mandrel 2.58.0 → 2.60.0

package/.agents/rules/known-tooling-behavior.md

@@ -66,16 +66,30 @@ commands** the script knows nothing about, so a locally green
 | `check-dead-exports.js` | no | no | **yes** |
 | `check-dead-exports.js --production` | no | no | **yes** |
 | `check-context-budget.js` | no | no | **yes** |
-| `check-workflow-citations.js` | no | no | **no** |
+| `check-workflow-citations.js` | no | no | **yes** |
 | `check-cyclomatic.js` | no | no | **yes** |
 | `check-schema-references.js` | no | no | **yes** |
 | `check-knip-entries.js` | no | no | **yes** |
 | `check-baseline-scope.js` | no | no | **no** |
 
-Two are still in **no** local aggregate command —
-`check-workflow-citations.js` and `check-baseline-scope.js`. Each is reachable
-only as its own npm script (`check:workflow-citations`, `baselines:scope`) or a
-direct invocation.
+One is still in **no** local aggregate command — `check-baseline-scope.js`,
+reachable only as its own npm script (`baselines:scope`) or a direct
+invocation. `check-workflow-citations.js` joined `verify` when Story #5340
+demoted it to a report; before that it was exempted because
+`tests/check-workflow-citations.test.js` re-ran its ratchet through the `test`
+step.
+
+**Two of the nine cannot fail this job any more** (Story #5340, ADR
+20260917-5340). `check-workflow-citations.js` prints a per-file provenance
+count and always exits 0 — it reads no baseline at all, and the one it used
+to ratchet against is deleted.
+`check-context-budget.js` fails on the `alwaysLoaded` tier alone: its
+`workflow` and `mandatoryRead` tiers are printed, and the per-file 8 KB
+agent-boot ceiling and the row-vs-tree drift gate are gone. A green run of
+either is therefore **not** evidence the numbers held — read the report. Both
+ratchets used to fail a *rise*, so a prose fix had to be paid for with an
+unrelated trim in the same commit; the always-loaded gate is the one that
+stays because every session and every subagent spawn re-pays that closure.
 
 `prune-baseline-orphans.js --check` left this table in v2.32.0: it no longer
 runs in CI in any mode. It reports the same absent / out-of-scope rows as
@@ -84,25 +98,24 @@ the required job made the scope gate's inherited-divergence warning
 unreachable — a stale row on `main` red every open PR regardless of who landed
 it. It stays the operator's remedy, via `npm run baselines:prune`.
 
-`check-context-budget.js` additionally runs in `.husky/pre-push`. It is
-zero-tolerance in **both** directions — a change that *shrinks* the
-always-loaded doc closure reds it exactly as growth does, and the remedy is a
-committed baseline refresh, not a smaller diff.
+`check-context-budget.js` and `check-workflow-citations.js` additionally run in
+`.husky/pre-push`, as reports. A shrink in the always-loaded closure no longer
+reds anything either (Story #5313): it is reported, and the close's write-back
+seam commits the lower total on the Story branch.
 
 **Reproduce.**
 
 ```bash
 node .agents/scripts/check-baselines.js --format text   # names the 3 gates it ran
 sed -n '/name: baselines/,/windows-smoke/p' .github/workflows/ci.yml | grep 'js'
-grep "label: '" .agents/scripts/run-verify.js            # the 9 steps verify covers
+sed -n '/^const STEPS/,/^];/p' .agents/scripts/run-verify.js  # the steps verify covers
 ```
 
 **Safe move.** `npm run verify` is the closest local mirror. Run
-`node .agents/scripts/check-workflow-citations.js` alongside it when the change
-touches workflow prose under `.agents/workflows/`, and
-`npm run baselines:scope && npm run baselines:prune -- --check` when it adds,
-deletes or moves files inside a scored `targetDirs` root. Reproducing only the
-`.agentrc.json` command before a push is a false green.
+`npm run baselines:scope && npm run baselines:prune -- --check` alongside it
+when the change adds, deletes or moves files inside a scored `targetDirs`
+root. Reproducing only the `.agentrc.json` command before a push is a false
+green.
 
