mandrel 2.25.0 → 2.26.0

package/.agents/workflows/helpers/deliver-reference.md

@@ -12,6 +12,34 @@ Reference-only detail split out of [`deliver.md`](../deliver.md) so the
 always-resident spine stays lean. Nothing here is a new MUST —
 it is the mechanics an operator consults when the matching lever is engaged.
 
+## Ranges (`4922 - 4926`) {#ranges}
+
+A contiguous span is how an operator reads a plan run, so the dash range is a
+first-class id shape rather than prose to interpret — `/deliver 4922 - 4926`
+means exactly the five ids in it.
+
+**Pass the span through; never expand it by hand.** Every id-list flag on the
+delivery path takes range tokens — `resolve-stories.js --ids`,
+`stories-wave-tick.js --stories` and `--dispatched`, and
+`plan-run-epilogue.js --stories`. Normalize the operator's spacing away and hand
+the scripts one unspaced token (`--ids 4922-4926`), mixed freely with singles
+and commas (`--ids 4901,4922-4926`); overlaps dedupe. A hand-typed enumeration
+is where an id gets dropped or invented, and the drop is silent.
+
+The shared expander (`lib/util/parse-id-list.js`) refuses rather than guesses,
+so a typo fails where it was typed instead of resolving the wrong set:
+
+| Input | Outcome |
+| --- | --- |
+| `4922-4926`, `4922 - 4926` | Expands to the inclusive span. En and em dashes, and a `#` on either endpoint, are accepted too. |
+| `4926-4922` | Refused — write it low-to-high. |
+| `1-4926` | Refused — above the 50-id span cap (`MAX_RANGE_SPAN`). |
+| `4922-`, `-4926`, `4922-4923-4924` | Refused as a malformed token. |
+
+The cap is per range token, not per run: a genuine 60-Story delivery is still
+expressible as two ranges, but a slipped digit cannot fan out into a live
+resolution sweep of thousands of issues.
+
 ## Sequencing edge cases (`stories-wave-tick.js`)
 
 **What "discovered, not declared" means concretely.** `resolve-stories.js` reads
@@ -63,6 +91,18 @@ different operator holds, unless you pass `--steal`. Assignee-based withholding
 needs `github.operatorHandle` set (in `.agentrc.local.json`); without it the
 probe logs a warning and leans on init's lease refusal alone.
 
+**Overlapping footprints are reserved across beats, not just within one.** A
+Story sharing a **concrete** path with a still-implementing Story is withheld
+and named in `inFlightReservation: { available, withheld: [{ id, blockedBy,
+reason }], note }`, where `reason` is `in-flight-earlier-beat` or
+`foreign-lease`. Like `foreignHeld` this is neither a failure nor a wedge — the
+Story re-admits automatically once its blocker leaves the in-flight set — and
+it exists so an unfilled slot is explained rather than mysterious. A **glob**
+footprint (or the UNKNOWN sentinel for an unparseable body) reserves nothing
+across beats; it still serializes its own beat. Reservation needs the in-flight
+Stories' footprints, so it is a `--probe-live` capability: under `--dag` the
+report is `available: false` and selection de-conflicts within the beat only.
+
 ## Dispatch mechanics (role-scoped by default)
 
 **A single-Story run executes inline.** Sub-agent isolation is
@@ -96,6 +136,18 @@ footprint keeps the fresh acceptance critic), and the `route::lite` label
 remains a human-visible hint only, never the control signal — a lost or
 never-written label cannot misroute delivery.
 
