mandrel 2.36.0 → 2.38.0

package/.agents/scripts/lib/orchestration/planning/authoring-context.js

@@ -2,7 +2,7 @@
  * phases/authoring-context.js — emit-context phase.
  *
  * Builds the authoring context the host LLM (or the
- * `/plan` author step) needs to write the Tech Spec.
+ * `/mandrel-plan` author step) needs to write the Tech Spec.
  * Returns a plain JSON-serialisable object; never hits the network beyond
  * the provider call needed to load the Epic.
  */
@@ -31,12 +31,12 @@ import { buildMemoryPoolAdvisory } from './memory-pool-advisory.js';
  * pull-on-demand, mirroring the Story #4324 delivery-children cutover).
  *
  * Ensures a docs digest exists at the **same** per-Epic temp path the
- * `/deliver` story sub-agents already consume
+ * `/mandrel-deliver` story sub-agents already consume
  * (`<tempRoot>/epic-<seedIssueId>/docs-digest.md`) via the shared
  * `ensureDocsDigest` export in `docs-digest.js` — one generator, one file,
  * reused across the planning and delivery surfaces for the same Epic. The
  * envelope carries only the digest path, not embedded doc content; the
- * planner (host LLM driving the `/plan` author step) reads the digest
+ * planner (host LLM driving the `/mandrel-plan` author step) reads the digest
  * and pulls a full file/section on demand when it bears on the decision.
  *
  * Returns `null` — a silent no-op — when `project.docsContextFiles` is not
@@ -80,7 +80,7 @@ async function buildPlanningDocsContext({ seedIssueId, settings, cwd }) {
  * Story #4977 — the scan itself stays uncapped (a faithful `.feature`
  * index), but the envelope-bound result is truncated to
  * `BDD_SCENARIOS_BYTE_BUDGET` via `capBddScenarios` so a mature Gherkin
- * corpus cannot alone consume the `/plan` context-envelope ceiling. Callers
+ * corpus cannot alone consume the `/mandrel-plan` context-envelope ceiling. Callers
  * needing the raw count read `totalScenarios` vs `includedScenarios`.
  *
  * @returns {ReturnType<typeof capBddScenarios>}
@@ -97,7 +97,7 @@ function scanBddScenariosBestEffort() {
 
 /**
  * Build the authoring context the host LLM (or the
- * `/plan` author step) needs to write the Tech Spec.
+ * `/mandrel-plan` author step) needs to write the Tech Spec.
  *
  * `docsContext` is digest-first (Story #4433): a pointer at the per-Epic
  * docs digest rather than embedded doc content, `null` when
@@ -133,7 +133,7 @@ export async function buildAuthoringContext(
   const githubCfg = opts.github ?? null;
 
   // Story #4952 — these five gathers share no data, so they run under bounded
-  // concurrency instead of five sequential awaits on the interactive `/plan`
+  // concurrency instead of five sequential awaits on the interactive `/mandrel-plan`
   // path. `concurrentMap` preserves input order, so the destructuring is
   // positional and the produced context is identical to the serial build:
   //
@@ -149,7 +149,7 @@ export async function buildAuthoringContext(
   //      memory corpus is delivery retrospectives whose subject IS a delivered
   //      Story — and its directory (`~/.claude/projects/<repo>/memory/`) never
   //      resolved, because harness project dirs are cwd-slugs. This renders no
-  //      per-entry verdict at all: it stats and counts, and the `/plan` spine
+  //      per-entry verdict at all: it stats and counts, and the `/mandrel-plan` spine
   //      surfaces `recommend` at Gate #1. Filesystem-only and total;
   //   5. Story #2554 — open meta feedback issues, so retro signals are routed
   //      into durable substrates rather than lost in chat. Best-effort:

package/.agents/scripts/lib/orchestration/planning/decomposer-context.js

@@ -1,5 +1,5 @@
 /**
- * decomposer-context.js — system-prompt builder for the `/plan` authoring
+ * decomposer-context.js — system-prompt builder for the `/mandrel-plan` authoring
  * pass (Story #2466; v2 keeps the prompt helper, drops the Epic-fetch
  * decomposition envelope).
  *

