dev-loops 0.4.0 → 0.5.0

package/.claude/skills/docs/retrospective-checkpoint-contract.md

@@ -93,17 +93,48 @@ When `workflow.requireRetrospectiveGate` is enabled in `.devloops` at repo root,
 - has `state: "complete"`
 - includes a `behavioralReview` with `mergeApproved: true`, `followedWorkingAgreement` (boolean), `gateQualityAcceptable` (boolean), and `drifts` (array)
 - includes `mergeRecommendation` (non-empty string)
+- **(developer mode only)** includes `behavioralReview.internalToolingOnly: true` (boolean) and `behavioralReview.rawCallViolations: []` (empty array) — see **Internal-tooling-only rule** below
 
-The enforcement function is `evaluateRetrospectiveMergeApproval(checkpoint)` in `packages/core/src/loop/pr-gate-coordination.mjs`, called from `evaluatePrGateCoordination` at each merge-ready boundary.
+The enforcement function is `evaluateRetrospectiveMergeApproval(checkpoint, { developerMode })` in `packages/core/src/loop/pr-gate-coordination.mjs`, called from `evaluatePrGateCoordination` at each merge-ready boundary.
+
+### Internal-tooling-only rule (issue #982) — DEVELOPER MODE ONLY, opt-in
+
+This is a **developer-mode** retro step: it enforces the dev-loops maintainers' own dogfooding discipline and is **opt-in** via `workflow.requireRetrospectiveInternalTooling` (**default OFF**). **Consumers of the dev-loops extension are never blocked by it** — a consumer may legitimately use raw `gh`/`python`/`node -e` in their own workflow, so when the flag is OFF (the consumer default) the `internalToolingOnly`/`rawCallViolations` fields are neither required nor enforced, and a complete, merge-approved checkpoint passes exactly as it would without these fields. The check only requires and enforces them when `workflow.requireRetrospectiveInternalTooling: true` (the dev-loops repo opts in via its own repo-root `.devloops`).
+
+When developer mode is ON, the loop's own execution must use internal dev-loops tooling only. **Agent-level top-level raw calls are all the same breach and all block the merge gate:**
+
+- `gh ...` (including `gh api`, `gh ... --jq`) — use `gate capture-threads`, `gate reply-resolve`, `dev-loops loop info`, `queue …`, or a `node scripts/...` wrapper instead
+- `python` / `python3` — parse JSON with the tool's `--jq`/built-in output, not an inline Python one-liner
+- `node -e` / `node --eval` — inline eval to read/parse tool output is a violation; `node scripts/foo.mjs` is **not**
+
+**Allowed (NOT violations):** dev-loops subcommands, and `node scripts/*.mjs` invocations — those scripts legitimately call `gh`/GraphQL internally; that is the tooling. The rule targets the agent's own top-level shell calls, not a script's internals.
+
+The retrospective author records the result in the checkpoint:
+- `behavioralReview.internalToolingOnly` (boolean) — `true` only when the run used internal tooling throughout
+- `behavioralReview.rawCallViolations` (array of strings) — one entry per agent-level raw call; empty when clean
+
+**Fail-closed (developer mode only):** when `requireRetrospectiveInternalTooling` is ON, the merge gate blocks (`retrospective_gate_pending`) if `internalToolingOnly` is not `true` **OR** `rawCallViolations` is missing/non-empty. `state: "complete"` alone is **not** sufficient — a clean tooling record is also required. When the flag is OFF (consumer default) this check is inert: it never blocks, even for a checkpoint that omits the fields or records a violation.
+
+**Back-compat:** an OLD checkpoint that omits `internalToolingOnly`/`rawCallViolations` only fails closed in developer mode; with the flag OFF (consumer default) it passes unchanged. Re-record the retrospective with the new fields to clear a developer-mode block.
+
+**Write-op allowlist (verifier only):** `gh pr merge`, `gh pr ready`, `gh issue create`, `gh issue edit` have no internal wrapper today. The deterministic verifier (below) records these as `allowedWriteOps` rather than violations so the gate is not blocked forever on an unavoidable gap. Close the gap with a wrapper to remove an entry. The gate itself still fails on any non-empty `rawCallViolations` the author records.
+
+### Deterministic verifier
+
+`node scripts/loop/check-retro-tooling.mjs [--transcript <path>] [--json]` reads a newline-delimited transcript of the shell commands the agent ran (one top-level command per line, via `--transcript` or stdin) and reports agent-level raw `gh`/`python`/`python3`/`node -e`/`node --eval` calls. Use it to derive the `rawCallViolations` array the retrospective author records. Exit code `1` when violations are found, `0` when clean. The pure `analyzeTranscript(transcript)` export returns `{ violations, allowedWriteOps, internalToolingOnly }`.
+
+Matching rules: a tool name at the start of a command segment (start of line, or after `&&`/`||`/`|`/`;`); `node` is a violation only with `-e`/`--eval`. Before classifying, the verifier normalizes the segment head — it strips leading `NAME=value` env-assignment prefixes (`GH_TOKEN=x gh api`), strips a leading wrapper binary from `{sudo, env, xargs, time, nice, command}` (`sudo gh api`, `xargs gh api`), and reduces a path-prefixed binary to its basename (`./node_modules/.bin/gh`, `/usr/bin/python3`) — so the common prefixed/wrapped raw-call forms are caught. Known limitation: it does NOT fully parse shell quoting/substitution. A separator inside a quoted argument can over-report; deeply obfuscated calls (command substitution `$(...)`, aliases, `eval`) may evade it — prefer single-line, single-purpose commands in transcripts.
 
 ### Merge gate states
 
 | Retrospective state | Merge gate result |
 |---|---|
 | No checkpoint file | Blocked: `retrospective_gate_pending` |
-| `state: "complete"` with `mergeApproved: true` and valid fields | Allowed: proceeds to `FINAL_APPROVAL_READY` |
+| `state: "complete"` with `mergeApproved: true` and valid base fields (consumer default; internal-tooling check OFF) | Allowed: proceeds to `FINAL_APPROVAL_READY` |
 | `state: "complete"` without `mergeApproved: true` | Blocked |
-| Missing required fields (`followedWorkingAgreement`, `gateQualityAcceptable`, `drifts`, `mergeRecommendation`) | Blocked |
+| **Developer mode ON** + `internalToolingOnly` not `true`, or `rawCallViolations` missing/non-empty | Blocked |
+| **Developer mode OFF** (consumer default) + missing `internalToolingOnly`/`rawCallViolations` or a recorded violation | Allowed (check is inert) |
+| Missing base required fields (`followedWorkingAgreement`, `gateQualityAcceptable`, `drifts`, `mergeRecommendation`) | Blocked |
 | `state: "skipped"` or `state: "required"` | Blocked |
 
 ### Configuration
@@ -111,8 +142,9 @@ The enforcement function is `evaluateRetrospectiveMergeApproval(checkpoint)` in
 ```yaml
 # .devloops at repo root
 workflow:
-  requireRetrospective: true        # startup/resume gate
-  requireRetrospectiveGate: true    # merge gate after pre_approval_gate
+  requireRetrospective: true                  # startup/resume gate
+  requireRetrospectiveGate: true               # merge gate after pre_approval_gate
+  requireRetrospectiveInternalTooling: true    # developer-mode internal-tooling-only check (default OFF; consumers leave unset)
 ```
 
 ## Durable artifact format
@@ -130,11 +162,25 @@ The checkpoint file is written by `.pi/extensions/dev-loop-behavioral-review.ts`
 
 ### After retrospective is done (written by operator or skill)
 
+A minimal completion clears the startup gate. To also clear the **merge gate**
+(`requireRetrospectiveGate`), include the `behavioralReview`, `gateQuality`, and
+`mergeRecommendation` fields:
+
 ```json
 {
   "state": "complete",
   "completedAt": "2026-05-29T16:30:00.000Z",
-  "notes": "Loop followed working agreement; minor drift on thread resolution."
+  "notes": "Loop followed working agreement; minor drift on thread resolution.",
+  "gateQuality": "Real gates with concrete findings and follow-through.",
+  "mergeRecommendation": "Merge approved — all gates passed clean.",
+  "behavioralReview": {
+    "mergeApproved": true,
+    "followedWorkingAgreement": true,
+    "gateQualityAcceptable": true,
+    "drifts": [],
+    "internalToolingOnly": true,
+    "rawCallViolations": []
+  }
 }
 ```
 