+**Issue a beat's spawns in one turn.** A wave tick hands you a ready set, not a
+queue: those Stories have no dependency edge between them (the resolver already
+withheld any that do) and no shared write paths (each owns its own worktree and
+branch). Dispatch them the way
+[`parallel-tooling.md`](parallel-tooling.md) Rule 3 prescribes — **N `Agent`
+calls issued together in a single assistant turn**, one per ready Story, not
+`Agent` → wait → `Agent`. Serial dispatch is compliant with every other rule on
+this page and costs the run a full Story's implementation time per sibling for
+nothing; the wave aggregator is built for the parallel shape. Respect
+`delivery.deliverRunner.concurrencyCap`: when the ready set exceeds it, slice
+into batches of `cap` and dispatch each batch in its own turn.
+
 **Dispatch each `ready` Story (role-scoped by default).** When
 `delivery.routing.roleScopedAgents` is enabled (the **default**) and the host
 exposes agent dispatch, spawn each ready Story as its own
@@ -235,8 +287,32 @@ A slow-CI consumer can opt the close into `"async"` mode so the merge wait
 probes once for ~60s (catching an instant merge or an instantly-red required
 check) and then returns `pending` instead of burning ~5 minutes of the host
 tool slot polling a merge that lands after the wait would have expired anyway.
-When a worker returns that `pending` envelope, launch its `nextCommand` as a
+When a close returns that `pending` envelope, launch its `nextCommand` as a
 **background** invocation (host background Bash — its completion re-invokes the
 agent) and move on to the next Story; `single-story-confirm-merge.js` is
 idempotent and owns the whole tail. Do not foreground-poll the merge. The
 default `"sync"` behaviour is unchanged.
+
+**On a multi-Story run, pass `--merge-watch-mode async` on every close.** Close
+sees one Story and cannot see run topology, so it cannot make this call for
+itself — you can. Implementation runs in parallel but the close tail is
+serialized one at a time, and under `sync` each of those closes holds the
+foreground for its full merge wait before the next Story's close may start.
+That is the run's dominant serialized cost, and it is paid per sibling:
+
+```bash
+node <main-repo>/.agents/scripts/single-story-close.js \
+  --story <storyId> --cwd <main-repo> --merge-watch-mode async
+```
+
+The flag overrides `delivery.mergeWatch.mode` for that invocation only — the
+config default stays `"sync"`, which is right for the solo delivery that has no
+sibling waiting behind it. It composes with `--max-wait-seconds`: pass both and
+the explicit bound still wins over the async probe cap. An unrecognized value
+exits non-zero before any phase runs, so a typo cannot silently drop the run
+back onto synchronous waiting. Expect a `pending` envelope from each async
+close — that is the designed ending here, not a failure; background its
+`nextCommand` and move to the next Story's close immediately.
+
+A one-Story run should keep the `sync` default: there is no sibling to unblock,
+and the foreground wait is the cheapest path to `landed`.

package/.agents/workflows/helpers/deliver-story-reference.md

@@ -203,6 +203,29 @@ round cap, proceed / redraft / block — not an independent additional pass
 over the criteria. The M4-B floor holds: one verdict per cluster, the
 cluster count owned by `acceptance-clusters.js` alone.
 
+**One round = N cluster critics → ONE merged verdict → ONE gate call.** The
+clusters are how a round is _authored_; they are not how it is _scored_.
+Concatenate every cluster's records into a single `criteria[]` ordered by
+`index` — exactly one per `acceptance[]` item, under one `storyId`,
+`schemaVersion`, `round` and `commitSha` — and hand that merged file to the
+gate once, with `--expected-criteria` set to the Story's `acceptance[]` count:
+
+```bash
+node <main-repo>/.agents/scripts/acceptance-eval.js \
+  --story <storyId> --verdict <merged-verdict-path> \
+  --expected-criteria <acceptance[] count>
+```
+
+The flag is what makes the merge enforceable: `assertCriteriaCoverage` returns
+early on the `null` default, so **omitting it leaves the guard inert** and a
+single cluster's verdict handed over unmerged scores a fraction of the criteria
+and still reports `proceed`. A length mismatch is rejected before scoring and
+consumes no round. Calling the gate once per cluster instead spends a round
+_per cluster_ — a Story past the cluster ceiling would burn its whole redraft
+budget on cluster arithmetic — and N concurrent calls race the Story-scoped
+round ledger. Full per-round mechanics, including the parallel dispatch and the
+merge shape: [`acceptance-self-eval.md`](acceptance-self-eval.md).
+
 **Critic evidence-share.** When the critic runs a `verify[]`
 command that is byte-identical to a close gate (`lint` / `typecheck`), it
 records the pass into the Story evidence keyspace via `--standalone` so
@@ -318,12 +341,23 @@ The wait probes the checks every poll: a red required check fails fast as
 base is brought up to date within `updateAttempts` tries.
 
 **Async merge-confirm mode (`delivery.mergeWatch.mode: "async"`).** Under
-the default `"sync"` the merge wait runs in the foreground as
-described above. When a consumer's CI routinely takes longer than the host
-tool ceiling (~10 min) can hold a single close invocation, the foreground
-wait almost always expires `pending` after burning ~5 minutes of the slot —
-so `"async"` makes that async confirm a designed mode instead of an expiry
-accident. In async mode the close arms auto-merge, runs one short **~60s
+the default `"sync"` the merge wait runs in the foreground as described above.
+
+**On a multi-Story run, async is the posture — pass `--merge-watch-mode async`
+on every close.** This is not a slow-CI opt-in. Implementation fans out, but
+the close tail is serialized one Story at a time, and under `sync` each close
+holds the foreground for its full merge wait before the next may start; that
+is the run's dominant serialized cost, paid once per sibling. Close sees one
+Story and cannot see run topology, so the orchestrator — which can — makes the
+call per invocation while the config default stays `"sync"`, which is right for
+a solo delivery with no sibling waiting behind it
+([`deliver-reference.md`](deliver-reference.md) § Async merge-confirm mode).
+A slow-CI consumer reaches for the same mode for the separate reason that a
+foreground wait longer than the host tool ceiling (~10 min) almost always
+expires `pending` after burning ~5 minutes of the slot — `"async"` makes that
+confirm a designed ending instead of an expiry accident.
+
+In async mode the close arms auto-merge, runs one short **~60s
 probe window** (long enough to catch an instant merge and, via the
 head-anchored required-check predicate, an instantly-red required check),
 then returns the standard `pending` terminal with a `nextCommand`. When you
@@ -362,6 +396,13 @@ judgment that help text cannot carry.
 - `--max-wait-seconds <n>` — from a headless caller with no host
   tool-invocation ceiling, to keep single-block semantics
   without editing the consumer's config.
+- `--merge-watch-mode <sync|async>` — the per-invocation override of
+  `delivery.mergeWatch.mode`. **Pass `async` on every close of a multi-Story
+  run** (above); leave it off for a solo delivery. It composes with
+  `--max-wait-seconds` — pass both and the explicit bound still wins over the
+  async probe cap. An unrecognized value is refused before any phase runs, so a
+  typo cannot silently drop the run back onto synchronous waiting; the refusal
+  reports a `failed` terminal envelope at `phase: init`, mutating nothing.
 