package/.agents/scripts/lib/orchestration/planning/memory-pool-advisory.js

@@ -1,5 +1,5 @@
 /**
- * memory-pool-advisory.js — the `/plan` Phase 0 memory-hygiene advisory.
+ * memory-pool-advisory.js — the `/mandrel-plan` Phase 0 memory-hygiene advisory.
  *
  * Replaces the retired memory-freshness pre-flight (Story #2557 / #4414) in
  * the same slot, fixing both of that design's defects:
@@ -10,7 +10,7 @@
  *      replaced by `-` — so the old path never resolved in any consumer and
  *      the scan was a silent no-op everywhere.
  *   2. **A named consumer.** The retired scanner emitted a per-entry staleness
- *      verdict nothing read. This emits one advisory the `/plan` spine
+ *      verdict nothing read. This emits one advisory the `/mandrel-plan` spine
  *      surfaces at Gate #1, recommending `/memory-consolidate`.
  *
  * It also drops the semantic that made the old scanner unfixable: it renders
@@ -131,7 +131,7 @@ function countEntries({ poolDir, fsImpl }) {
  * Build the `memoryPoolAdvisory` envelope field.
  *
  * Advisory only — it carries **no routing authority**, mirroring
- * `deliverLightSuggestion`. The `/plan` spine surfaces `recommend` at Gate #1;
+ * `deliverLightSuggestion`. The `/mandrel-plan` spine surfaces `recommend` at Gate #1;
  * nothing auto-runs, and nothing here mutates the operator's memory store.
  *
  * @param {object} [opts]

package/.agents/scripts/lib/orchestration/pr-watch.js

@@ -50,35 +50,6 @@ import { applyBehindUpdate } from './behind-recovery.js';
  * values collapse to `'skipped'` so any future GitHub state we haven't
  * enumerated still maps into the vocabulary.
  */
-/**
- * The raw check-state tokens `normalizeCheckState` recognizes (lowercased).
- * A token absent from this set is one we have NOT enumerated — the watch
- * path collapses it to `'skipped'` (validate-anything), but a fail-closed
- * consumer (the auto-merge arming probe) must treat it as unknown-therefore-
- * blocking rather than trust the `'skipped'` collapse. Exported so that
- * stricter consumer lives here as the single vocabulary owner.
- */
-export const RECOGNIZED_CHECK_STATES = Object.freeze(
-  new Set([
-    '',
-    'pending',
-    'queued',
-    'in_progress',
-    'requested',
-    'waiting',
-    'success',
-    'completed',
-    'failure',
-    'startup_failure',
-    'neutral',
-    'cancelled',
-    'timed_out',
-    'action_required',
-    'stale',
-    'skipped',
-  ]),
-);
-
 export function normalizeCheckState(raw) {
   const v = String(raw ?? '')
     .trim()
@@ -325,7 +296,7 @@ export function allTerminal(outcomes) {
  * red check): a `'still-running'` map means "re-arm the watch / hand off
  * to `/loop`," never "the change is broken."
  */
-export const STILL_RUNNING = 'still-running';
+const STILL_RUNNING = 'still-running';
 
 /**
  * Promote any leftover `'pending'` outcomes to the schema-valid

package/.agents/scripts/lib/orchestration/reassert-status-column.js

@@ -38,14 +38,14 @@ import { ColumnSync, columnForLabels } from './column-sync.js';
  * long enough to outlast the typical GitHub built-in workflow fire
  * timing without ballooning the close-path wall clock.
  */
-export const DEFAULT_POLL_ATTEMPTS = 4;
+const DEFAULT_POLL_ATTEMPTS = 4;
 
 /**
  * Default delay between drift checks (ms). Five seconds is a sweet
  * spot: short enough that the operator doesn't perceive the close
  * path as hung, long enough that one re-fire usually wins the race.
  */
-export const DEFAULT_POLL_DELAY_MS = 5000;
+const DEFAULT_POLL_DELAY_MS = 5000;
 
 /**
  * Production sleep function. Tests inject `sleepFn: () => Promise.resolve()`

package/.agents/scripts/lib/orchestration/remote-verifier.js

@@ -3,11 +3,11 @@
  * remote-verifier.js — deterministic "is there a live, pushable remote?"
  * evidence for the delivery entry seams. Issue #4483.
  *
- * `/deliver` could silently shortcut the entire orchestration — building
+ * `/mandrel-deliver` could silently shortcut the entire orchestration — building
  * the delivery inline and committing to local `main` without pushing —
  * when the driving agent *perceived* the environment had no live GitHub
  * remote. The judgment was vibes, not fact. This module gives the entry
- * seams (`single-story-init.js` for v2 `/deliver`) a verified
+ * seams (`single-story-init.js` for v2 `/mandrel-deliver`) a verified
  * probe result to record in their envelopes so the workflow can branch on
  * `remoteVerified: true|false` deterministically: use the remote, or
  * transition to `agent::blocked` quoting the probe output — never a