@@ -153,7 +199,9 @@ The checkpoint file is written by `.pi/extensions/dev-loop-behavioral-review.ts`
 | Artifact | Location |
 |---|---|
 | Checkpoint state machine | `packages/core/src/loop/retrospective-checkpoint.mjs` (internal core module; not part of the public package exports surface) |
-| Tests | `packages/core/test/retrospective-checkpoint.test.mjs` |
+| Merge gate evaluator | `packages/core/src/loop/pr-gate-coordination.mjs` — `evaluateRetrospectiveMergeApproval` |
+| Internal-tooling verifier | `scripts/loop/check-retro-tooling.mjs` (+ `test/loop/check-retro-tooling.test.mjs`) |
+| Tests | `packages/core/test/retrospective-checkpoint.test.mjs`, `packages/core/test/pr-gate-coordination.test.mjs` |
 | Extension (writes required marker, fires review prompt) | `.pi/extensions/dev-loop-behavioral-review.ts` |
 | Checkpoint file | `.pi/dev-loop-retrospective-checkpoint.json` |
 | AGENTS.md repo contract | [Agent Instructions](../../AGENTS.md) — concise repo contract and working rules |

package/.claude/skills/docs/spike-mode-contract.md

@@ -0,0 +1,237 @@
+# Spike-mode contract
+
+A spike is a time-boxed, exploratory dev-loop run started from a local question with no GitHub issue. Its deliverable is a findings document. When the exploration reaches a recommendation, the operator concludes it with one of two exits: **discard** (drop it with zero tracker artifacts) or **graduate** (emit a #947-style local-first plan file that enters the existing plan→PR promotion path).
+
+This doc is the canonical operator sequence for spike mode and carries a worked example showing both exits. The spike artifact lives outside the tracker; it is exempt from the production-gate ceremony at entry and runs under a relaxed gate profile (`gates.spike`). The [Artifact Authority Contract](artifact-authority-contract.md) owns the local-planning model a graduated spike feeds into.
+
+The spike track shipped across three phases: P1 (#964) the `--spike` intake state machine, P2 (#965) the relaxed gate profile and the discard/graduate exits, P3 (#966) this doc.
+
+## Shipped surfaces
+
+| Stage | Helper script | Pure logic |
+|---|---|---|
+| Validate | `scripts/refine/validate-spike-file.mjs` | `validateSpikeFile` / `validateSpikeExplorationSections` (in the helper script) |
+| Start | `scripts/loop/resolve-dev-loop-startup.mjs --spike <path>` | `buildSpikeInput` (in the script) + `evaluateSpikeIntakeState` (`@dev-loops/core/loop/spike-intake-contract`) |
+| Exit | `scripts/refine/exit-spike.mjs` | `evaluateSpikeExit` / `buildGraduatedPlanBody` (`@dev-loops/core/loop/spike-exit-contract`) |
+
+The spike findings artifact carries four authoring sections (`scripts/refine/validate-spike-file.mjs`, `SPIKE_FILE_BASE_SECTIONS`):
+
+- `## Question` — what the spike investigates
+- `## Approach` — how the exploration was carried out
+- `## Findings` — the evidence gathered
+- `## Recommendation` — the exit marker
+
+The first three (`SPIKE_FILE_EXPLORATION_SECTIONS`) are the exploration scaffold required to **start** a spike. `Recommendation` (`SPIKE_FILE_EXIT_MARKER_SECTION`) is filled in during the spike; once it is present and non-empty, the intake state flips to ready-for-exit.
+
+## Intake states
+
+`evaluateSpikeIntakeState` (`packages/core/src/loop/spike-intake-contract.mjs`) classifies the artifact into one of three `SPIKE_INTAKE_STATE` values from two facts — whether the exploration scaffold is valid, and whether a Recommendation is present:
+
+| State | Meaning |
+|---|---|
+| `spike_in_progress` | Scaffold (Question/Approach/Findings) valid; no Recommendation yet |
+| `spike_ready_for_exit` | Scaffold valid and a non-empty Recommendation is present; an exit decision can be made |
+| `ambiguous_fail_closed` | Scaffold invalid (a missing or empty base section); fail closed, no routing |
+
+## Operator sequence
+
+### 1. Author the exploration scaffold
+
+Write a findings artifact carrying `## Question`, `## Approach`, and `## Findings` with non-empty bodies. `## Recommendation` is left for the spike to fill in. Check the full set of base sections with the validator:
+
+```
+node scripts/refine/validate-spike-file.mjs --input <spike.md> --json
+```
+
+The payload is `{ checker: "validate-spike-file", ok, errors }`, with a distinct `missing_*` code per absent or empty section (`missing_question`, `missing_approach`, `missing_findings`, `missing_recommendation`). An in-progress spike reports `missing_recommendation` here and is still startable, because startup gates on the exploration scaffold (`validateSpikeExplorationSections`), which omits the Recommendation.
+
+### 2. Start the spike
+
+Hand the artifact to startup with `--spike` (mutually exclusive with `--issue`, `--pr`, `--input`, `--plan-file`):
+
+```
+node scripts/loop/resolve-dev-loop-startup.mjs --spike <spike.md>
+```
+
+`buildSpikeInput` (`scripts/loop/resolve-dev-loop-startup.mjs`) reads the file read-only — no tracker mutation, no GitHub call, no issue or PR number. It requires the exploration scaffold; a missing/unreadable file, or one failing `validateSpikeExplorationSections`, throws and startup fails closed (exit 1, no readiness bundle). On success it builds a `local_phase` startup input with the resolved spike path as the target `phase` and threads `spikeIntakeState` onto the output. A spike with no Recommendation reports `spike_in_progress`. The spike path is exempt from the worktree-isolation guard because there is no issue to key a worktree on (`planFileExempt: true`).
+
+### 3. Run under the relaxed gate profile
+
+A spike runs under `gates.spike`, a lighter angle set than the production draft → pre-approval gates. See [Relaxed gate profile](#relaxed-gate-profile) below.
+
+### 4. Reach a Recommendation, then exit
+
+When the exploration concludes, fill in `## Recommendation`. The intake state is now `spike_ready_for_exit`, and `exit-spike.mjs` can conclude the spike. It recomputes the intake facts (`validateSpikeExplorationSections` for the scaffold, `extractSection` for the Recommendation), runs `evaluateSpikeExit`, and routes to one of two dispositions.
+
+```
+node scripts/refine/exit-spike.mjs --spike-file <spike.md> --disposition <discard|graduate> [--plan-file <path>] [--json]
+```
+
+`evaluateSpikeExit` (`packages/core/src/loop/spike-exit-contract.mjs`) is eligible only from `spike_ready_for_exit`. From any other state it fails closed with `not_ready_for_exit`; an unrecognized disposition fails closed with `unknown_disposition`. On a fail-closed path the CLI makes zero tracker mutation, writes no plan file, and exits 1.
+
+## Exit dispositions
+
+The two dispositions are `SPIKE_EXIT_DISPOSITION.DISCARD` (`"discard"`) and `SPIKE_EXIT_DISPOSITION.GRADUATE` (`"graduate"`).
+
+### discard
+
+The recommendation is "don't pursue". The spike is dropped with zero tracker artifacts; the findings document on disk is the whole record. The CLI writes nothing and creates no GitHub artifact. `--plan-file` is irrelevant for a discard.
+
+### graduate
+
+The recommendation is "pursue this". `buildGraduatedPlanBody` builds a #947-consumable plan-file body from the spike's `Question`, `Approach`, `Findings`, and `Recommendation`, and `exit-spike.mjs` writes it to the required `--plan-file` path. The mapping:
+
+- `Question` + `Approach` become the Objective's context.
+- `Recommendation` becomes the In-scope work.
+- `Findings` are recorded as supporting evidence.
+- A fixed Explicit non-goals block keeps the plan from re-opening the concluded exploration.
+
+The emitted body carries the four plan-file base sections (`## Status`, `## Objective`, `## In scope`, `## Explicit non-goals` — see [Plan-file Contract](plan-file-contract.md)), so it passes `validatePlanFile` and enters the local-first plan→PR promotion path (#952) unchanged: refine it, hold the local human-review checkpoint, then `promote-plan.mjs`. Graduation is idempotent — `buildGraduatedPlanBody` is pure, so re-running reproduces the same plan file. `buildGraduatedPlanBody` fails closed (throws) on an empty required section, so a graduate exit cannot emit a plan the validator would reject. `--plan-file` is required for graduate (the CLI never guesses an output path).
+
+## Relaxed gate profile
+
+A spike's deliverable is a findings doc, so it does not carry the full production draft → pre-approval → Copilot gate set. `gates.spike` is a relaxed profile, shipped in `packages/core/src/config/extension-defaults.yaml`:
+
+```yaml
+gates:
+  spike:
+    angles:
+      - scope
+      - docs
+    excludeAngles: []
+    required: false
+    requireCi: false
+    mandatoryAngles: []
+```
+
+| Knob | Spike value | Production draft/pre-approval | Why for a spike |
+|---|---|---|---|
+| `angles` | `scope`, `docs` | the full production angle set | A findings doc needs a scope check and a docs check, so the angle set is small and docs-first |
+| `required` | `false` | `true` | The spike's record is the findings doc, so the gate is advisory and the loop proceeds without a passing verdict |
+| `requireCi` | `false` | `true` (pre-approval always requires CI) | A spike produces no production code to run CI against |
+
+`gates.spike` resolves through the same config-merge layering and the same `resolveGateConfig(config, "spike")` path as `draft` and `preApproval` (`packages/core/src/config/config.mjs`) — no new strategy→knob resolver. It is `optional()` in the schema (`packages/core/src/config/config.mjs`, `schemas/dev-loop-config.schema.json`) and absent for non-spike work, so production gates are unaffected. A repo `.devloops` can override any of these knobs.
+
+## Worked example
+
+One spike, `spike-cache.md`, from question to a Recommendation, shown through both exits.
+
+### Authored scaffold
+
+```markdown
+# Spike: response cache
+
+## Question
+
+Would an in-process LRU cache on the metadata fetch measurably cut p95 latency?
+
+## Approach
+
+Add a throwaway `@lru_cache` wrapper behind a flag, replay a captured request
+trace, and compare p95 with and without the cache.
+
+## Findings
+
+The trace replay shows p95 dropping from 240ms to 90ms with the cache, with a
+~3MB steady-state memory cost and no correctness regressions on the replay set.
+```
+
+Validation reports the spike is still in progress — the Recommendation is not yet written, so the full-section validator flags it while the scaffold itself is complete:
+
+```
+$ node scripts/refine/validate-spike-file.mjs --input spike-cache.md --json
+{ "checker": "validate-spike-file", "ok": false, "errors": [ { "code": "missing_recommendation", "message": "..." } ] }
+```
+
+Startup accepts it anyway, because the entry gate is the exploration scaffold:
+
+```
+$ node scripts/loop/resolve-dev-loop-startup.mjs --spike spike-cache.md
+# ... startup output carries spikeIntakeState: "spike_in_progress"
+```
+
+### Recommendation reached
+
+The operator fills in the exit marker:
+
+```markdown
+## Recommendation
+
+Adopt the cache behind a default-on flag; the latency win is large and the
+memory cost is small. Ship it as a follow-up plan.
+```
+
+The intake state is now `spike_ready_for_exit`, so the spike is exitable.
+
+### Exit A — discard (leaves nothing behind)
+
+Had the findings come out the other way — say the latency win was within noise — the recommendation would be "don't pursue", and the operator discards:
+
+```
+$ node scripts/refine/exit-spike.mjs --spike-file spike-cache.md --disposition discard --json
+{ "ok": true, "action": "discard", "spikeFile": "/abs/spike-cache.md" }
+```
+
+No plan file is written, no GitHub artifact is created, and no tracker entry is made. The `spike-cache.md` findings doc on disk is the entire record of the exploration.
+
+### Exit B — graduate (emits a plan file)
+
+With the "adopt it" recommendation above, the operator graduates the spike into a plan file:
+
+```
+$ node scripts/refine/exit-spike.mjs --spike-file spike-cache.md --disposition graduate --plan-file docs/phases/phase-43.md --json
+{ "ok": true, "action": "graduate", "spikeFile": "/abs/spike-cache.md", "planFile": "/abs/docs/phases/phase-43.md" }
+```
+
+`docs/phases/phase-43.md` is written with the spike's sections mapped onto the plan-file base sections:
+
+```markdown
+# Graduated spike plan
+
+## Status
+
+Draft (graduated from a spike). Needs refinement before promotion.
+
+## Objective
+
+Act on the spike's recommendation. The spike asked: Would an in-process LRU
+cache on the metadata fetch measurably cut p95 latency?
+
+Approach explored:
+
+Add a throwaway `@lru_cache` wrapper behind a flag, replay a captured request
+trace, and compare p95 with and without the cache.
+
+## In scope
+
+Adopt the cache behind a default-on flag; the latency win is large and the
+memory cost is small. Ship it as a follow-up plan.
+
+Supporting findings from the spike:
+
+The trace replay shows p95 dropping from 240ms to 90ms with the cache, with a
+~3MB steady-state memory cost and no correctness regressions on the replay set.
+
+## Explicit non-goals
+
+- Re-running the spike's exploration; that question is concluded.
+- Work beyond the recommendation above.
+```
+
+This plan file passes `validatePlanFile`, so it enters the local-planning flow as a new plan needing refinement: refine it, hold the local human-review checkpoint, then `promote-plan.mjs` opens the draft PR (see the [Local-Planning Worked Example](local-planning-worked-example.md)). Re-running the graduate exit reproduces the same plan body, because `buildGraduatedPlanBody` is pure.
+
+## Relationship to other docs
+
+| Doc | Relationship |
+|---|---|
+| [Artifact Authority Contract](artifact-authority-contract.md) | Canonical artifact model; a graduated spike feeds the local-planning tier |
+| [Local-Planning Flow](local-planning-flow.md) | The operator sequence a graduated plan file then follows |
+| [Local-Planning Worked Example](local-planning-worked-example.md) | One plan file through promotion, where a graduated spike lands |
+| [Plan-file Contract](plan-file-contract.md) | Plan-file format and base sections the graduated body satisfies |
+| [Public Dev Loop Contract](public-dev-loop-contract.md) | Canonical routing contract; `--spike` is one startup input |
+
+## Non-goals
+
+- Changing spike intake, exit, or gate behavior (this is a docs-only surface).
+- Defining a tracker-backed spike — a spike is local by construction.
+- A discard that records a tracker artifact — discard is zero-mutation by contract.