 ---
 

package/.agents/workflows/helpers/plan-reference.md

@@ -215,12 +215,14 @@ ceilings on `STORY_SHAPE_CEILINGS` in
   against the repo before overriding one (authoring `creates` for a file
   that exists at base is a validator rejection). The persist gates stay
   authoritative: they probe the base branch ref, not the working tree.
-- **Keep `## Spec` near contract-level prose.** Persist emits an
-  **advisory** warning past ~250 words (`SPEC_SOFT_WORD_BUDGET`) — it never
-  fails the persist, but it is the nudge toward a contract-level
+- **Keep `## Spec` near contract-level prose.** **Aim for ~250 words; an
+  advisory warning fires past 350** (`SPEC_SOFT_WORD_BUDGET`). Two numbers,
+  two jobs: ~250 is the authoring target — the nudge toward a contract-level
   Spec (interfaces, invariants, load-bearing constraints; no per-file
-  behavior narration). The hard fail-closed ceiling (~1500 tokens,
-  `spec-spill.js`) is unchanged.
+  behavior narration) — while 350 is the slacker threshold at which persist
+  actually warns, so the warning marks a real outlier instead of ordinary
+  variance. Neither fails the persist. The hard fail-closed ceiling
+  (~1500 tokens, `spec-spill.js`) is unchanged.
 
 A faithfully-filled skeleton — placeholders replaced, pre-resolved entries
 kept, tags valid — passes the persist ticket validators with no
@@ -313,6 +315,14 @@ output shape standalone. When the kill-switch is off
 (`roleScopedAgents: false`) or the host cannot spawn at this depth, fall back
 to a generic sub-agent and hand it the same charter (the `consolidation` /
 `pre-mortem` definitions in [`plan-critic.md`](../../agents/plan-critic.md)).
+**When both critics fire, dispatch them in a single turn.** Consolidation and
+pre-mortem read the same immutable draft, share no write path, and neither
+consumes the other's verdict — the textbook independent fan-out of
+[`parallel-tooling.md`](parallel-tooling.md) Rule 3. Issue both `Agent` calls
+together in one assistant turn rather than awaiting the first verdict before
+spawning the second; serialized critics double the round's wall clock and buy
+nothing, because you fold both verdicts into the same re-author round anyway.
+
 Either way the critic is **maker-blind**: hand it the draft artifacts
 (`stories.json`, and `techspec.md` when present) — never the authoring
 transcript or the reasons the planner believed its own draft is sound. A