 ## 3. The two dead-export passes disagree, and the production pass is silent without `!`
 
@@ -148,3 +161,40 @@ hand-edit `baselines/dead-exports*.json`: a hand-written row set is the one
 input no gate re-derives. `.agents/rules/test-seams.md` governs which seams
 are sanctioned. Never remove the `!` suffixes from `knip.json` to quieten the
 production pass.
+
+## 4. A green pre-push is only evidence the two scopes agreed because `--ref` outranks config
+
+**Behavior.** `.husky/pre-push` captures coverage and then scores it, both
+against a literal `origin/main`. The CRAP half of `quality-preview.js` is a
+function of complexity **and** coverage, so the preview is only reading its
+own tree if the artifact under it was captured over the same change set.
+Both steps resolve that set through **one** rule, stated once in
+`resolveChangedFilesRef` (`.agents/scripts/lib/changed-files.js`): the ref the
+caller named wins, and
+`delivery.quality.gates.crap.incrementalCoverage.baseRef` is the default for a
+caller that names none — the close-validation gate, which passes no `--ref`.
+Both `coverage-capture` paths and the preview's CRAP baseline join call it, so
+one hook invocation cannot resolve two refs.
+
+Before Story #5365 the configured value outranked the flag. The preview has
+no config ref to consult, so a consumer that set `baseRef` captured against
+one ref while the preview scored another, and the preview could read an
+artifact whose scope was not its own — the stale-artifact read the
+capture-before-preview ordering (Story #5356) exists to prevent, reopened by
+configuration rather than by editing the hook. **This repository sets no
+`baseRef`**, so the divergence was invisible here: a green local run proved
+nothing about a consumer's.
+
+**Reproduce.**
+
+```bash
+node -e "import('./.agents/scripts/lib/changed-files.js').then(({ resolveChangedFilesRef }) => { const crap = { incrementalCoverage: { baseRef: 'develop' } }; console.log(resolveChangedFilesRef({ crap, ref: 'origin/main' })); console.log(resolveChangedFilesRef({ crap, ref: null })); })"
+# → origin/main   (the hook's flag wins over a configured baseRef)
+# → develop       (config still answers a caller that named no ref)
+grep -n 'origin/main' .husky/pre-push   # the same literal on both steps
+```
+
+**Safe move.** Read a green pre-push as evidence about CRAP only when both
+hook steps still carry the same literal ref. Moving one means moving the
+other; adding another consumer of the change set means routing it through
+`resolveChangedFilesRef` rather than reading `baseRef` directly.

package/.agents/runtime-deps.json

@@ -1,12 +1,17 @@
 {
-  "description": "Single source of truth for the third-party npm packages the .agents/ framework scripts import at runtime. This manifest ships *inside* .agents/ so it travels with the mandrel package into consumer projects. Consumers must provide these packages in the node_modules resolvable from their repository root (the framework scripts free-ride on the consumer's install). The `dependencies` block is fail-fast enforced by the preflight guard (.agents/scripts/lib/runtime-deps/ensure-installed.js); the `optionalDependencies` block lists packages that are imported behind graceful-degradation paths (try/catch or dev-only watch tooling) and are therefore declared but never preflight-blocked. A drift test (tests/scripts/runtime-deps-drift.test.js) asserts every third-party import under .agents/scripts/** is declared here. Version ranges mirror the framework's own root package.json.",
+  "description": "Single source of truth for the third-party npm packages the .agents/ framework scripts import at runtime. This manifest ships *inside* .agents/ so it travels with the mandrel package into consumer projects. Consumers must provide these packages in the node_modules resolvable from their repository root (the framework scripts free-ride on the consumer's install). The `dependencies` block is fail-fast enforced by the preflight guard (.agents/scripts/lib/runtime-deps/ensure-installed.js); the `optionalDependencies` block lists packages that are imported behind graceful-degradation paths (try/catch or dev-only watch tooling) and are therefore declared but never preflight-blocked. A drift test (tests/scripts/runtime-deps-drift.test.js) asserts every third-party import under .agents/scripts/** is declared here. Version ranges mirror the framework's own root package.json. Two entries need their reason recorded. `@babel/parser` is pinned to ^7 because the kernel's fixed plugin list does not parse under 8.x, and the range is documentation rather than enforcement \u2014 `.agents/` resolves from the consumer's own node_modules \u2014 so the kernel asserts the resolved major at load. `babel-runtime` is declared even though no framework script imports it: four of the metric-core packages require it and none of them declares it, so it resolved only by being hoisted from the parse/dispatch plumbing this framework used to depend on. Declaring it makes the closure explicit and preflight-enforced. The import-vs-manifest drift guard only flags imported-but-undeclared, so a declared-but-unimported peer repair like this is legal by design.",
   "dependencies": {
+    "@babel/parser": "^7.29.3",
     "ajv": "^8.20.0",
     "ajv-formats": "^3.0.1",
+    "babel-runtime": "^6.26.0",
+    "escomplex-plugin-metrics-module": "^0.1.0",
+    "escomplex-plugin-syntax-babylon": "^0.1.0",
     "js-yaml": "^4.3.1",
     "minimatch": "^10.0.0",
     "picomatch": "^4.0.4",
-    "typhonjs-escomplex": "^0.1.0"
+    "typhonjs-ast-walker": "^0.2.1",
+    "typhonjs-escomplex-commons": "^0.1.1"
   },
   "optionalDependencies": {
     "@commitlint/load": "^21.0.0",