package/.agents/scripts/lib/orchestration/resolve-stories.js

@@ -2,7 +2,7 @@
  * lib/orchestration/resolve-stories.js — resolve a set of Story ids into the
  * `{ stories, dag, done }` envelope the delivery scheduler consumes.
  *
- * This is the ONE resolution step for `/deliver`. It generalizes the
+ * This is the ONE resolution step for `/mandrel-deliver`. It generalizes the
  * envelope shaping proven by the retired `resolve-plan-run.js` and fixes the
  * two defects that shipped with it:
  *
@@ -77,7 +77,7 @@ export function toStoryRecord(issue, requestedId) {
   if (!labels.includes(TYPE_LABELS.STORY)) {
     throw new Error(
       `[resolve-stories] Issue #${id} is not a Story (labels: ${labels.join(', ') || 'none'}). ` +
-        `/deliver accepts ${TYPE_LABELS.STORY} tickets only — close it or re-plan it as a v2 Story.`,
+        `/mandrel-deliver accepts ${TYPE_LABELS.STORY} tickets only — close it or re-plan it as a v2 Story.`,
     );
   }
   const body = String(issue?.body ?? '');
@@ -351,7 +351,7 @@ export function buildStoriesEnvelope({
   return {
     kind: 'stories',
     // `dispatchMode` (Story #4722, #4736, #4829): the resolver reports the
-    // per-Story execution mode so `/deliver` reads one field — `inline` (run
+    // per-Story execution mode so `/mandrel-deliver` reads one field — `inline` (run
     // deliver-story in the router's own session: no story-worker /
     // acceptance-critic sub-agent boots) or `subagent` (the conservative
     // default). Model-side fan-out only; close gates are untouched.
@@ -385,7 +385,7 @@ export function buildStoriesEnvelope({
 /**
  * Parse and validate the `--ids` list, expanding any `A-B` dash range.
  *
- * A contiguous span is how an operator names a plan run — `/deliver 4922 -
+ * A contiguous span is how an operator names a plan run — `/mandrel-deliver 4922 -
  * 4926` — so the range is expanded here rather than transcribed by the host.
  * `stories-wave-tick.js --stories` reads through this same function, which is
  * what keeps the sequencing set identical to the resolved one.

package/.agents/scripts/lib/orchestration/review-providers/codex.js

@@ -57,7 +57,7 @@ export const CODEX_REMEDIATIONS = Object.freeze({
  *
  * Exported so tests can extend the list without monkey-patching `os`.
  */
-export const DEFAULT_PLUGIN_MARKERS = Object.freeze([
+const DEFAULT_PLUGIN_MARKERS = Object.freeze([
   path.join(os.homedir(), '.claude', 'plugins', 'codex-plugin-cc'),
   path.join(os.homedir(), '.claude', 'plugins', 'openai', 'codex-plugin-cc'),
 ]);
@@ -194,7 +194,7 @@ export function parseCodexFindings(rawStdout) {
  * @param {{ baseRef: string, headRef: string, depth?: import('./types.js').ReviewDepth }} args
  * @returns {string}
  */
-export function buildCodexReviewPrompt({ baseRef, headRef, depth }) {
+function buildCodexReviewPrompt({ baseRef, headRef, depth }) {
   return (
     `/codex:review --base ${baseRef} --head ${headRef} --wait ` +
     `${renderDepthDirective(depth)}`
@@ -214,7 +214,7 @@ export function buildCodexReviewPrompt({ baseRef, headRef, depth }) {
  * @param {{ baseRef: string, headRef: string, depth?: import('./types.js').ReviewDepth }} args
  * @returns {{ status: number, stdout: string, stderr: string }}
  */
-export function defaultInvokeCodexReview({ baseRef, headRef, depth }) {
+function defaultInvokeCodexReview({ baseRef, headRef, depth }) {
   const cliArgs = [
     '--print',
     buildCodexReviewPrompt({ baseRef, headRef, depth }),

package/.agents/scripts/lib/orchestration/review-providers/findings-renderer.js

@@ -102,7 +102,7 @@ export function renderFinding(finding) {
  * @param {ReadonlyArray<string>} messages
  * @returns {string[]}  lines to append (empty when no messages)
  */
-export function renderManualPromptsSection(messages) {
+function renderManualPromptsSection(messages) {
   const filtered = Array.isArray(messages)
     ? messages.filter((m) => typeof m === 'string' && m.trim().length > 0)
     : [];