package/.agents/workflows/memory-consolidate.md

@@ -0,0 +1,116 @@
+---
+description: >-
+  Attended consolidation pass over this project's agent memory pool — merge
+  duplicates, verify claims against the current tree, prune with operator
+  confirmation, rewrite the index, and stamp the pool so the /plan advisory
+  goes quiet.
+---
+
+# /memory-consolidate [--dry-run]
+
+Agent memory drifts as the codebase moves: a workaround gets fixed, a budget
+changes, two sessions save the same lesson twice. This is the sweep that
+corrects it — and the **only** sanctioned writer of the memory pool outside a
+session's own point-of-write corrections.
+
+**Attended by contract.** Mandrel ships the trigger, the grounding and the
+receipt; it never mutates the operator's personal memory store unsupervised.
+Nothing here runs on a schedule, and no step deletes a file before Gate #2.
+
+## Step 0 — Resolve the pool
+
+`MANDREL_MEMORY_DIR` wins outright; otherwise the pool is
+`~/.claude/projects/<cwd-slug>/memory/`, where `<cwd-slug>` is the absolute
+working directory with every `/` and `.` replaced by `-`.
+
+Report the resolved path and the entry count. **No pool → say so in one line
+and stop** — an absent substrate is a clean no-op, never an error, and never a
+reason to create one.
+
+Each project (and each worktree) has its own pool: consolidate the pool of the
+checkout you are in, and never reach across into another.
+
+## Step 1 — Read the pool
+
+Read `MEMORY.md` and every `*.md` entry. Group them by subject so duplicates
+and supersessions surface together. Do not judge an entry from its index line —
+the index is a pointer, the file is the claim.
+
+## Step 2 — Classify each entry
+
+Verify before you keep. Run `ls` / `grep` against the **current tree** for the
+paths, flags, budgets and gate names an entry cites; re-measure a number rather
+than trusting it (`wc -c` a file whose byte budget is quoted).
+
+| Verdict | Means | Action |
+| --- | --- | --- |
+| `fresh` | Claims re-confirmed against the tree | Leave byte-identical |
+| `correct` | Core lesson holds, a detail has moved | Rewrite the detail in place, keep the entry |
+| `duplicate` | Same lesson as another entry | Merge into the richer one, propose the thinner for pruning |
+| `superseded` | The thing it warns about is fixed or retired | Strike the claim, keep the correction visible |
+| `dead` | Subject no longer exists at all | Propose for pruning |
+
+**A closed issue is the subject of a retrospective, never evidence of
+staleness.** These pools are largely delivery write-ups — "delivered this
+Story", "fixed by that PR". The referenced issue being closed is what the
+memory is *about*: the normal, healthy shape. A retired predecessor of this
+workflow marked exactly that pattern stale, which is why it no longer exists.
+Use `github.owner` / `github.repo` from `.agentrc.json` when an issue
+genuinely needs probing, and treat an unreachable probe as unknown — never
+as dead.
+
+Prefer `correct` over `dead`. A rewritten entry keeps hard-won context; a
+deleted one costs a future session the same mistake.
+
+## Step 3 — Gate #1: present the plan
+
+Write the full proposal to `temp/memory/consolidation-<date>.md` — per entry:
+verdict, evidence, and the exact edit or merge proposed. Summarize inline:
+counts per verdict and the specific prune candidates by name.
+
+**STOP.** Do not edit until the operator responds.
+
+## Step 4 — Apply the edits
+
+On confirmation, apply the `correct`, `duplicate` and `superseded` rewrites.
+Preserve each entry's frontmatter contract (`name`, `description`, `metadata`)
+and keep `[[wiki-link]]` references resolvable — a merge that orphans a link
+has moved the problem, not fixed it.
+
+## Step 5 — Gate #2: confirm every deletion
+
+**Deletion is never implied by Gate #1.** Name each file proposed for pruning
+with its one-line reason, and offer **mark-vs-delete**: marking the claim
+superseded in place is the default and always available. Wait for an explicit
+yes; delete only what the operator names. Silence is not consent.
+
+## Step 6 — Rewrite the index and stamp
+
+Rewrite `MEMORY.md` so it holds exactly one pointer line per surviving entry —
+title, link, hook. It is loaded into context every session, so it carries
+pointers only, never memory content.
+
+Then write the receipt to `.consolidation-stamp.json` in the pool root:
+
+```json
+{ "lastConsolidatedAt": "<ISO-8601 timestamp>" }
+```
+
+The `/plan` Phase 0 advisory reads this file; until it is written, the nudge
+keeps firing. Write it **only** after Gate #2 — the stamp asserts an operator
+reviewed the pass, so writing it early makes it a lie.
+
+Close with counts: entries read, corrected, merged, pruned, and the new total.
+
+## Constraints
+
+- `--dry-run` runs Steps 0–3 and stops: the report is written, nothing mutates.
+- Never delete an entry the operator did not name. Never write the stamp on a
+  dry run or a declined gate.
+- Never invent memories here — this pass corrects and prunes what sessions
+  wrote; new memories come from the sessions that learned them.
+
+## See also
+
+[`/plan`](plan.md) (surfaces the advisory at its Gate #1),
+[`/deliver`](deliver.md) (point-of-write corrections as work lands).

package/.agents/workflows/plan.md

@@ -75,6 +75,9 @@ decision-made-by-default.
 **Gate #1** — STOP to confirm the sharpened plan intent and any
 duplicate-candidate review. Under `--yes`, auto-proceed.
 
+On a truthy `memoryPoolAdvisory.recommend`, name [`/memory-consolidate`](memory-consolidate.md)
+as an operator option, quoting its `reasons[]` — advisory, never invoked here.
+
 On a truthy `deliverLightSuggestion.suggested`, offer — advisory, never an
 automatic reroute — to deliver the seed instead; on confirm, route **in this
 session** into [`helpers/deliver-light.md`](helpers/deliver-light.md), its gate

package/README.md

@@ -182,7 +182,9 @@ dimensions, run model, and how to benchmark a new version.
 ## Contributors
 