package/.agents/schemas/acceptance-eval-verdict.schema.json

@@ -2,7 +2,7 @@
   "$schema": "https://json-schema.org/draft/2020-12/schema",
   "$id": "acceptance-eval-verdict",
   "title": "Acceptance Self-Eval Verdict",
-  "description": "Structured output of one independent (fresh-context) acceptance self-eval round (Story #3819). Produced by the critic pass in the Story-implementation phase (helpers/deliver-story.md Step 1a), written under the gitignored temp tree, and consumed by acceptance-eval.js to enforce the bounded round cap and decide whether the Story proceeds to `closing`, redrafts, or escalates to `agent::blocked`. The author MUST NOT be the evaluator — the verdict is an independent judgment of the caller-injected change set against each inline acceptance[] item.",
+  "description": "Structured output of one acceptance self-eval round (Story #3819). One file per Story covering every inline acceptance[] item exactly once, authored by the round's single verdict owner — the inline self-eval under ceremonyProfile minimal/standard, a fresh-context maker-blind critic under strict (Story #5343) — in the Story-implementation phase (helpers/deliver-story.md Step 1a). Written under the gitignored temp tree and consumed by acceptance-eval.js, which enforces the bounded round cap and decides whether the Story proceeds to `closing`, redrafts, or escalates to `agent::blocked`. Whoever owns it scores the caller-injected change set afresh against each criterion, treating the implementation reasoning as untrusted; the gate is the scorer of that verdict, never a second pass over the criteria.",
   "type": "object",
   "required": ["storyId", "schemaVersion", "round", "criteria"],
   "properties": {