package/.agents/scripts/lib/orchestration/review-providers/native.js

@@ -267,8 +267,14 @@ function isJsMaintainabilityFile(relPath) {
  * `reportFn` forces the serial path (the injected scorer cannot cross the
  * worker boundary).
  *
+ * `serialThreshold` overrides the pool-vs-serial cutover for one call. It is
+ * the seam the parity test uses to drive the pooled branch: Story #5109
+ * retuned `POOL_SERIAL_THRESHOLD` from 8 to 256 against measured data, and a
+ * fixture sized to clear the old number would otherwise have gone quietly
+ * serial while still claiming to cover the pool.
+ *
  * @param {string[]} changedFiles
- * @param {{ reportFn?: Function, classifier?: Function, runOnPoolFn?: typeof runOnPool, headRef?: string|null, gitSpawnFn?: typeof gitSpawn, readHeadSourceFn?: typeof readHeadSource }} [deps]
+ * @param {{ reportFn?: Function, classifier?: Function, runOnPoolFn?: typeof runOnPool, headRef?: string|null, gitSpawnFn?: typeof gitSpawn, readHeadSourceFn?: typeof readHeadSource, serialThreshold?: number }} [deps]
  * @returns {Promise<{ totalFiles: number, jsFiles: number, maintainability: object[], criticalFindings: Finding[], mediumFindings: Finding[] }>}
  */
 export async function analyzeChangedFiles(
@@ -280,6 +286,7 @@ export async function analyzeChangedFiles(
     headRef = null,
     gitSpawnFn = gitSpawn,
     readHeadSourceFn = readHeadSource,
+    serialThreshold = SERIAL_THRESHOLD,
   } = {},
 ) {
   const results = {
@@ -308,7 +315,7 @@ export async function analyzeChangedFiles(
   const customReportFn = reportFn != null;
 
   // Serial path: small batches, or whenever a caller injects its own scorer.
-  if (jsFiles.length < SERIAL_THRESHOLD || customReportFn) {
+  if (jsFiles.length < serialThreshold || customReportFn) {
     for (let i = 0; i < jsFiles.length; i += 1) {
       const relPath = jsFiles[i];
       const source = sources[i];

package/.agents/scripts/lib/orchestration/review-providers/review-depth.js

@@ -28,7 +28,7 @@
  *
  * @type {Readonly<Record<ReviewDepth, string>>}
  */
-export const DEPTH_DIRECTIVES = Object.freeze({
+const DEPTH_DIRECTIVES = Object.freeze({
   light:
     'Review depth: LIGHT. Run a single pass focused on spec adherence over the ' +
     'changed surface — confirm the change matches its stated intent. Reduce the ' +
@@ -54,7 +54,7 @@ export const DEPTH_DIRECTIVES = Object.freeze({
  * @param {unknown} depth
  * @returns {ReviewDepth}
  */
-export function normalizeDepth(depth) {
+function normalizeDepth(depth) {
   return depth === 'light' || depth === 'deep' ? depth : 'standard';
 }
 

package/.agents/scripts/lib/orchestration/review-providers/security-review.js

@@ -59,7 +59,7 @@ export const SECURITY_REVIEW_REMEDIATIONS = Object.freeze({
  * @param {{ spawnFn?: typeof spawnSync }} [opts]
  * @returns {boolean}
  */
-export function defaultProbeClaudeCli(opts = {}) {
+function defaultProbeClaudeCli(opts = {}) {
   const spawnFn = opts.spawnFn ?? spawnSync;
   try {
     const result = spawnFn('claude', ['--version'], {
@@ -162,7 +162,7 @@ export function parseSecurityReviewFindings(rawStdout) {
  * (Story #3937) so a high-risk Epic instructs the model toward a deeper
  * second-pass review while a low-risk one keeps it light.
  */
-export const SECURITY_REVIEW_INVOKE_PROMPT =
+const SECURITY_REVIEW_INVOKE_PROMPT =
   'Run /security-review against the diff `{baseRef}`...`{headRef}` ' +
   'for {scopeLabel} #{ticketId}. {depthDirective} After the review, emit ' +
   'ONLY a JSON array of findings on stdout with this exact shape:\n\n' +
@@ -208,7 +208,7 @@ export function buildSecurityReviewPrompt(input) {
  * @param {ReviewInput} input
  * @returns {{ status: number, stdout: string, stderr: string }}
  */
-export function defaultInvokeSecurityReview(input) {
+function defaultInvokeSecurityReview(input) {
   const prompt = buildSecurityReviewPrompt(input);
   const result = spawnSync('claude', ['--print', prompt], {
     encoding: 'utf-8',

