policy entry is refused, not guessed. The default
+is merged first, so a conflict surfaces against the broader base first. On a conflict it
 STOPS and splits the conflicted paths into generated, remeasured and authored —
 the first has one correct resolution (regenerate), the second's is to re-run the
 measurement on the merged tree (a ratchet baseline, where picking a side is how

package/dist/agent-src/contexts/execution/terminal-states.md

@@ -10,7 +10,7 @@ identically to a win: a loop that hit its iteration cap, a gate that could not
 run, an approval that never came — each ending with a summary that says the work
 is done.
 
-## The six states
+## The seven states
 
 | state | what it means | what it is NOT |
 |---|---|---|
@@ -20,6 +20,23 @@ is done.
 | `approval-required` | The work is ready and the next action crosses a gate the agent may not cross alone. | Not `blocked`: nothing is missing, and the run produced a deliverable. |
 | `exhausted` | A declared budget ran out — iterations, tokens, spend, wall-clock. | **Never** `success`, whatever partial progress exists. The cap firing IS the outcome. |
 | `stagnated` | The budget has not run out and progress has stopped: the same failure signature is repeating. | Not `exhausted`. Stagnation is detected by a NO-PROGRESS signal, not by a counter. |
+| `premise-invalidated` | The plan premise moved: the situational-awareness fingerprint the run engaged under (`origin/main` plus every open PR head) was re-observed and differs. | Not `exhausted` — the budget is untouched. Not `stagnated` — progress may have been fine. Not `blocked` — nothing is missing; what the run knew went stale. |
+
+**Why `premise-invalidated` is not `blocked` with a reason field.** Added
+2026-09-03 (AI council, unanimous) when the continuation ladder needed a word for
+a run whose base ref and open-PR set had moved under it. Reusing `blocked` keeps
+the arity at six and loses two things that are not decorative: the ability to
+COUNT premise invalidation separately from a missing precondition, and
+type-enforced exhaustive handling in the consumers that switch on the state. The
+remedies differ as sharply as `exhausted` and `stagnated` do — `blocked` says
+obtain the missing thing, `premise-invalidated` says re-probe and re-plan against
+the world as it now is.
+
+A consumer that cannot represent it reports `blocked`, which is the DECLARED
+downgrade in `_lib/outcome_vocabularies.ts` and not an ad-hoc substitution. The
+value domain is versioned (`RUN_TERMINAL_VOCABULARY_VERSION`, v2), and a reader
+meeting an unrecognised persisted value reports it as not-recorded rather than
+crashing or guessing.
 
 **Why `exhausted` and `stagnated` are separate**, when both stop a loop: they
 have different remedies and different next actions. `exhausted` says *the work
@@ -40,8 +57,8 @@ budget-exhausted stop indistinguishable from a completed one in a dashboard.
 | `[~]` | deferred | `blocked` or `approval-required`, undistinguished |
 | `[-]` | cancelled | *(a decision, not a run outcome)* |
 
-**Three states have no glyph, and they are exactly the ones a validation budget
-and a hard-blocker class produce:**
+**Four states have no glyph, and they are exactly the ones a validation budget,
+a hard-blocker class and a drift detector produce:**
 
 1. **`exhausted`** — the run stopped at its cap with the step still open. The
    glyph stays `[ ]`, which is identical to *never attempted*. A reader cannot
@@ -51,6 +68,8 @@ and a hard-blocker class produce:**
 3. **`approval-required`** — collapses into `[~]` alongside `blocked`, so a
    roadmap cannot show that the work is FINISHED and waiting on a human, versus
    not startable at all.
+4. **`premise-invalidated`** — same `[ ]` again, and the reader cannot see that
+   the step was not attempted because the plan it belonged to went stale.
 
 **This is recorded rather than fixed.** Adding glyphs would change a format that
 `update_roadmap_progress`, `check_roadmap_trackable`, `lint_empty_roadmaps` and
@@ -75,11 +94,13 @@ applies unchanged.
 