 The published `mandrel` package ships three directories — `.agents/`, `bin/`,
-and `lib/` (see the `files` array in [`package.json`](package.json)).
+and `lib/` — plus the single file `docs/CHANGELOG.md`, and excludes the
+`__tests__` subtrees under `.agents/` and `lib/` (see the `files` array in
+[`package.json`](package.json)).
 `.agents/` is the payload `mandrel sync` materializes into a consumer's
 `./.agents/` directory; `bin/mandrel.js` and its `lib/` implementation stay
 inside `node_modules/mandrel/` and back the `npx mandrel …` CLI used
@@ -193,8 +195,8 @@ tooling and is not published.
 Common commands while developing the framework itself:
 
 ```bash
-npm run lint           # markdown + biome
-npm run format         # auto-format
+npm run lint           # biome + markdownlint + repo ratchets + generated-doc drift
+npm run format         # biome format — JavaScript/JSON only, not markdown
 npm test               # framework tests
 npm run test:coverage  # tests with coverage gate
 ```
@@ -226,9 +228,14 @@ against malicious lifecycle scripts in compromised transitive packages
 install scripts for a specific install, run
 `npm install --ignore-scripts=false` for that invocation only.
 
-CRAP and Maintainability gates fire at three sites — close-validation
-(story close), `.husky/pre-push`, and CI (`ci.yml`, push + PR) — against
-the same thresholds from `delivery.quality.*` in `.agentrc.json`.
+CRAP and Maintainability gates fire at four sites — `.husky/pre-commit`
+(`quality-preview.js --staged`, scored on the staged index), `.husky/pre-push`
+(diff-scoped against `origin/main`), close-validation (story close), and CI
+(`ci.yml`, push + PR) — against the same thresholds from
+`delivery.quality.*` in `.agentrc.json`. All four are **blocking**: the
+pre-commit hook fires earliest and refuses the commit on a threshold
+violation, and a green `npm run lint` / `npm test` / `check-baselines.js`
+run does not predict it.
 
 ## License
 

package/docs/CHANGELOG.md

@@ -2,6 +2,70 @@
 
 All notable changes to this project will be documented in this file.
 