package/.agents/schemas/agentrc.schema.json

@@ -904,7 +904,7 @@
                         "baseRef": {
                           "type": "string",
                           "minLength": 1,
-                          "description": "Git ref the changed-file set is computed against. Omitted falls back to the gate’s own `--ref` (`main`)."
+                          "description": "Default git ref the changed-file set is computed against, for a caller that passes no `--ref`. A `--ref` the caller named wins over this value, so a caller that anchors another gate on the same ref — `.husky/pre-push`, which passes `--ref origin/main` and then previews `--changed-since origin/main` — resolves ONE scope for both steps (Story #5365). Omitted and unpassed, the gate’s own default `main` applies."
                         }
                       },
                       "additionalProperties": false
@@ -1742,17 +1742,12 @@
         },
         "feedbackLoop": {
           "type": "object",
-          "description": "Opt-out toggles for the close-time auto-file graduators. All default to auto-filing on.",
+          "description": "Opt-in toggle for the close-time retro auto-file graduator, plus the friction recurrence window. Auto-filing defaults to OFF (Story #5341).",
           "properties": {
-            "auditResultsAutoFile": {
-              "type": "boolean",
-              "description": "When true (default), the close-time audit-results graduator auto-files non-blocking audit-results findings as follow-up issues routed by source classification. Set to false to suppress auto-filing; findings remain accessible in the structured comments on the Story.",
-              "default": true
-            },
             "retroProposals": {
               "type": "boolean",
-              "description": "When true (default), the retro auto-files its actionable routed proposals as meta::<framework-gap|consumer-improvement> + friction::<category> issues via the graduator pre-parsed-findings seam, and the rendered retro sections list the filed issue numbers instead of paste-ready gh command stanzas. Set to false to fall back to the command stanzas.",
-              "default": true
+              "description": "When true, the retro auto-files its actionable routed proposals as meta::<framework-gap|consumer-improvement> + friction::<category> issues via the graduator pre-parsed-findings seam, and the rendered retro sections list the filed issue numbers instead of paste-ready gh command stanzas. Defaults to false (Story #5341), which renders the command stanzas instead.",
+              "default": false
             },
             "frictionWindowDays": {
               "type": "integer",
@@ -1843,7 +1838,7 @@
         },
         "routing": {
           "type": "object",
-          "description": "v2 delivery-spawn routing: role-scoped boot contexts and the ceremony profile. The v1 singleDelivery epic-route kill-switch was removed in Stage 6; the freshCriticSampleRate sampling floor was retired in Story #5313.",
+          "description": "v2 delivery-spawn routing: role-scoped boot contexts and the ceremony profile. The v1 singleDelivery epic-route kill-switch was removed in Stage 6; the freshCriticSampleRate sampling floor was retired in Story #5313 and the derived-level ceremony routing in Story #5343.",
           "properties": {
             "roleScopedAgents": {
               "type": "boolean",
@@ -1853,7 +1848,7 @@
             "ceremonyProfile": {
               "type": "string",
               "enum": ["minimal", "standard", "strict"],
-              "description": "Acceptance-ceremony depth. minimal = always inline critic; strict = always fresh-context critic; standard (default) = routed off the change level derived from the Story diff: high or underivable → fresh, low → inline.",
+              "description": "Acceptance-ceremony depth — who authors the Story acceptance verdict. minimal and standard (default) = the inline self-eval, whatever the diff touches; strict = a fresh-context maker-blind critic. Review depth is a separate decision and still derives `deep` for any sensitive path (review-depth.js).",
               "default": "standard"
             },
             "closeAndLand": {

package/.agents/schemas/crap-baseline.schema.json

@@ -18,7 +18,7 @@
     },
     "escomplexVersion": {
       "type": "string",
-      "description": "Version of the typhonjs-escomplex dependency used to produce this baseline. A mismatch against the installed dependency fails the gate closed (different kernel semantics are not negotiable)."
+      "description": "Version of the package that computes the metrics (escomplex-plugin-metrics-module) used to produce this baseline. Informational: the axis that compared it was removed in Story #5336 — the v2 envelope does not carry the field, so it could not fire. scoringSemantics is the axis that rejects an incompatible scorer."
     },
     "tsTranspilerVersion": {
       "type": "string",

package/.agents/schemas/crap-report.schema.json

@@ -14,7 +14,7 @@
     },
     "escomplexVersion": {
       "type": "string",
-      "description": "typhonjs-escomplex version used for the scan."
+      "description": "Version of the package that computes the metrics (escomplex-plugin-metrics-module) used for the scan."
     },
     "summary": {
       "type": "object",