package/CHANGELOG.md

@@ -4,6 +4,49 @@ All notable changes to this project will be documented in this file.
 
 ## Unreleased
 
+## 0.5.0 - 2026-06-28
+
+### Added
+
+- **Conflict-free (mergeable) is a required gate precondition + deterministic auto-resolve (#980).** A PR that conflicts with its base gets no `pull_request` CI run (GitHub can't compute the merge ref), so the gate silently stalls green-less. Mergeability is now a required precondition checked at every gate (draft + pre-approval) and again before merge: the detect layer fetches `gh pr view --json mergeable,mergeStateStatus` and the evaluator (`evaluatePrGateCoordination`) blocks a `CONFLICTING`/`DIRTY`/`BEHIND` PR (`gateBoundary: conflict_resolution`, `nextAction: resolve_merge_conflicts`). Because GitHub computes `mergeable` asynchronously, a freshly-pushed head briefly reads `UNKNOWN`; `fetchPrFactsWithSettledMergeable` re-polls a bounded number of times and, if it never settles, the gate fails closed to a recheck (`nextAction: wait_for_ci`) — an unsettled merge state is never treated as clean. A new conservative helper `scripts/loop/resolve-pr-conflicts.mjs` merges `origin/<base>` into the PR branch and resolves ONLY the safe additive case — a `CHANGELOG.md` conflict where both sides only ADD list/section entries (keep BOTH sides, in order) — then runs `npm run test:docs` and (with `--push`) pushes; ANY other conflicted path, or a non-additive CHANGELOG edit, FAILS CLOSED naming the conflicted paths (no general conflict-resolution engine). `loop info` surfaces a `Mergeable:` line (mergeStateStatus included) so a conflict is diagnosed immediately rather than mistaken for missing CI. `node:test` coverage: CONFLICTING/UNKNOWN/MERGEABLE evaluator cases, the bounded UNKNOWN re-poll, and real-git fixtures for the additive-CHANGELOG resolve, the non-CHANGELOG fail-closed (path named), the non-additive-CHANGELOG fail-closed, and a clean merge. Docs: [merge-preconditions](skills/docs/merge-preconditions.md) documents the conflict-check-and-resolve step (before CI/Copilot and before merge).
+
+### Changed
+
+- **Pages decks auto-deploy on push to `main`** (#941, closes #940; supersedes the operator-gated deploy from #930). Pages is enabled (Source: GitHub Actions), so the deploy job no longer needs the `workflow_dispatch` gate `#930` shipped while Pages was off. `.github/workflows/pages.yml` now scopes the deploy job to the main ref (`if: github.ref == 'refs/heads/main'`) instead of `if: github.event_name == 'workflow_dispatch'`, so a merge to `main` publishes the assembled `site/` automatically. A manual `workflow_dispatch` still deploys, but only when run against `main`; dispatching any other branch/tag builds the artifact and skips deploy. The workflow header comment, the deploy-job inline comment, and `docs/presentations/README.md` were updated to describe auto-deploy on merge in place of the one-time operator "Run workflow" step.
+- **Public materials reframed around the pull principle** (#991). The handoff-framing passages across the four public pieces now lead with the Kanban-style insight that the next step is always known: the authoritative resolver/state graph computes the next action for any change at any time, `loop info` surfaces it, and the board mirrors it, so whoever is free — agent or human — pulls the next bounded step from the visible state. The handoff becomes optional, at most additive: when one happens it adds a note and stays a recorded decision, and it is no longer a load-bearing blocking step, which is what dissolves the waiting-for-handoff. Edited in sync (`.md` + `.html`): the intro article's lede, "Model-agnostic by construction", and "Where to go deeper" pointer (`docs/articles/introducing-dev-loops.md`/`.html`); the deep-dive article's Part 1 — the two-part intro, "The one idea" (retitled "the next step is always known"), "What a known next step buys you", "Why a state graph beats a prompt" (recast as the thing that makes the next step always-known and pullable), and the close (`docs/articles/dev-loops-deep-dive.md`/`.html`); the intro deck's hero, the idea slide (retitled "The Next Step Is Always Known"), the model-agnostic slide, and the companion-pieces note (`docs/presentations/introducing-dev-loops.html`); and the deep-dive deck's hero, core-idea slide (retitled "The Next Step Is Always Known"), why-graphs slide, and close (`docs/presentations/dev-loops-deep-dive.html`). The "still a recorded decision when it happens" point is kept, now subordinate to the pull framing. Diagrams and the Diagram 1–8 caption numbering are unchanged; the deck section ids and 390px mobile-fit are intact (deep-dive deck Playwright spec green). The A/B-contrast deslop step (`docs/ab-contrast-deslop-step.md`) was applied to all rewritten prose — the pull idea is stated as plain declaratives with no "pull, not push" / "rather than" / "instead of" antithesis.
+- **Public materials consolidated to ONE deep dive per format** (#978). The two deep-dive articles (`eliminating-coordination-delay` + `make-the-waiting-visible`) merged into one `docs/articles/dev-loops-deep-dive.md`/`.html`, and the two deep-dive decks (`applied-dev-loops.html` + `process-observability.html`) merged into one `docs/presentations/dev-loops-deep-dive.html`. Each merged piece runs in two parts — Part 1 eliminating coordination delay (explicit handoffs, fan-out/fan-in review, mid-flight steering, the state graph), then Part 2 make the waiting visible (interrupt cost, handoff discovery, the git blind spot, the four fields, the measurement loop, grounding in real mechanisms) — under one title/hero, one intro, and one close. The deep-dive article reuses the article design system (glass cards, inline-flow nodes, inline SVG diagrams) and the deck reuses the deck design system with stable per-slide section ids and mobile-fit layout. The old four published files are removed (and the orphaned `*-notes.md` review records for the two removed articles). `scripts/pages/build-site.mjs` `ARTICLES`/`DECKS`/`NAV_LINKS` collapse to the single deep-dive in each format (nav labels "Deep dive" / "Deep dive (deck)"); the article and deck share the source basename `dev-loops-deep-dive.html` under different `docs/` dirs, so the deck publishes as `dev-loops-deep-dive-deck.html` to avoid clobbering the article in `site/`. The Pages nav is now Intro article (landing) + Deep-dive article + Intro deck + Deep-dive deck. The two old deck Playwright specs/configs (`applied-deck`, `observability-deck`) and the `test:playwright:deck`/`:obs-deck` scripts are replaced by one `deep-dive-deck.spec.mjs` + `playwright.dev-loops-deep-dive.config.mjs` + `test:playwright:deep-dive` (section-ids, 390px mobile fit/no-clip, CSP, guard-the-guard). The intro article's "Where to go deeper", `docs/index.md`, and `docs/presentations/README.md` point at the consolidated pages. The A/B-contrast deslop step was applied to the merged prose (new part-divider headlines and bridge sentences kept clean; the source pieces were already deslopped).
+- **Local-first low-noise posture made the coherent default (#953, builds on #949/#950/#951/#952).** The shipped extension-defaults layer (`packages/core/src/config/extension-defaults.yaml`) — the local-first opinion that sits above the built-in github-first code defaults and below repo config — now sets three existing knobs to their low-noise values, each with an inline intent comment: `autonomy.humanMergeOnly: true` (local-first never auto-merges; a human always merges), `queue.maxAutoFiledIssues: 1` (local-first is PR-first per #952, so auto-filing issues is near-zero; a low cap keeps tracker noise minimal), and an explicit `gates.postFindingsComments: true` (gate findings live ON the PR as review evidence, not tracker noise — keep them on). No new resolver and no new `strategy → knob` coupling: the values come purely from the existing config-merge layering (built-in < extension < repo `.pi/dev-loop/defaults.*` < repo `.devloops`), which already permits all three knobs at the file/extension layer (the runtime Zod schema needed no change). The published JSON-schema artifact `schemas/dev-loop-config.schema.json` was reconciled to match that runtime contract — its `gates` block (which used `additionalProperties: false`) now lists `postFindingsComments` plus the previously-missing `requireFanoutEvidence`/`maxFanoutReviewers`, so the schema no longer rejects the shipped config. The github-first/built-in posture is unchanged — `BUILT_IN_DEFAULTS` still yields `humanMergeOnly: false`, `maxAutoFiledIssues: 10`, and a resolved `postFindingsComments: true`. The post-promotion draft→pre-approval→human-merge flow is untouched. `node:test` coverage locks it: a merged-config assertion (`humanMergeOnly === true`, `maxAutoFiledIssues === 1`, `postFindingsComments === true` under the shipped defaults), built-in-default assertions for the unchanged github-first posture, and an explicit local-first phase-doc intake test (strategy from the shipped extension default, inputSource phase-docs from repo settings) asserting via a logging `gh` stub that NO `gh issue create` / `gh pr create` and NO Copilot request fire before promotion.