package/.agents/scripts/lib/orchestration/run-epilogue.js

@@ -1,5 +1,5 @@
 /**
- * run-epilogue.js — real per-run closeout for `/deliver --run`.
+ * run-epilogue.js — real per-run closeout for `/mandrel-deliver --run`.
  *
  * After the last Story in a multi-Story plan-run lands, this module:
  *   1. Selects the cross-Story audit lens roster over the combined landed
@@ -100,7 +100,7 @@ export function planRunEpilogue({ planRunId, stories } = {}) {
     };
   }
 
-  // Positional `/deliver 101 102` has no plan-run label. Synthesize a
+  // Positional `/mandrel-deliver 101 102` has no plan-run label. Synthesize a
   // stable adhoc id from the sorted Story set so the epilogue still
   // anchors comments / audit roster without requiring `--run`.
   const effectiveRunId =

package/.agents/scripts/lib/orchestration/single-story-close/phases/base-sync.js

@@ -83,7 +83,7 @@ export async function runBaseSyncPhase({
           : syncResult.stderr
             ? `: ${syncResult.stderr.slice(0, 200)}`
             : '') +
-        `. Story transitioned to ${AGENT_LABELS.BLOCKED}; resolve in ${syncCwd} and re-run \`/deliver ${storyId}\`.`,
+        `. Story transitioned to ${AGENT_LABELS.BLOCKED}; resolve in ${syncCwd} and re-run \`/mandrel-deliver ${storyId}\`.`,
     );
   }
   progress('SYNC', `✅ Synced from origin/${baseBranch} (${syncResult.kind}).`);
@@ -182,7 +182,7 @@ export function buildSyncFailureCommentBody({
   const lines = [
     `### ${heading}`,
     '',
-    '`/deliver` close-validation passed, but the pre-push',
+    '`/mandrel-deliver` close-validation passed, but the pre-push',
     `sync against \`origin/${baseBranch}\` could not complete. The Story has`,
     `been transitioned to \`agent::blocked\`. To resume:`,
     '',

package/.agents/scripts/lib/orchestration/single-story-close/phases/conventional-subject.js

@@ -343,7 +343,7 @@ function scanForBreaking(text, readHeaderBang) {
  *     writing the commit that does the breaking.
  *   - **The Story body** — the same footer, written as its own line anywhere
  *     in the Story issue (the `## Spec` block is the natural home). This is
- *     the declarative path: `/plan` can state the break up front and close
+ *     the declarative path: `/mandrel-plan` can state the break up front and close
  *     propagates it even when no individual commit remembered the footer.
  *     Only the footer form counts; prose describing a break does not, because
  *     a keyword-free sentence is exactly what release-please cannot parse.

package/.agents/scripts/lib/orchestration/single-story-close/phases/normalize-pr-title.js

@@ -191,7 +191,11 @@ function synthesizeSubject({ description, storyId, commitMessages }) {
  * @returns {string}
  */
 function buildPrBody({ storyId, breakingNotes }) {
-  const lines = [`Closes #${storyId}`, '', '_Auto-opened by `/deliver`._'];
+  const lines = [
+    `Closes #${storyId}`,
+    '',
+    '_Auto-opened by `/mandrel-deliver`._',
+  ];
   if (breakingNotes.length > 0) {
     lines.push('', `BREAKING CHANGE: ${breakingNotes.join(' ')}`);
   }