+<!--
+  Machine-generated by release-please (`changelog-path` in
+  `release-please-config.json`); only the header above is hand-owned — every
+  release prepends a section beneath it. The three rules below are properties
+  of conventional-changelog-writer's output, not of this repository's prose
+  style: `*` list markers (MD004), a double blank line before each `###`
+  section (MD012), and `*`-wrapped commit subjects that read as unbalanced
+  emphasis (MD037). They would be reintroduced on every release cut, failing
+  lint on release-please's own PR, so they are exempted file-wide rather than
+  hand-repaired. Do not extend this list to cover hand-written prose defects.
+-->
+<!-- markdownlint-disable-file MD004 MD012 MD037 -->
+
+## [2.26.0](https://github.com/dsj1984/mandrel/compare/mandrel-v2.25.0...mandrel-v2.26.0) (2026-08-03)
+
+
+### Added
+
+* /audit-baselines lens: ratchet-tightening audit workflow over the committed baseline surface ([#4903](https://github.com/dsj1984/mandrel/issues/4903)) ([#4913](https://github.com/dsj1984/mandrel/issues/4913)) ([420fb09](https://github.com/dsj1984/mandrel/commit/420fb09df605edfd8ace86e6b2295c7bd6909c3f))
+* **acceptance-eval:** merge parallel cluster verdicts into one gate call per round (refs [#4951](https://github.com/dsj1984/mandrel/issues/4951)) ([#4954](https://github.com/dsj1984/mandrel/issues/4954)) ([56b69dd](https://github.com/dsj1984/mandrel/commit/56b69dde0a5faee418d5f49f4b0c3eab5f682245))
+* **audit:** add the read-only baseline hotspot engine (refs [#4902](https://github.com/dsj1984/mandrel/issues/4902)) ([#4910](https://github.com/dsj1984/mandrel/issues/4910)) ([22a2245](https://github.com/dsj1984/mandrel/commit/22a22458645823705a2cfcf94ea6d1a7acb4546e))
+* **deliver:** accept dash ranges in delivery Story-id lists (refs [#4927](https://github.com/dsj1984/mandrel/issues/4927)) ([#4928](https://github.com/dsj1984/mandrel/issues/4928)) ([9000044](https://github.com/dsj1984/mandrel/commit/90000442d5f578a4895b5950784e2fb8493d316a))
+* **deliver:** add --merge-watch-mode and mandate one-turn fan-out dispatch (refs [#4949](https://github.com/dsj1984/mandrel/issues/4949)) ([#4957](https://github.com/dsj1984/mandrel/issues/4957)) ([fa1c426](https://github.com/dsj1984/mandrel/commit/fa1c426c181cb3696b4ea34166c99719cea877c2))
+* make the coverage instrument measure: widen CI's run to every test tier, configure the gate, and reseed the baseline ([#4922](https://github.com/dsj1984/mandrel/issues/4922)) ([#4931](https://github.com/dsj1984/mandrel/issues/4931)) ([bc94fbf](https://github.com/dsj1984/mandrel/commit/bc94fbf65b41d53098606aff7f7b58dc1679ea5d))
+* repair the inert quality instruments and tighten the ratchets behind the reclaimed slack ([#4923](https://github.com/dsj1984/mandrel/issues/4923)) ([#4933](https://github.com/dsj1984/mandrel/issues/4933)) ([3cc3a30](https://github.com/dsj1984/mandrel/commit/3cc3a30b5e54b3dac78649dd9536ad617086ee0c))
+* retire friction-event.schema.json or wire it, and stop four docs citing a schema nothing compiles ([#4938](https://github.com/dsj1984/mandrel/issues/4938)) ([#4941](https://github.com/dsj1984/mandrel/issues/4941)) ([259f6e6](https://github.com/dsj1984/mandrel/commit/259f6e6cc47db88434b0612aa9327638afb73911))
+* retire the memory-freshness scanner and ship the visible memory-hygiene loop ([#4919](https://github.com/dsj1984/mandrel/issues/4919)) ([#4921](https://github.com/dsj1984/mandrel/issues/4921)) ([54cf87c](https://github.com/dsj1984/mandrel/commit/54cf87c1a4e929681f194fc899decb24636bd5f2))
+
+
+### Fixed
+
+* **baselines:** settle the exit through runAsCli so a piped report is not truncated ([#4932](https://github.com/dsj1984/mandrel/issues/4932)) ([d3670de](https://github.com/dsj1984/mandrel/commit/d3670dec493c2ba60a60e365c1ab809bde58ce43)), closes [#4783](https://github.com/dsj1984/mandrel/issues/4783)
+* close three harness gaps: unbounded CI jobs, divergent runner flags, and inert pretest hooks ([#4936](https://github.com/dsj1984/mandrel/issues/4936)) ([#4942](https://github.com/dsj1984/mandrel/issues/4942)) ([d71040d](https://github.com/dsj1984/mandrel/commit/d71040dc0b558906a42327dcd9bd88ccfc26694c))
+* exclude transpiled-coordinate rows from the 0%-coverage fill, and refuse a baseline that predates provenance stamping ([#4901](https://github.com/dsj1984/mandrel/issues/4901)) ([#4905](https://github.com/dsj1984/mandrel/issues/4905)) ([adc135b](https://github.com/dsj1984/mandrel/commit/adc135b3338550e2911e7139d638be8670eb5bab))
+* **friction:** refuse a quoted --cmd as a usage error and split ENOBUFS from timeout (refs [#4915](https://github.com/dsj1984/mandrel/issues/4915)) ([#4918](https://github.com/dsj1984/mandrel/issues/4918)) ([64989aa](https://github.com/dsj1984/mandrel/commit/64989aa4a5cf746d959b649e760bf93f96f2ead0))
+* **git:** give the git runners an explicit 64 MB stdout ceiling, unblocking Story close ([#4953](https://github.com/dsj1984/mandrel/issues/4953)) ([75160a4](https://github.com/dsj1984/mandrel/commit/75160a491704acc2de127d9135bcfcbc846bdfee))
+* give the raised GitHub write fan-outs transient-retry backoff and single-source their concurrency bound ([#4961](https://github.com/dsj1984/mandrel/issues/4961)) ([#4964](https://github.com/dsj1984/mandrel/issues/4964)) ([459a3f1](https://github.com/dsj1984/mandrel/commit/459a3f12da0280881b757c0855d33d0a89221320))
+* let a dangling-citation anchor count inside a code span (refs [#4906](https://github.com/dsj1984/mandrel/issues/4906)) ([#4911](https://github.com/dsj1984/mandrel/issues/4911)) ([a96ae9f](https://github.com/dsj1984/mandrel/commit/a96ae9fbed4b701ff45d13e268d91e960ebbe595))
+* **lint:** lint docs/ so the gate matches the code-review lens ([#4971](https://github.com/dsj1984/mandrel/issues/4971)) ([b1f4e65](https://github.com/dsj1984/mandrel/commit/b1f4e6551670f145fa6f6e688a16130e8da644d7))
+* make git hooks fire in linked worktrees and stop bypassing pre-push at close ([#4943](https://github.com/dsj1984/mandrel/issues/4943)) ([#4946](https://github.com/dsj1984/mandrel/issues/4946)) ([2ac0597](https://github.com/dsj1984/mandrel/commit/2ac05979241ca2ad061e14fbeda2b71c1b299da7))
+* make the ratchet trend roll-up count what it measures and baseline staleness commit-aware ([#4962](https://github.com/dsj1984/mandrel/issues/4962)) ([#4967](https://github.com/dsj1984/mandrel/issues/4967)) ([fff36bf](https://github.com/dsj1984/mandrel/commit/fff36bf79ab4c75a2f4274c60228ab0942ded65a))
+* **observability:** attribute terminal friction to the emitting CLI ([#4966](https://github.com/dsj1984/mandrel/issues/4966)) ([5aa986a](https://github.com/dsj1984/mandrel/commit/5aa986aaee97e858329ab8f948a1a59db8aeb1c4))
+* **observability:** classify a bare framework-script basename as framework (refs [#4916](https://github.com/dsj1984/mandrel/issues/4916)) ([#4920](https://github.com/dsj1984/mandrel/issues/4920)) ([a8e0bbc](https://github.com/dsj1984/mandrel/commit/a8e0bbc22ad1cc8e0ab3ce117e597c537be96d7c))
+* **plan-persist:** re-carry audit provenance onto dependent Stories (refs [#4935](https://github.com/dsj1984/mandrel/issues/4935)) ([#4939](https://github.com/dsj1984/mandrel/issues/4939)) ([6866b0c](https://github.com/dsj1984/mandrel/commit/6866b0c368de0ff438b565d5049a38273b56ef8c))
+* **quality:** key anonymous CRAP methods by enclosing scope, not escomplex ordinal (refs [#4969](https://github.com/dsj1984/mandrel/issues/4969)) ([#4970](https://github.com/dsj1984/mandrel/issues/4970)) ([ca5da99](https://github.com/dsj1984/mandrel/commit/ca5da9940da57aaf20d34844afa835f6a8b127f0))
+* read prior baseline rows per kind so duplication's diff-scope preservation and epsilon damping actually work ([#4937](https://github.com/dsj1984/mandrel/issues/4937)) ([#4940](https://github.com/dsj1984/mandrel/issues/4940)) ([e7d13b5](https://github.com/dsj1984/mandrel/commit/e7d13b551fb958c38833a7946785929dd17907a3))
+* retune the advisory Spec word budget to 350 and repair its report channel (refs [#4907](https://github.com/dsj1984/mandrel/issues/4907)) ([#4909](https://github.com/dsj1984/mandrel/issues/4909)) ([3c914ba](https://github.com/dsj1984/mandrel/commit/3c914ba9430743f2e5399ded6ac74aee4c4148d1))
+* **scripts:** make --help short-circuit before check-test-temp-hygiene does work ([#4968](https://github.com/dsj1984/mandrel/issues/4968)) ([5e873ac](https://github.com/dsj1984/mandrel/commit/5e873acbc316a37e5a173a94bb2fe52090614493))
+* stop check-baselines silently disabling its compare arm when the base baseline cannot be read ([#4914](https://github.com/dsj1984/mandrel/issues/4914)) ([#4917](https://github.com/dsj1984/mandrel/issues/4917)) ([d62a334](https://github.com/dsj1984/mandrel/commit/d62a334555bccd605bd37bd9d8a544fdcc4e3710))
+* stop the git-cleanup branches phase announcing a DRY RUN before it deletes ([#4898](https://github.com/dsj1984/mandrel/issues/4898)) ([3edad1e](https://github.com/dsj1984/mandrel/commit/3edad1e8c9a8906c1abc5f02e79e7ee4792775b2))
+* **wave-runner:** exempt glob footprints from cross-beat reservation (refs [#4960](https://github.com/dsj1984/mandrel/issues/4960)) ([#4963](https://github.com/dsj1984/mandrel/issues/4963)) ([ca935e2](https://github.com/dsj1984/mandrel/commit/ca935e223f211340c671668a53738de634c12c04))
+* **wave-runner:** reserve in-flight Story footprints in the ready-set guard (refs [#4950](https://github.com/dsj1984/mandrel/issues/4950)) ([#4955](https://github.com/dsj1984/mandrel/issues/4955)) ([7f8ffb3](https://github.com/dsj1984/mandrel/commit/7f8ffb31c523dba01fb9cffe376a5149d92d7c08))
+
+
+### Performance
+
+* **plan:** run the /plan path's independent gathers and write loops concurrently (refs [#4952](https://github.com/dsj1984/mandrel/issues/4952)) ([#4956](https://github.com/dsj1984/mandrel/issues/4956)) ([8c8ec6c](https://github.com/dsj1984/mandrel/commit/8c8ec6c1134f23fbf8214d138438b3febc6014c8))
+
+
+### Changed
+
+* **baselines:** migrate the duplication CLI onto refreshBaseline (refs [#4944](https://github.com/dsj1984/mandrel/issues/4944)) ([#4945](https://github.com/dsj1984/mandrel/issues/4945)) ([38e43d2](https://github.com/dsj1984/mandrel/commit/38e43d2362031cfaebbfeeaa8bec75c3c5ecf56b))
+* burn down the cross-gate hotspot clusters and remove the shipped seams nothing calls ([#4926](https://github.com/dsj1984/mandrel/issues/4926)) ([#4934](https://github.com/dsj1984/mandrel/issues/4934)) ([970c7f8](https://github.com/dsj1984/mandrel/commit/970c7f88f5d198b685f90ede5358beea5b2896ce))
+
 ## [2.25.0](https://github.com/dsj1984/mandrel/compare/mandrel-v2.24.0...mandrel-v2.25.0) (2026-08-01)
 