+- **Both presentation decks deslopped — shorter headlines + A/B-contrast prose removed** (#936, applies the #944 standard step). Every slide headline on `docs/presentations/applied-dev-loops.html` and `docs/presentations/process-observability.html` was shortened to a short 3–7 word claim and stripped of the binary-contrast antipattern: "The Work Is One Loop Inside Another — and a Handoff Is Never Guessed" → "Loops Inside Loops"; "Prompt-Only Workflows Drift; a State Graph Can't" → "State Graphs Pin Behavior"; "One Interrupt Costs Five Transitions, Not Five Minutes" → "One Interrupt, Five Transitions"; "Those Four Fields Aren't a Wish — They're Where the Work Already Lives" → "The Fields Already Live in the Work"; "The Cheapest Speed-Up Is Making the Waiting Visible" → "Make the Waiting Visible" (and the rest). Slide body text in the publishable HTML renders lost the same "X, not Y" / negation-by-contrast construction in both orderings ("verified, never assumed" → "verified against the real result"; "don't add up. They multiply" → "multiply"; "Stop optimizing… Start measuring…" → "Measure how long the work waits"), with load-bearing distinctions kept plain. The HTML renders are the published source of truth; the Slidev `*-presentation.md` sources mirror the shortened headlines only (their body text already predates the HTML restructure — see `docs/presentations/README.md`). The dark visual identity, section ids, CSP guard, mobile-fit responsiveness, and all six Playwright deck tests are unchanged. Deploys to GitHub Pages on merge.
+
+### Added
+- **First-class `--jq` field-selection, `--silent`/`-s` exit-code checks, and concise output across the read/action scripts** (#981, subsumes #963). One shared helper `scripts/lib/jq-output.mjs` (`evaluateJqFilter`, `emitResult`, `JQ_OUTPUT_PARSE_OPTIONS`, `JQ_OUTPUT_USAGE`, `JqFilterError`) gives the loop a token-economical way to read tool JSON so it never falls back to `gh api | python3` or inline `node -e`. `--jq <filter>` applies a gh-style jq-subset filter (field access, `.a.b` chains, `.[]`/`.field[]` iteration, `.[N]` index, `|` pipes, `select(...)`, `==`/`!=`/`<`/`<=`/`>`/`>=`, `length`, `keys`) to the script's result and prints only the filtered value; an unsupported filter fails closed (stderr + exit `2`). `--silent`/`-s` suppresses stdout and maps the result to an exit code only (`0` pass/truthy, `1` fail/falsy) for zero-output yes/no checks — composing with `--jq` as a predicate (`… --jq '.ciStatus=="success"' -s; echo $?`), where an invalid filter still fails closed at exit `2`, distinct from a clean predicate-false silent exit `1`. Without these flags every script's JSON shape is unchanged. The full subset (not full jq — overkill for the loop's field/predicate needs) is applied uniformly to `scripts/github/capture-review-threads.mjs`, `scripts/loop/copilot-pr-handoff.mjs`, `scripts/loop/pr-runner-coordination.mjs`, `scripts/github/upsert-checkpoint-verdict.mjs`, and the queue scripts `scripts/projects/add-queue-item.mjs`/`move-queue-item.mjs`/`list-queue-items.mjs` (the #963 operator-action-scripts-JSON-only goal folded in here). In addition, the two scripts that emitted a large blob with no concise mode — `scripts/loop/run-watch-cycle.mjs` and `scripts/github/probe-copilot-review.mjs` — gain a `--concise`/`--summary` human-readable mode covering loop state, Copilot round count, unresolved/actionable thread counts, round-cap-clean eligibility, CI status, next action, and (the field `loop info --pr` omits) the current round's NEW Copilot comment bodies. The token-economical reading convention (subcommand/concise → `--silent` → `--jq` → `gh --jq` → never `| python3`/`node -e`) is documented in `skills/dev-loop/SKILL.md` and referenced from the round-cap gate-cadence rule. `node:test` coverage: a helper unit spec (subset extraction success, fail-closed on unsupported syntax, `emitResult` jq/silent/invalid-filter exit-code semantics), an end-to-end CLI spec (`--jq` extraction, invalid filter exit `2` + stderr, `--silent` pass exit `0` silent, predicate-false exit `1` silent, silent+invalid-filter exit `2`, unchanged JSON shape), and concise-mode specs for run-watch-cycle/probe-copilot-review.
+- **Developer-mode internal-tooling-only retrospective check (#982).** A new opt-in workflow flag `workflow.requireRetrospectiveInternalTooling` (default **OFF**, defined in `packages/core/src/config/config.mjs`, shipped explicitly false in `extension-defaults.yaml`, declared in `schemas/dev-loop-config.schema.json`) gates an internal-tooling-only check in the retrospective merge gate. **It is scoped to developer mode — the dev-loops maintainers dogfooding the tooling on themselves — and never blocks consumers** of the extension, who may legitimately use raw `gh`/`python`/`node -e` in their own workflow. When the flag is ON, the merge-approval evaluator (`evaluateRetrospectiveMergeApproval(checkpoint, { developerMode })` in `packages/core/src/loop/pr-gate-coordination.mjs`, routed through at every merge-ready boundary in `evaluatePrGateCoordination`) requires `behavioralReview.internalToolingOnly: true` and `behavioralReview.rawCallViolations: []`, blocking (`retrospective_gate_pending`) when `internalToolingOnly` is not `true` **OR** `rawCallViolations` is missing/non-empty. When the flag is OFF (the consumer default) the check is inert: a complete, merge-approved checkpoint passes exactly as before — even if it omits the fields or records a violation — so a consumer's state changes are never blocked (this also resolves the draft-review back-compat concern: old checkpoints only fail closed in developer mode). The dev-loops repo opts in via its own repo-root `.devloops` (`workflow.requireRetrospectiveInternalTooling: true`), so the discipline still applies to our own retros. The rule treats agent-level top-level raw `gh` (incl. `gh api`/`--jq`), `python`/`python3`, and `node -e`/`node --eval` as the same breach; `node scripts/*.mjs` and dev-loops subcommands are allowed (the scripts call gh/GraphQL internally — that IS the tooling). The dependency-free, deterministic verifier `scripts/loop/check-retro-tooling.mjs` (pure `analyzeTranscript(transcript)` export) is unchanged: it reads a newline-delimited transcript of the agent's shell commands (`--transcript <path>` or stdin) and returns the `rawCallViolations` list, distinguishing script-internal gh from agent-level raw calls and keeping a small explicit write-op allowlist (`gh pr merge`/`gh pr ready`/`gh issue create`/`gh issue edit`, recorded as `allowedWriteOps`). `node:test` coverage: developer-mode-ON gate tests for a recorded violation / `internalToolingOnly:false` / missing-field-old-checkpoint (all block) and a clean record (allows `FINAL_APPROVAL_READY`), the key developer-mode-OFF test asserting a consumer is NOT blocked even with missing fields OR a recorded violation, plus the unchanged verifier spec (`test/loop/check-retro-tooling.test.mjs`). The retrospective-checkpoint contract doc (`skills/docs/retrospective-checkpoint-contract.md` + generated `.claude/` mirror) documents the developer-mode scoping, the opt-in flag, and that consumers are never blocked. Known verifier limitation: segment splitting does not parse shell quoting, so a separator inside a quoted argument can over-report (never under-report) — the fail-closed direction.
+- **"Introducing dev-loops" presentation deck** (#973, companion to the #971 overview article). A self-contained, CSP-safe `docs/presentations/introducing-dev-loops.html` (9 slides: the coordination-delay problem, how manual handoffs compound, every handoff a recorded decision, one bounded decision per change, model-agnostic by construction (open-source models drive the work), the aggregate proof data, install/run, and the companion pieces) reusing the existing decks’ dark glass-card design system, CSP meta, and mobile-fit responsiveness. Added to `scripts/pages/build-site.mjs` `DECKS` (navLabel "Intro (deck)") so it publishes and joins the shared Pages nav; a mirror Playwright spec + config + `test:playwright:intro-deck` script match the existing per-deck tests. Aggregate data only (no internal PR/issue numbers or script paths in slide prose); the A/B-contrast deslop step was applied.
+- **Model-agnostic section in the intro article + open-source model proof** (#979). A new "Model-agnostic by construction" section in `docs/articles/introducing-dev-loops.md`/`.html` (the Pages landing page) explains that the guardrails — every handoff a recorded decision, the same draft → review → green-CI gate, fail-closed on ambiguity, a person merging — bound how far any single step can stray, so a strong open-source model can drive most or all of the work (cheaper, self-hostable, no single-vendor dependence; the gate holds the quality bar). Names the concrete proof: the loop has driven real work end to end on DeepSeek V4, Kimi K2.6, MiniMax M3, Qwen 3.6, and GLM 5.2. Aggregate/conceptual only; A/B-contrast deslop applied.