package/.agents/scripts/lib/orchestration/single-story-close/phases/options.js

@@ -87,6 +87,50 @@ export function resolveWaitForMerge({
   };
 }
 
+/**
+ * Flags this CLI advertised in its `--help` descriptor and never implemented
+ * (Story #5100). `--dry-run` was transcribed into the usage block by the
+ * #4750 self-describing sweep and read by nothing in this pipeline —
+ * `git log -S dryRun` over it is empty in every revision — so an operator who
+ * trusted it got a real base-sync merge, an `agent::blocked` Story, and a
+ * terminal envelope that then captured `deliver-recover`'s routing.
+ * `--no-evidence` names a working flag on the gate wrappers
+ * (`evidence-gate.js`), but the slot this CLI parsed had no reader at all.
+ *
+ * Deleting them from the descriptor is NOT enough. `parseSprintArgs` runs
+ * `parseArgs` with `strict: false`, so an unknown flag is silently ignored and
+ * the close proceeds for real — the identical injury, now with no help text to
+ * explain it. Fail closed instead, exactly as `parseMergeWatchMode` does for an
+ * unrecognized mode: throw during option parsing, before the first phase, so
+ * the CLI boundary reports `failed` at phase `init` and nothing is mutated.
+ */
+const RETIRED_FLAGS = Object.freeze({
+  '--dry-run':
+    'this pipeline has never had a dry-run mode; it was advertised in error.',
+  '--no-evidence':
+    'per-close evidence control was never wired here; the working flag of that name belongs to the gate wrappers.',
+});
+
+/**
+ * Reject a retired flag before any phase runs.
+ *
+ * Matches `--flag` and `--flag=value`; a bare `--` terminator or a positional
+ * that merely contains the text is not a flag and does not trip it.
+ *
+ * @param {string[]} argv argv tail (`process.argv.slice(2)`)
+ * @throws {Error} naming the flag, why it is gone, and that nothing was mutated
+ */
+function assertNoRetiredFlags(argv) {
+  for (const [flag, why] of Object.entries(RETIRED_FLAGS)) {
+    const present = argv.some((a) => a === flag || a.startsWith(`${flag}=`));
+    if (!present) continue;
+    throw new Error(
+      `${flag} was retired: ${why} Nothing was mutated — no branch, label, ` +
+        `comment, or PR was touched. Re-run without it to close for real.`,
+    );
+  }
+}
+
 /**
  * Parse and resolve all CLI / injection options for `runSingleStoryClose`.
  *
@@ -116,7 +160,14 @@ export function parseCloseOptions({
   // which is precisely what `resolveFlag` already does below, preferring the
   // param over the parsed slot. One expression per flag now serves both
   // callers, so a new flag is added in one place instead of two that can drift.
-  const parsed = storyIdParam === undefined ? parseSprintArgs() : {};
+  // The guard runs on the argv door only: an injecting caller supplies
+  // `storyIdParam` and never reads argv, so the host process's flags (a test
+  // runner's, say) are none of its business.
+  let parsed = {};
+  if (storyIdParam === undefined) {
+    assertNoRetiredFlags(process.argv.slice(2));
+    parsed = parseSprintArgs();
+  }
   // Preserve undefined so resolveWaitForMerge can apply the closeAndLand
   // config default when neither flag was supplied.
   const waitForMergeExplicit = resolveFlag(

package/.agents/scripts/lib/orchestration/single-story-close/phases/review-block.js

@@ -19,7 +19,7 @@ export async function handleCriticalReviewBlock({
     '### Code review blocked delivery',
     '',
     `The Story-scope review reported **${criticalCount} critical blocker(s)** on ${prUrl}.`,
-    'Remediate the posted findings, then re-run `/deliver`.',
+    'Remediate the posted findings, then re-run `/mandrel-deliver`.',
     '',
     'If you have reviewed a finding and judged it wrong, re-run close with',
     '`--override-review-block "<reason>"` rather than merging by hand — see',

package/.agents/scripts/lib/orchestration/single-story-close/phases/wrong-tree-guard.js

@@ -1,7 +1,7 @@
 /**
  * phases/wrong-tree-guard.js — detect worktree/main-checkout edit divergence.
  *
- * Story #3364 — `/single-story-deliver` (and `/deliver`) materializes a
+ * Story #3364 — `/single-story-deliver` (and `/mandrel-deliver`) materializes a
  * per-Story worktree and instructs the agent to `cd` into it before editing.
  * On Windows that guidance is silently insufficient: `cd <workCwd>` steers the
  * Bash tool's working directory, but the path-based Edit/Write tools operate on
@@ -232,7 +232,7 @@ export function formatWrongTreeFinding({ storyId, strayFiles, worktreePath }) {
     `**Recovery:** relocate these edits into the worktree (re-apply them under ` +
     `\`${worktreePath}\`), restore the main checkout ` +
     `(\`git -C <main-repo> checkout -- <files>\`), then re-run ` +
-    `\`/deliver ${storyId}\`.`
+    `\`/mandrel-deliver ${storyId}\`.`
   );
 }
 