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "mandrel",
-  "version": "2.25.0",
+  "version": "2.26.0",
   "description": "Claude Code-first opinionated workflow framework: instructions, skills, rules, and SDLC workflows that govern AI coding assistants.",
   "files": [
     ".agents/",
@@ -15,9 +15,7 @@
     "provenance": true
   },
   "scripts": {
-    "pretest": "node .agents/scripts/test-wrapper.js && node .agents/scripts/validate-skills.js",
-    "pretest:quick": "node .agents/scripts/test-wrapper.js",
-    "pretest:integration": "node .agents/scripts/test-wrapper.js",
+    "//test-preflight": "There are deliberately NO pretest* scripts (Story #4936). .npmrc sets ignore-scripts=true as supply-chain defence (CWE-1357), and that suppresses every pre*/post* lifecycle script for `npm run` too — so a pretest entry here would look like a gate and never execute. Each tier's preflight is declared once in .agents/scripts/lib/test-runner-contract.js and invoked by the runners themselves, which is why it fires for `npm test`, `npm run test:coverage` and a bare `node .agents/scripts/run-tests.js` alike.",
     "test": "node .agents/scripts/run-tests.js",
     "test:quick": "node .agents/scripts/run-tests.js --tier quick",
     "test:integration": "node .agents/scripts/run-tests.js --tier integration",
@@ -28,6 +26,7 @@
     "test:coverage": "node .agents/scripts/run-coverage.js",
     "coverage:check": "node .agents/scripts/check-baselines.js --gate coverage",
     "coverage:update": "node .agents/scripts/update-coverage-baseline.js",
+    "coverage:reanchor": "node .agents/scripts/update-coverage-baseline.js --full-scope",
     "baseline:agents-loc": "node baselines/agents-loc-baseline.mjs",
     "lint:md": "markdownlint-cli2 \".agents/**/*.md\" \"*.md\" \"!node_modules/**\" \"!.worktrees/**\"",
     "lint": "node .agents/scripts/run-lint.js && npm run docs:check",
@@ -39,13 +38,19 @@
     "format:check": "biome ci .",
     "maintainability:check": "node .agents/scripts/check-baselines.js --gate maintainability",
     "maintainability:update": "node .agents/scripts/update-maintainability-baseline.js",
+    "maintainability:reanchor": "node .agents/scripts/update-maintainability-baseline.js --full-scope",
     "check:arch": "node .agents/scripts/check-arch-cycles.js",
     "check:context-budget": "node .agents/scripts/check-context-budget.js",
+    "check:cyclomatic": "node .agents/scripts/check-cyclomatic.js",
+    "cyclomatic:update": "node .agents/scripts/check-cyclomatic.js --update",
+    "check:schema-refs": "node .agents/scripts/check-schema-references.js",
     "check:workflow-citations": "node .agents/scripts/check-workflow-citations.js",
     "crap:check": "node .agents/scripts/check-baselines.js --gate crap",
     "crap:update": "node .agents/scripts/update-crap-baseline.js",
+    "crap:reanchor": "node .agents/scripts/update-crap-baseline.js --full-scope",
     "duplication:check": "node .agents/scripts/check-baselines.js --gate duplication",
     "duplication:update": "node .agents/scripts/update-duplication-baseline.js",
+    "duplication:reanchor": "node .agents/scripts/update-duplication-baseline.js --full-scope",
     "quality:preview": "node .agents/scripts/check-dead-exports.js && node .agents/scripts/quality-preview.js",
     "quality:watch": "node .agents/scripts/quality-watch.js",
     "sync:commands": "node bin/mandrel.js sync-commands",