package/.agents/schemas/story-deliver-terminal.schema.json

@@ -2,7 +2,7 @@
   "$schema": "https://json-schema.org/draft/2020-12/schema",
   "$id": "https://github.com/dsj1984/mandrel/blob/main/.agents/schemas/story-deliver-terminal.schema.json",
   "title": "story-deliver-terminal",
-  "description": "The single terminal envelope a Story delivery invocation emits (Story #4543). Before this schema, the delivery tail had two divergent prose return contracts — one in .agents/workflows/helpers/deliver-story.md, a different one in .agents/agents/story-worker.md — and neither was validated by anything, so a caller could not tell a landed Story from a parked one without re-probing GitHub. This is the SSOT both now reference rather than restate. status is exactly one of landed | pending | blocked | failed | escalated; phase names where the run ended; tail carries per-step booleans so a partial-tail degradation is visible without failing an otherwise-landed merge; nextCommand names the single command that resumes or remediates the run, drawn from the same vocabulary deliver-recover.js prints. escalated (Story #4746) is the one status emitted BEFORE a Story exists — /deliver-light's suitability gate refusing an over-scope prompt — which is why storyId is null exactly there and non-null everywhere else.",
+  "description": "The single terminal envelope a Story delivery invocation emits (Story #4543). Before this schema, the delivery tail had two divergent prose return contracts — one in .agents/workflows/helpers/deliver-story.md, a different one in .agents/agents/story-worker.md — and neither was validated by anything, so a caller could not tell a landed Story from a parked one without re-probing GitHub. This is the SSOT both now reference rather than restate. status is exactly one of landed | pending | blocked | failed | escalated; phase names where the run ended; tail carries per-step booleans so a partial-tail degradation is visible without failing an otherwise-landed merge; nextCommand names the single command that resumes or remediates the run, drawn from the same vocabulary deliver-recover.js prints. escalated (Story #4746) is the one status emitted BEFORE a Story exists — /deliver-light's suitability gate refusing a prompt — which is why storyId is null exactly there and non-null everywhere else.",
   "type": "object",
   "required": [
     "kind",
@@ -21,7 +21,7 @@
     },
     "status": {
       "type": "string",
-      "description": "landed — the PR merged, the Story is agent::done, and the post-land tail was attempted. pending — a bounded wait expired with the PR still in flight; NO label was mutated and no merge.unlanded event was emitted, so the run is resumable via nextCommand. blocked — a classified hard block; the Story carries agent::blocked and blocked.blockClass names the class. failed — a phase crashed; phase names which one. escalated — the /deliver-light suitability gate refused an over-scope prompt under --yes; nothing was created and the session ENDS here, nextCommand naming the /mandrel-plan invocation that owns the work instead.",
+      "description": "landed — the PR merged, the Story is agent::done, and the post-land tail was attempted. pending — a bounded wait expired with the PR still in flight; NO label was mutated and no merge.unlanded event was emitted, so the run is resumable via nextCommand. blocked — a classified hard block; the Story carries agent::blocked and blocked.blockClass names the class. failed — a phase crashed; phase names which one. escalated — /deliver-light's suitability gate refused a prompt on an un-waivable risk rule or an un-ledgered verdict; nothing was created and the LIGHT PATH ends here, nextCommand naming the /mandrel-plan invocation that owns the work instead. Story #5344 narrowed what refuses (the predicted-shape ceilings are gone) and loosened what follows (that command may be run in the same session, seeded with escalation.reasons) without touching this shape.",
       "enum": ["landed", "pending", "blocked", "failed", "escalated"]
     },
     "phase": {
@@ -179,7 +179,7 @@
     },
     "nextCommand": {
       "type": ["string", "null"],
-      "description": "The single command that advances this work from where it stopped, or null when status === \"landed\" and nothing remains. Shares its vocabulary with deliver-recover.js so recovery and normal resumption speak one language. For status escalated it is the /mandrel-plan invocation the operator runs in a FRESH session — the one case where the command is a slash command rather than a script, because the work needs planning, not resumption.",
+      "description": "The single command that advances this work from where it stopped, or null when status === \"landed\" and nothing remains. Shares its vocabulary with deliver-recover.js so recovery and normal resumption speak one language. For status escalated it is the /mandrel-plan invocation that owns the work — the one case where the command is a slash command rather than a script, because the work needs planning, not resumption. Story #4746 required it to run in a FRESH session; Story #5344 permits the same session when it is seeded with escalation.reasons, and helpers/deliver-light.md carries the seeding contract.",
       "minLength": 1
     },
     "elapsedSeconds": { "type": "number", "minimum": 0 },

package/.agents/scripts/README.md

@@ -13,7 +13,17 @@ GitHub Actions surfaces first. `check-knip-entries.js` derives that
 caller set mechanically, so a CLI no invoker names is dead, not
 operator-only.
 