@@ -307,7 +307,7 @@ async function abortWrongTree({
       `diff-path set while the worktree (${worktreePath}) is the active work ` +
       `tree. Close aborted to avoid an empty-diff PR. Stray files: ` +
       `${strayFiles.join(', ')}. Relocate the edits into the worktree, ` +
-      `restore the main checkout, then re-run /deliver ${storyId}.`,
+      `restore the main checkout, then re-run /mandrel-deliver ${storyId}.`,
   );
 }
 

package/.agents/scripts/lib/orchestration/single-story-close/runner.js

@@ -84,7 +84,7 @@ async function emitTerminal({ terminal, result, config }) {
  * issue as `completed` (the `Closes #<id>` footer firing on merge — the work
  * IS on the base branch) or as `not_planned` (superseded by a re-plan,
  * abandoned — nothing ever merged). Reporting the second as `landed` told
- * `/deliver` that work had reached `main` when no PR ever merged, which would
+ * `/mandrel-deliver` that work had reached `main` when no PR ever merged, which would
  * also satisfy any dependent Story waiting on it. Fail loudly instead: being
  * handed an abandoned Story is an input error only the operator can resolve.
  *
@@ -283,7 +283,7 @@ async function openAndReviewPr({
     });
     throw new Error(
       `[single-story-close] Story-scope review reported ${criticalCount} critical blocker(s) on PR ${prUrl}. ` +
-        'Auto-merge was not enabled. Remediate the findings posted to the PR and re-run `/deliver`. ' +
+        'Auto-merge was not enabled. Remediate the findings posted to the PR and re-run `/mandrel-deliver`. ' +
         'If you have reviewed a finding and judged it wrong, re-run with ' +
         '`--override-review-block "<reason>"` rather than merging by hand.',
     );

package/.agents/scripts/lib/orchestration/single-story-lease-guard.js

@@ -105,7 +105,7 @@ export async function acquireStoryLease({
     steal,
     renderRefusal: (result) =>
       `single-story lease: Story #${storyId} is currently held by @${result.owner}. ` +
-      'Another /deliver run owns this Story. Coordinate with that ' +
+      'Another /mandrel-deliver run owns this Story. Coordinate with that ' +
       'operator, or re-run with --steal to forcibly transfer the claim once you ' +
       'have confirmed the other run is dead. (A foreign assignee always ' +
       'blocks unless stolen.)',

package/.agents/scripts/lib/orchestration/story-close/format-autofix.js

@@ -63,7 +63,7 @@ export const FORMAT_AUTOFIX_TIMEOUT_EXIT_CODE = 124;
  * @param {(args: string[], opts: object) => string} git
  * @returns {string[]}
  */
-export function listDirtyPaths(cwd, git) {
+function listDirtyPaths(cwd, git) {
   const out = git(['status', '--porcelain'], {
     cwd,
     encoding: 'utf8',
@@ -91,10 +91,7 @@ export function listDirtyPaths(cwd, git) {
  * }} [opts]
  * @returns {{ writeCmdString: string, writeCmd: string, writeArgs: string[] }}
  */
-export function resolveFormatterCmd({
-  commands,
-  dropTrailingDot = false,
-} = {}) {
+function resolveFormatterCmd({ commands, dropTrailingDot = false } = {}) {
   // `resolveFormatWriteCommand` reads `config.project.commands`; wrap the
   // caller-supplied `commands` map into that canonical shape.
   const writeCmdString = resolveFormatWriteCommand({ project: { commands } });

package/.agents/scripts/lib/orchestration/story-deliver-terminal.js

@@ -47,7 +47,7 @@ export const TERMINAL_ENVELOPE_KIND = 'story-deliver-terminal';
  * `pending` is **3**, not 0 and not 1. Zero would tell a headless caller the
  * Story landed when it did not; one would conflate a resumable slow-CI wait
  * with a hard block and send the operator to diagnose branch protection that
- * is working fine. A distinct code is what lets `/deliver` resume without
+ * is working fine. A distinct code is what lets `/mandrel-deliver` resume without
  * classifying.
  */
 export const TERMINAL_EXIT_CODES = Object.freeze({
@@ -61,7 +61,7 @@ export const TERMINAL_EXIT_CODES = Object.freeze({
    * terminal, not a new one, so giving it a fresh code would fork a vocabulary
    * callers already branch on. It stays distinct from `landed` (nothing was
    * delivered) and from `blocked`/`failed` (nothing is wrong — the work simply
-   * belongs to `/plan`).
+   * belongs to `/mandrel-plan`).
    */
   escalated: 2,
 });
@@ -74,12 +74,12 @@ export const TERMINAL_STATUSES = Object.freeze([
   'escalated',
 ]);
 