+
+- **Spike-mode contract doc + worked example** (#966, docs-only; phase 3 of 3 of the spike-mode track #955, closing it; builds on #964/#965). A new canonical operator-sequence doc [`skills/docs/spike-mode-contract.md`](skills/docs/spike-mode-contract.md) walks spike mode end to end against the merged surfaces: author the exploration scaffold and validate it (`scripts/refine/validate-spike-file.mjs`, base sections Question/Approach/Findings/Recommendation), start it (`scripts/loop/resolve-dev-loop-startup.mjs --spike`, `buildSpikeInput` + `evaluateSpikeIntakeState`, intake states `spike_in_progress`/`spike_ready_for_exit`/`ambiguous_fail_closed`), run it under the relaxed `gates.spike` profile, then exit via `scripts/refine/exit-spike.mjs` with disposition `discard` (zero tracker artifacts; the findings doc is the whole record) or `graduate` (`buildGraduatedPlanBody` emits a #947-consumable plan file that enters the existing plan→PR promotion path #952). The doc documents the relaxed gate profile and its rationale — `angles: [scope, docs]`, `required: false`, `requireCi: false` (`packages/core/src/config/extension-defaults.yaml`), resolved through the same `resolveGateConfig(config, "spike")` path and absent for non-spike work — and carries one worked example (`spike-cache.md`) from a question through findings to BOTH exits, showing the actual commands/states and, for graduate, the resulting plan-file body. Cross-referenced from the [Artifact Authority Contract](skills/docs/artifact-authority-contract.md) relationship table and `docs/index.md`; the generated `.claude/skills/docs/` mirror was regenerated. No behavior, schema, defaults, or scripts changed.
+- **Spike-mode relaxed gate profile + discard/graduate exits** (#965, phase 2 of the spike-mode track #955; builds on #964 and reuses #951/#952/#953). A spike is now runnable to a conclusion: a lighter gate posture plus an explicit exit that either discards (zero tracker artifacts) or graduates (emits a #947-consumable plan file). **Relaxed gate profile (config):** a new `gates.spike` profile (same `GateConfig` shape as `gates.draft`/`gates.preApproval`) resolved through the SAME `resolveGateConfig(config, "spike")` path and the existing config-merge layering — no new `strategy → knob` resolver. The shipped extension default (`packages/core/src/config/extension-defaults.yaml`) is intentionally lighter than the production draft → pre-approval → Copilot set: a small docs-first angle set (`scope`, `docs`), `required: false`, and `requireCi: false`, because a spike's deliverable is a findings doc, not production code. Runtime zod (`GatesConfig`/`FileGatesConfig` in `packages/core/src/config/config.mjs`) and the published JSON-schema artifact (`schemas/dev-loop-config.schema.json`, `gates.spike` → `$defs/gateConfig`) were kept in sync (#953's lesson). The github-first/production posture is UNCHANGED: `draft`/`preApproval` stay `required: true`/`requireCi: true`, and `gates.spike` is absent for non-spike work. **Discard/graduate exit contract (pure):** a new `packages/core/src/loop/spike-exit-contract.mjs` (no `scripts/` import, no fs/network/gh; mirrors spike-intake-contract) exports frozen `SPIKE_EXIT_DISPOSITION`/`SPIKE_EXIT_ACTION` plus `evaluateSpikeExit({ spikeIntakeState, disposition }) -> { ok, action?, reason?, spikeIntakeState? }`, eligible ONLY from `spike_ready_for_exit` and failing closed with `not_ready_for_exit` (non-ready/ambiguous state) or `unknown_disposition` (anything but `discard`/`graduate`). `buildGraduatedPlanBody({ question, approach, findings, recommendation })` maps the spike's four sections onto a #947-consumable plan-file body with the four base authoring sections (Status/Objective/In scope/Explicit non-goals) so it passes `validatePlanFile` and enters the existing plan→PR promotion path (#952) unchanged; it is idempotent and fails closed (throws) on an empty required section. The export is added to `packages/core/package.json`. **Thin CLI:** `scripts/refine/exit-spike.mjs` (`--spike-file`/`--disposition`/`--plan-file`/`--json`, mirroring promote-plan.mjs) owns all I/O — it reads the spike, computes intake-state facts via the P1 surfaces (`validateSpikeExplorationSections` for the scaffold + the Recommendation exit-marker), classifies via the pure contract, and on graduate writes the local-first plan file; discard creates ZERO tracker artifacts and graduation opens nothing on the tracker (it only writes a local plan file, then promotes via the existing path). Structured success/error JSON matches the sibling refine/promote scripts. `node:test` coverage: the pure exit contract (discard/graduate eligible from ready; fail-closed on unknown disposition + non-ready/ambiguous/missing state) + the plan-body builder (base-valid output passes `validatePlanFile`, carries the spike content, idempotent, throws on empty sections); the gate-profile resolution (spike profile resolves; shipped default is relaxed and lighter than draft, with draft/preApproval unchanged); and the CLI (discard → empty gh-stub log + exit 0, graduate → base-valid plan file written + empty gh log + idempotent re-run, fail-closed not-ready/unknown-disposition → empty gh log + exit 1, arg validation). Narrative skill docs are #966 (NOT in this phase); no production-flow change for non-spike work.
+- **Top-of-funnel article "Introducing dev-loops" + CSP-safe HTML render** (#971, sibling of the #942/#943 article PRs). `docs/articles/introducing-dev-loops.md` is the single what/why/how-do-I-start entry point the two concept articles lacked: a plain-language introduction to the idea (coordination delay is where AI-assisted lead time goes; the loop makes every handoff an explicit, recorded decision; a person merges by default; work starts from a durable plan or issue), a "what it does to the work" section that uses the repository’s own ~two-week history (~100 merged PRs, ~7/day, about seven in eight tied to a tracked item, every one drafted, reviewed, green-CI gated, and human-merged) to show the straightforward, repeatable process the loop produces, and concrete setup instructions (the Claude Code plugin slash commands, the Pi extension install, the plain-language `start/auto/continue dev loop` entrypoint, and a minimal `.devloops` covering start-local vs issue intake, Copilot rounds on/off, and human-merge). A self-contained, CSP-safe `docs/articles/introducing-dev-loops.html` ports the dark glass-card identity from the sibling renders (no font/CDN/remote assets). Cross-linked from a new Articles section in `docs/index.md`, and the two existing articles gain a one-line "start here" pointer so they read as deep dives beneath the intro. Aggregate figures are gh-verified; the A/B-contrast deslop step (`docs/ab-contrast-deslop-step.md`) was applied (zero flagged constructions). The intro article is also made the **GitHub Pages landing page**: `scripts/pages/build-site.mjs` now publishes `docs/articles/introducing-dev-loops.html` as `site/index.html` (superseding the prior generated "Presentation Decks" card index from #930), publishes the two deep-dive articles and two decks alongside it, and injects a shared navigation bar into the article pages linking the other resources (`injectNav`, which fails closed if a page lacks its `<style>`/`<body>` anchors); the decks are copied as-is. `test/pages/build-site.test.mjs` updated. On merge, `pages.yml` rebuilds and deploys the new landing page.