 ## Reporting contract
 
-A closing report names the state **by one of the six words above**, and where the
+A closing report names the state **by one of the seven words above**, and where the
 state is not `success` it names what would change it:
 
 - `exhausted` — the budget that ran out, its value, and what was achieved inside it.
 - `stagnated` — the repeating failure signature, and the attempts it survived.
+- `premise-invalidated` — which fingerprint was claimed, which was observed, and
+  that the next act is a re-probe rather than another iteration.
 - `blocked` — the missing thing, by name.
 - `approval-required` — the exact action awaiting approval.
 - `clean-no-op` — what was scanned, so "nothing to do" is distinguishable from

package/dist/agent-src/rules/missing-skill-recovery.md

@@ -31,8 +31,8 @@ and dropped **402 entries** from the model-visible skills list
 projection moved that to 330 and stripped the descriptions anyway.
 
 402 is the **host's own count against the host's own denominator**, which is not
-this package's: the same install projects 297 skills, so the figure is a
-magnitude, never a subtraction anyone can perform. What it establishes is that
+this package's: that install projected 297 skills <!-- count: dated -->, so the figure
+is a magnitude, never a subtraction anyone can perform. What it establishes is that
 most of the catalogue is not model-visible — which is all this rule needs.
 
 So "I do not see a skill for this" is evidence about the **delivery**, never

package/dist/agent-src/rules/token-budget-discipline.md

@@ -57,8 +57,8 @@ A skill claiming `rich` MUST satisfy ALL three:
    carry a `load_context` pointer instead.
 
 3. **≤ 15 % of the suite's skills may claim `rich`** — the CI linter counts
-   rich-tagged skills and hard-fails if the ratio is exceeded. With ~290 skills
-   (288 as-of 2026-08), the cap is ~43. Do not claim `rich` for convenience; claim it only for
+   rich-tagged skills and hard-fails if the ratio is exceeded. With 299 skills
+   the cap is ~44. Do not claim `rich` for convenience; claim it only for
    irreducible-complexity skills.
 
 ## Telegraph-speak amendment

package/dist/agent-src/templates/agents/agent-project-settings.example.yml

@@ -39,7 +39,7 @@ schema_version: 1
 # CI guard: a release bump of `package.json` must update this value
 # in lockstep — see scripts/check_template_pin_drift.ts (road-to-
 # portable-runtime-and-update-check P3.3).
-agent_config_version: "14.14.0"
+agent_config_version: "14.15.0"
 
 # --- Project identity ---
 project:

package/dist/cli/registry.js

@@ -77,6 +77,7 @@ export const REGISTRY = [
     { name: 'hooks:install', disposition: 'delegate', synopsis: 'Install the combined pre-commit hook.' },
     { name: 'hooks:status', disposition: 'delegate', synopsis: 'Print the runtime hook matrix (per-platform install + bindings).' },
     { name: 'hooks:doctor', disposition: 'delegate', synopsis: 'Diagnose hook health: concerns, posture, missing trampolines (read-only).' },
+    { name: 'hooks:effect', disposition: 'delegate', synopsis: 'Does a bound concern actually FIRE here? Sandboxed synthetic dispatch per concern; five per-concern states, four host verdicts (read-only).' },
     { name: 'routing:doctor', disposition: 'delegate', synopsis: 'Live routing diagnosis: per-gate ACTIVE/INACTIVE with reason, chain, freshness, orchestration state (read-only).' },
     { name: 'workspace:doctor', disposition: 'delegate', synopsis: 'Workspace identity with provenance (repo root, main worktree, current worktree, branch, PR base) + session claim and worktree pressure (read-only).' },
     { name: 'route:explain', disposition: 'delegate', synopsis: 'Deterministic rule-routing trace for one prompt: matched triggers, tier, disposition, budget, rejected candidates (trigger-match level only).' },