-It reads the entry list from whatever configuration knip itself would
+The one script an operator-facing workflow drives **by name, repeatedly**, is
+[`deliver-run.js`](deliver-run.js): one beat of a multi-Story
+`/mandrel-deliver` run — it ticks the ready set from live state, writes each
+ready Story's dispatch prompt under `<tempRoot>/run-<id>/`, keeps the run
+ledger that replaces hand-maintained dispatch bookkeeping, and renders the
+`single-story-close.js` command for every hand-off. Everything else in the
+delivery chain (`resolve-stories.js`, `single-story-init.js`,
+`single-story-close.js`, `stories-wave-tick.js`) is reached through it or
+through a workflow step.
+
+`check-knip-entries.js` reads the entry list from whatever configuration knip itself would
 load — `knip.json`, `knip.jsonc`, `.knip.json(c)`, `knip.ts`, `knip.js`,
 `knip.config.ts`, `knip.config.js`, or `package.json#knip` — evaluating
 TS/JS modules rather than parsing them, and counting entries declared

package/.agents/scripts/acceptance-eval.js

@@ -3,12 +3,11 @@
 /**
  * acceptance-eval.js — bounded per-Story acceptance self-eval gate (Story #3819).
  *
- * The Story-implementation phase runs ONE verdict-owner per acceptance
- * cluster (Story #4723) — the fresh-context critic when
- * `ceremony-routing.js` sensitivity-routes the cluster fresh, the
- * contract-identical inline self-eval otherwise — which scores the
- * caller-injected change set against each inline `acceptance[]` item and
- * emits one verdict file per round
+ * The Story-implementation phase runs ONE verdict-owner per Story
+ * (Story #4723, narrowed by #5343) — the contract-identical inline self-eval
+ * under `ceremonyProfile` `minimal` / `standard`, a fresh-context maker-blind
+ * critic under `strict` — which scores the caller-injected change set against
+ * each inline `acceptance[]` item and emits one verdict file per round
  * (`.agents/schemas/acceptance-eval-verdict.schema.json`). This CLI is the
  * deterministic SCORER of that single authored verdict — it validates and
  * decides, it never re-scores the criteria as an independent additional
@@ -41,18 +40,17 @@
  * tier along with the per-AC-cluster `--epic <id> --cluster <id>` mode that
  * scored an Epic `## Acceptance Table` against a `main..epic/<id>` diff.)
  *
- * One gate call per round (Story #4951). A round may fan out into N parallel
- * maker-blind cluster critics, but their per-cluster verdicts are merged by
- * the caller into ONE verdict — `criteria[]` in acceptance-array order — and
- * scored here exactly once. Invoking the gate per cluster instead would burn
- * one Story-level round per cluster (distinct fingerprints defeat the replay
- * guard) and race the `signals.ndjson` round ledger. The gate reads the
- * Story's own `acceptance[]` count off its body (Story #5313) and rejects a
- * verdict whose `criteria[]` length differs **before** scoring, so the
- * mistake costs no round. `--expected-criteria` is still accepted but is
- * redundant with the derived count: when both are known they must agree.
- * An inline-owned verdict is one file scored in one call — the cluster
- * merge applies only to fresh critics.
+ * One gate call per round (Story #4951; Story #5343 retired the cluster
+ * protocol that used to fan a round out). The round's owner authors ONE
+ * verdict covering every `acceptance[]` item — `criteria[]` in
+ * acceptance-array order — and it is scored here exactly once. A second call
+ * inside one round would burn a Story-level round for nothing (distinct
+ * fingerprints defeat the replay guard) and race the `signals.ndjson` round
+ * ledger. The gate reads the Story's own `acceptance[]` count off its body
+ * (Story #5313) and rejects a verdict whose `criteria[]` length differs
+ * **before** scoring, so the mistake costs no round. `--expected-criteria` is
+ * still accepted but is redundant with the derived count: when both are known
+ * they must agree.
  *
  * CLI:
  *   --story <id>              Story ID (required).
@@ -180,13 +178,13 @@ function parseCliArgs(argv) {
 }
 