+- **Spike-mode entry + findings-artifact format** (#964, phase 1 of the spike-mode track #955; reuses #949/#950). A spike is a time-boxed exploratory loop startable from a local question with no GitHub issue. A new spike artifact format mirrors the plan-file format with base sections **Question / Approach / Findings / Recommendation**; a focused sibling validator `scripts/refine/validate-spike-file.mjs` exports `validateSpikeFile(markdownText) -> { checker: "validate-spike-file", ok, errors }` with a distinct `missing_*` code per absent/empty section (`missing_question`/`missing_approach`/`missing_findings`/`missing_recommendation`) plus a thin `--input`/`--json`/`--help` CLI mirroring `validate-plan-file.mjs`. The base-section-checking loop is shared with `validatePlanFile` via a small parameterized helper `checkBaseSections(markdownText, checker, sectionCodes)` in `scripts/refine/_refine-helpers.mjs` (DRY without forcing a shared abstraction across the distinct heading sets). `scripts/loop/resolve-dev-loop-startup.mjs` gains `--spike <path>`, mutually exclusive with `--issue`/`--pr`/`--input`/`--plan-file`: it reuses the `local_implementation` strategy and the existing `local_phase` target as a work-origin addition, makes ZERO `gh` calls / no tracker artifacts at entry (no production-gate ceremony), is exempt from the worktree-isolation `needs_reconcile` guard (a spike has no issue to key a worktree on), and fails closed (exit 1, no bundle) on a missing/unreadable spike file or one missing the exploration scaffold (Question/Approach/Findings). A new pure, deterministic contract `packages/core/src/loop/spike-intake-contract.mjs` (no `scripts/` import, no fs/network/gh) exports a frozen `SPIKE_INTAKE_STATE` and `evaluateSpikeIntakeState({ baseSectionsValid, hasRecommendation }) -> { state }` classifying the spike as `spike_in_progress` (exploration ongoing, no Recommendation yet — phase 2's discard exit) or `spike_ready_for_exit` (Recommendation reached — phase 2's discard/graduate exit seam), failing closed (`ambiguous_fail_closed`) on a malformed artifact. These states are DISTINCT from `PLAN_FILE_INTAKE_STATE` (a spike is not a plan-needing-refinement). The resolver threads the resulting `spikeIntakeState` onto its JSON output and the new core export is added to `packages/core/package.json`. `node:test` coverage: the pure contract spec, the validator spec (all sections ok / each missing→its code / empty-body / CLI), and resolver specs (valid spike → `spike_ready_for_exit` with zero `gh` calls via the empty-log stub, in-progress spike → `spike_in_progress`, missing/malformed → fail closed, mutual exclusivity with `--issue`/`--plan-file`). The relaxed gate profile + discard/graduate exits (#965) and narrative docs (#966) are NOT in this phase; no production-flow change.
+- **Local-first plan-file flow documented end to end + worked example** (#954, docs-only; builds on #949/#950/#951/#952/#953). The canonical [Artifact Authority Contract](skills/docs/artifact-authority-contract.md) (at `skills/docs/`; the `docs/` path in the original stub was stale) now documents the full local-first flow across P1–P5: the plan-file artifact + `localPlanning.plansDir` default (`docs/phases/`) and validator from P1; the `--plan-file` startup entry + `PLAN_FILE_INTAKE_STATE` states (`new_plan_needs_refinement` / `plan_refined_ready_for_promotion` / `ambiguous_fail_closed`) from P2; the in-place refine + `local_human_review` checkpoint (AC/DoD/`Coverage matrix`/`Docs-grill findings` sections) from P3; PR-first promotion + the bidirectional plan↔PR link (`prNumber` front-matter, `already_promoted` idempotency) from P4; and the low-noise extension defaults from P5. The contract was reconciled to the shipped posture: the effective default is `strategy.default: local-first` from `packages/core/src/config/extension-defaults.yaml` (decision #7 of #947), which sits above the `BUILT_IN_DEFAULTS` github-first fallback in `packages/core/src/config/config.mjs`; the stale `.pi/dev-loop/defaults.yaml` shipped-default claim and the "dev-loops is tracker-first" statement were corrected (the package ships defaults via the extension layer and the repo's own `.devloops` sets `local-first`). A new operator-sequence skill doc [`skills/docs/local-planning-flow.md`](skills/docs/local-planning-flow.md) walks validate-plan-file → resolve-dev-loop-startup `--plan-file` → refine-plan-file → promote-plan, and a worked example [`skills/docs/local-planning-worked-example.md`](skills/docs/local-planning-worked-example.md) shows one plan file (`docs/phases/phase-42.md`) evolving through every stage (base sections → refinement sections + coverage matrix + docs-grill findings → `prNumber` front-matter). All three docs are reachable from `docs/index.md` and cross-link the contract; the generated `.claude/skills/docs/` mirror was regenerated. No behavior, schema, defaults, or scripts changed.
+- **Local-first PR-FIRST plan promotion — no issue ever minted** (#952, builds on #949/#950/#951). A pure contract `packages/core/src/loop/plan-file-promote-contract.mjs` exports `evaluatePromoteEligibility({ baseSectionsValid, hasAcceptanceCriteria, hasDefinitionOfDone, existingPrNumber }) -> { ok, action, reason, planFileIntakeState, existingPrNumber }` (`PLAN_FILE_PROMOTE_ACTION` `promote`/`already_promoted`), `buildPromotionPrBody({ planDocPath, acceptanceCriteria, definitionOfDone })`, and a minimal additive plan front-matter parser/serializer (`parsePlanFrontMatter`, `readLinkedPrNumber`, `writeLinkedPrNumber`, `PLAN_FILE_PR_FRONT_MATTER_KEY`). Eligibility is fail-closed: promotion acts ONLY on P3's `plan_refined_ready_for_promotion` state (composed via P2's `evaluatePlanFileIntakeState`); any other state returns `not_ready_for_promotion` with no action so the caller makes zero GitHub mutation. The PR body is the self-contained spec-of-record — it references the committed plan-doc path and carries the FULL `Acceptance criteria` + `Definition of done` extracted from the refined plan, and never an issue-close keyword (no issue is minted). The module is pure (no GitHub, no network, no filesystem I/O). A thin CLI `scripts/refine/promote-plan.mjs` (`--plan-file <path> [--base <branch>] [--branch <name>] [--json]`) owns all I/O: it reads the plan, computes the section-presence facts with the P1 `validatePlanFile`/`extractSection` surfaces, runs the eligibility gate, and on `promote` commits the plan doc to a branch and opens EXACTLY ONE draft PR via the canonical `scripts/github/create-pr.mjs` wrapper (always `--draft`, self-assigned; never raw `gh pr create`), then writes the returned PR number back into the plan's `prNumber:` front-matter and commits the link — the bidirectional plan↔PR link (doc path in PR body, PR number in plan front-matter). The opened draft PR is a normal-shaped PR that enters the existing loop unchanged via `loop startup --pr <n>`; no new lifecycle state and no separate authority artifact. Idempotent: a plan already carrying `prNumber` returns `already_promoted`, opens nothing, and reports the existing PR. The front-matter is a minimal additive extension to P1's plan-file format (documented in `skills/docs/plan-file-contract.md`): plans without a leading `---` block are unchanged and fully valid. A `node:test` contract spec plus a CLI spec cover the ready-gate (refuse non-ready/ambiguous/partially-refined with zero gh calls), the single draft-PR open (one `gh pr create --draft`, asserted via a logging `gh` stub), the plan-doc commit + bidirectional link, idempotency, the fail-closed reasons, zero-pre-promotion activity (empty gh-stub log on refusal and on the idempotent no-op), and the PR body carrying the full AC + DoD.
+- **Local-first refine + human-review checkpoint (off-tracker)** (#951, builds on #949/#950, consumes #948). A pure contract `packages/core/src/loop/plan-file-refine-contract.mjs` exports `refinePlanFileInPlace({ markdownText, baseSectionsValid, hasAcceptanceCriteria, hasDefinitionOfDone, payload }) -> { ok, planFileIntakeState, refinedMarkdown, grillDispositions, stop }` plus `PLAN_FILE_REFINE_STOP`/`DOCS_GRILL_FINDINGS_HEADING`/`COVERAGE_MATRIX_HEADING`. It generalizes the proposal-first intake pattern (emit a local artifact, human-gate it, make zero tracker mutation, stop and ask) to plan files: starting from P2's `new_plan_needs_refinement`, it writes the refiner-produced `Acceptance criteria`, `Definition of done`, `Coverage matrix`, and recorded `Docs-grill findings` sections in place into the single canonical plan file, advances P2's intake state to `plan_refined_ready_for_promotion` via `evaluatePlanFileIntakeState`, and returns a `local_human_review` stop so the loop pauses for human review before any promotion. The docs-grill runs as a step of refinement: each finding is classified with #948's `classifyDocsGrillFinding` and recorded into the plan file, and a malformed finding fails the grill closed. The rewrite is idempotent — a re-run strips any prior copy of each refinement section before appending, so the section count stays at one each. It is pure (no GitHub, no network, no filesystem I/O): the caller supplies the section-presence facts and the refiner payload and writes the returned markdown back, keeping the zero-tracker-mutation guarantee structural. Fail-closed paths cover a non-`new_plan_needs_refinement` starting state, an ambiguous/base-invalid plan, missing AC/DoD/coverage-matrix payload pieces, and a failed grill — each surfaces a reason and advances/writes nothing. A thin CLI `scripts/refine/refine-plan-file.mjs` (`--plan-file <path> --payload <path> [--json]`) reads the plan and the refiner output, computes the facts with the P1 `validatePlanFile`/`extractSection` surfaces, calls the contract, writes the refined plan back in place on success (exit 0, `local_human_review` stop), and fails closed without writing on a fail-closed reason (exit 1). A `node:test` contract spec plus a CLI spec cover refine-from-needs-refinement, the docs-grill-as-a-step recording, in-place idempotency, state advance + local stop, the fail-closed cases, and a zero-tracker-mutation assertion (the logging `gh` stub's call log is empty across the whole path).