-/** Cap on the prompt echoed into an escalation's `/plan` next command. */
+/** Cap on the prompt echoed into an escalation's `/mandrel-plan` next command. */
 const PLAN_PROMPT_MAX = 200;
 
 /**
  * Render an operator prompt safe to sit inside the double quotes of the
- * `/plan "<prompt>"` next command: collapse newlines (a next command is one
+ * `/mandrel-plan "<prompt>"` next command: collapse newlines (a next command is one
  * line by contract), escape backslashes and double quotes so the quoting
  * cannot be broken out of, and cap the length so a long prompt does not turn
  * the envelope into a transcript. Total — a non-string yields the empty
@@ -133,7 +133,7 @@ export const NEXT_COMMANDS = Object.freeze({
   recover: (storyId) =>
     `node .agents/scripts/deliver-recover.js --story ${storyId}`,
   /**
-   * Hand over-scope work to `/plan` — the next command of an `escalated`
+   * Hand over-scope work to `/mandrel-plan` — the next command of an `escalated`
    * terminal (Story #4746).
    *
    * The only entry here that is a slash command rather than a script, and
@@ -143,7 +143,7 @@ export const NEXT_COMMANDS = Object.freeze({
    * the workflow's escalation section for why running it in the escalating
    * session is forbidden.
    */
-  escalateToPlan: (prompt) => `/plan "${quoteForPlan(prompt)}"`,
+  escalateToPlan: (prompt) => `/mandrel-plan "${quoteForPlan(prompt)}"`,
 });
 
 /**
@@ -260,15 +260,15 @@ export function buildTerminalEnvelope({
  * `/deliver-light`'s suitability gate already decided correctly when it
  * overrode a `lite` self-verdict on shape; what it lacked was an outcome a
  * session could not walk past. A mandrel-bench 2.13.0 light-arm run did
- * exactly that — it read the gate's `escalate-plan`, then invoked `/plan`
+ * exactly that — it read the gate's `escalate-plan`, then invoked `/mandrel-plan`
  * in the same session and delivered. The continuation was not harmless:
  * planning inside a session already framed as small work authored ONE Story
- * against the scenario's 3-5 contract, where a fresh `/plan` session on the
+ * against the scenario's 3-5 contract, where a fresh `/mandrel-plan` session on the
  * identical seed authored four. Escalation silently produced the very
  * under-decomposition the guard exists to prevent.
  *
  * So the outcome is a validated envelope with its own exit code, naming the
- * `/plan` command that owns the work, and asserting per artifact that nothing
+ * `/mandrel-plan` command that owns the work, and asserting per artifact that nothing
  * was started. Every guarantee here is enforced by the schema rather than by
  * prose: `storyId` must be null, `escalation.created.*` are pinned `false`.
  *
@@ -288,10 +288,10 @@ export function buildEscalationTerminal({
 }) {
   const quoted = quoteForPlan(prompt);
   if (quoted === '') {
-    // An escalation whose next command is `/plan ""` hands the operator
+    // An escalation whose next command is `/mandrel-plan ""` hands the operator
     // nothing — the same walk-past-able non-outcome in envelope clothing.
     throw new TypeError(
-      'buildEscalationTerminal: a non-empty prompt is required — the escalated terminal exists to name the /plan invocation that owns the work',
+      'buildEscalationTerminal: a non-empty prompt is required — the escalated terminal exists to name the /mandrel-plan invocation that owns the work',
     );
   }
   const recorded = (Array.isArray(reasons) ? reasons : []).filter(
@@ -305,7 +305,9 @@ export function buildEscalationTerminal({
       reasons:
         recorded.length > 0
           ? recorded
-          : ['predicted scope exceeds the light ceilings — escalate to /plan'],
+          : [
+              'predicted scope exceeds the light ceilings — escalate to /mandrel-plan',
+            ],
       // Not computed from anything: the escalation path returns before the
       // receipt/init call sites, so these are the assertion that it did.
       created: { receiptStory: false, storyBranch: false, worktree: false },

package/.agents/scripts/lib/orchestration/story-follow-ups.js

@@ -24,7 +24,7 @@ import {
 } from './retro-proposals.js';
 import { upsertStructuredComment } from './ticketing.js';
 
-export const FOLLOW_UPS_COMMENT_TYPE = 'follow-ups';
+const FOLLOW_UPS_COMMENT_TYPE = 'follow-ups';
 
 /** Milliseconds in one day — the unit `frictionWindowDays` is expressed in. */
 const MS_PER_DAY = 24 * 60 * 60 * 1000;