package/.agents/schemas/friction-event.schema.json

@@ -1,56 +0,0 @@
-{
-  "$schema": "http://json-schema.org/draft-07/schema#",
-  "title": "FrictionEvent",
-  "type": "object",
-  "additionalProperties": false,
-  "properties": {
-    "eventId": {
-      "type": "string",
-      "format": "uuid"
-    },
-    "timestamp": {
-      "type": "string",
-      "format": "date-time"
-    },
-    "sprintId": {
-      "type": "string"
-    },
-    "category": {
-      "type": "string",
-      "enum": [
-        "Prompt Ambiguity",
-        "Missing Skill",
-        "Incorrect Persona",
-        "Tool Limitation",
-        "Execution Error"
-      ]
-    },
-    "source": {
-      "type": "object",
-      "additionalProperties": false,
-      "properties": {
-        "tool": {
-          "type": "string"
-        },
-        "command": {
-          "type": "string"
-        }
-      }
-    },
-    "details": {
-      "type": "string",
-      "description": "Specific error message or observation."
-    },
-    "context": {
-      "type": "object",
-      "additionalProperties": false,
-      "properties": {
-        "protocolFile": {
-          "type": "string",
-          "description": "Path to relevant protocol file, e.g., '.agents/skills/core/tdd.md'"
-        }
-      }
-    }
-  },
-  "required": ["eventId", "timestamp", "sprintId", "category", "details"]
-}