 /**
- * The merge contract, stated once so both the flag error and the coverage
+ * The coverage contract, stated once so both the flag error and the coverage
  * error name the same shape the caller has to produce.
  */
 const MERGE_CONTRACT =
-  'One round = N parallel cluster critics -> ONE merged verdict -> ONE gate call: ' +
-  "merge every cluster's records into a single criteria[] in acceptance[] order, " +
-  'one per acceptance item, before scoring.';
+  'One round = ONE verdict -> ONE gate call: the verdict must carry one ' +
+  'criteria[] record per acceptance[] item, in acceptance-array order, ' +
+  'before scoring.';
 
 /**
  * Read the Story's `acceptance[]` count off its body (Story #5313), so the
@@ -292,8 +290,8 @@ export function resolveExpectedCriteria(raw) {
  * Reject a verdict that does not cover exactly `expectedCriteria` criteria.
  *
  * Called **before** `runAcceptanceEval`, which is where the round ledger is
- * read and appended — so a partial cluster verdict handed to the gate by
- * mistake costs no round and can never escalate a `redraft` into a `block`.
+ * read and appended — so a partial verdict handed to the gate by mistake
+ * costs no round and can never escalate a `redraft` into a `block`.
  *
  * Exported for tests.
  *
@@ -544,9 +542,9 @@ export async function runAcceptanceEvalCli(
 
   const verdict = validateVerdictImpl(parsed);
 
-  // Story #4951 / #5313: a merged verdict must cover every acceptance[] item,
-  // and the count comes from the Story body itself. This runs before the
-  // round ledger is touched, so a partial cluster verdict is a free mistake.
+  // Story #4951 / #5313: the verdict must cover every acceptance[] item, and
+  // the count comes from the Story body itself. This runs before the round
+  // ledger is touched, so a partial verdict is a free mistake.
   const config = resolveConfigImpl();
   const expected = await resolveExpectedCriteriaCount({
     storyId,

package/.agents/scripts/ceremony-derive.js

@@ -23,9 +23,16 @@
  *   { storyId, baseRef, headRef, files, enumerated, level, classes, profile,
  *     mode, reason, verdictOwner }
  *
- * `files` is the one change set every acceptance critic must be handed; the
- * caller never lets a critic re-enumerate it. `files: null` means the diff
- * could not be enumerated, which routes to the fail-safe fresh critic.
+ * `files` is the one change set the verdict owner must be handed; the caller
+ * never lets it re-enumerate the diff. `files: null` means the diff could not
+ * be enumerated.
+ *
+ * `level` / `classes` are what **review depth** reads (`resolveDepth`), and a
+ * sensitive class still resolves `deep`. They do not route the verdict owner:
+ * `verdictOwner` follows the ceremony profile alone (Story #5343) —
+ * `inline-self-eval` under `minimal` / `standard`, `fresh-critic` under
+ * `strict` — and since Story #5366 the resolver does not even accept the
+ * level, so the two signals cannot be confused for one decision.
  *
  * Exit codes: 0 on a derived decision (including the `null` fail-safe — an
  * unenumerable diff is a decision, not an error), 1 on a usage error.
@@ -114,13 +121,11 @@ export function deriveCeremony(
   const { level, classes } = deriveChangeLevelImpl({
     changedFiles: changeSet.files,
   });
-  // `derivedLevel` is the level STRING — handing the `{ level, classes }`
-  // object here was the transcription slip this CLI exists to retire.
-  const ceremony = resolveCeremonyImpl({
-    derivedLevel: level,
-    clusterIndex: 0,
-    ceremonyProfile,
-  });
+  // The level is still derived and still printed — review depth reads it —
+  // but the ceremony profile alone resolves the verdict owner, and the
+  // resolver accepts nothing else, so nothing here can be talked into a
+  // different owner by the diff.
+  const ceremony = resolveCeremonyImpl({ ceremonyProfile });
   return {
     storyId,
     baseRef,