+- **Autonomous in-loop docs-grill formalized as a standard step** (#948, sibling of #944 and #929). `docs/docs-grill-step.md` documents the repeatable step that interrogates a change against the repo's own contracts/docs while the loop runs — without a manual main-agent pass: what it checks (claims vs contracts, code-vs-doc drift, stale references, contract-surface accuracy), where it fires (the [refiner](agents/refiner.agent.md) cross-checks the active phase against the contracts it references during refinement, and the existing `docs` pre-approval angle resolves to the [docs persona](agents/docs.agent.md) review mode as one fan-out angle of the [gate review sub-loop](docs/gate-review-sub-loop-contract.md)), and the keep/fix rule. The rule is codified as a pure classifier `classifyDocsGrillFinding` in `scripts/loop/docs-grill-contract.mjs` (`DOCS_GRILL_FINDING_KINDS` `drift`/`stale_reference`/`cosmetic`; `DOCS_GRILL_DISPOSITIONS` `record_finding`/`fix_in_place`/`route_followup`/`ignore_cosmetic`): real drift between code/behavior and a contract claim is recorded as a finding, doc-only drift is fixed in place or routed as a follow-up, and a cosmetic nit never blocks a gate; an unknown kind fails closed (`invalid_finding`). The refiner agent gains an explicit docs cross-check line pointing at the step. The two firing surfaces already ran in-loop, so this captures the contract without adding a new entrypoint. The local-first epic (#947) tree refinement is recorded as the first run (each node was grilled against the contracts it reuses during the refiner+grill fan-out). Cross-linked from `docs/index.md`.
+- **`--plan-file` startup entry + local-planning intake state machine** (#950, builds on #949). `scripts/loop/resolve-dev-loop-startup.mjs` gains `--plan-file <path>`, mutually exclusive with `--issue`/`--pr`/`--input`. It reuses the `local_implementation` strategy and the existing `local_phase` target as a work-origin addition: a valid plan resolves to a `local_phase` bundle with the plan path carried as the target `phase` and no issue/PR number. The path is read-only (zero tracker mutation, no `gh` calls) and exempt from the worktree-isolation `needs_reconcile` guard, because a pre-promotion plan has no issue to key a worktree on. A missing/unreadable plan, or one failing the `validatePlanFile` base-section contract, fails closed (exit 1, no readiness bundle). A new pure, deterministic intake evaluator `evaluatePlanFileIntakeState({ baseSectionsValid, hasAcceptanceCriteria, hasDefinitionOfDone }) -> { state }` (frozen `PLAN_FILE_INTAKE_STATE` enum, no I/O) in `packages/core/src/loop/plan-file-intake-contract.mjs` classifies the plan as `new_plan_needs_refinement` (base sections only), `plan_refined_ready_for_promotion` (base sections plus `Acceptance criteria` + `Definition of done`), or `ambiguous_fail_closed` (invalid base, or exactly one of the two refinement sections). The resolver detects refined-vs-needs-refinement via `extractSection` for each `PLAN_FILE_REFINEMENT_SECTIONS` heading and threads the resulting `planFileIntakeState` onto its JSON output.
+- **Plan-file contract + validator for local-planning mode** (#949). The persisted markdown plan file reuses the existing phase-doc format under `docs/phases/`; a new `localPlanning` config family adds `plansDir` (default `docs/phases/`) across all four config locations (`DevLoopConfigSchema`, `FileConfigSchema`, `BUILT_IN_DEFAULTS`, and the packaged `extension-defaults.yaml`), with `resolvePlansDir(config)` returning the configured directory or the default. A pure validator `scripts/refine/validate-plan-file.mjs` exports `validatePlanFile(markdownText) -> { checker: "validate-plan-file", ok, errors }` that checks the base authoring sections a plan carries before refinement (`Status`, `Objective`, `In scope`, `Explicit non-goals`), reporting each absent or empty-body section under a distinct `missing_*` code; its thin CLI takes `--input`/`--json`/`--help`, reports the verdict in the JSON payload, and exits non-zero on argument or path errors. The format and required sections are documented in `skills/docs/plan-file-contract.md` (mirrored to `.claude/skills/docs/`) and linked from the Artifact Authority Contract.
+- **A/B (binary-contrast) prose removal formalized as a standard deslop step + applied to the two articles** (#944). `docs/ab-contrast-deslop-step.md` documents a repeatable editorial step that removes the binary-contrast / negation-by-contrast antipattern in both orderings ("X, not Y" and "Y, not X") — the strongest single AI tell in generated prose: the detection spec, the parallel-analysis → final-check → human-likeness flow, and the keep-load-bearing-distinctions/cut-the-scaffolding rewrite rule (with a keep-vs-cut table). It is the first documented sub-step of the broader deslop step tracked in #936. Applied as the first run to both Medium articles (`eliminating-coordination-delay`, `make-the-waiting-visible`) and their HTML renders: an audit found ~120 constructions across the articles and decks; the article prose now states each point directly while keeping real technical distinctions plain (the merge signal stays read-from-CI, automation stays state-gated), followed by a human-likeness pass to keep the cadence natural. Each article's `-notes.md` records the pass; a residual check finds zero flagged constructions in the article `.md`/`.html`.
+- **Presentation decks publishable to GitHub Pages** (#930). `.github/workflows/pages.yml` runs the standard Pages pipeline (`actions/configure-pages` + `actions/upload-pages-artifact` + `actions/deploy-pages`, with `permissions: { pages: write, id-token: write, contents: read }` and `concurrency: { group: pages }`). A committed, reusable build script (`scripts/pages/build-site.mjs`, exported `buildSite`/`DECKS`) deterministically assembles `site/` — copying the self-contained deck renders (`docs/presentations/applied-dev-loops.html`, `process-observability.html`) and generating a CSP-safe dark-aesthetic `site/index.html` linking both. The deck HTML stays the single source of truth; `site/` is assembled (gitignored), never hand-maintained. The **build** job (assemble + upload, which validates the assembly) runs on push to `main` + `workflow_dispatch`. At the time of this PR the **deploy** job was gated to `workflow_dispatch` (`if: github.event_name == 'workflow_dispatch'`) until an operator enabled Pages, with the build job validating the assembly on every push; **#941 (below) superseded that gate** once Pages was enabled, so the deploy job now auto-publishes on every push to `main`. `docs/presentations/README.md` documents the decks, local viewing, the deploy flow, and the private-repo plan caveat (a public Pages site from a private repo needs Pro/Team/Enterprise). Also adds the missing `@media (prefers-reduced-motion: reduce)` guard to `applied-dev-loops.html` for parity with the observability deck (disables `scroll-behavior: smooth` on `html`, `scroll-snap-type` on `body`, `scroll-snap-align` on `.slide`). A `node:test` spec (`test/pages/build-site.test.mjs`) asserts the build produces `index.html` + both decks and that the index links both.
+- **Medium-ready long-form article: "Eliminating Coordination Delay in AI-Assisted Dev Workflows"** (#934). `docs/articles/eliminating-coordination-delay.md` is a public-audience, ~1500-word prose piece derived from the Applied dev-loops narrative — zoomed-out and durable (no version-pegging, no raw enum/identifier dumps): hook (cheap code, leaky guessed handoffs) → the one idea (never guess a handoff; make every handoff an explicit, observable decision) → what it buys (safe pauses, mid-flight steering, parallel review → one consolidated verdict, "done" that means merged and read from the real merge signal) → why a state graph beats prompt-only → a one-line close. It carries four captioned `mermaid` diagrams (nested-loops state diagram, PR gate lifecycle flowchart, fan-out/fan-in evidence→verdict flow, mid-flight steering flow) plus a "Rendering the diagrams on Medium" note. A self-contained, CSP-safe preview render `docs/articles/eliminating-coordination-delay.html` ports the dark glass-card identity (navy gradient, glass cards, violet accent, blue kicker, system fonts) in an article layout (~65ch measure) with the four diagrams as inline CSS-flow / inline SVG (no mermaid runtime, no remote resources, wide diagrams scroll in their own container). One prose-adapted storytelling/editorial review pass is recorded and applied in `docs/articles/eliminating-coordination-delay-notes.md`.
+- **Long-form article "Make the Waiting Visible" + a self-contained HTML preview render** (#935). `docs/articles/make-the-waiting-visible.md` adapts the [Process Observability deck](docs/presentations/process-observability-presentation.md) into a Medium-ready public-audience essay (~1.6k words, optional title/dek/tags front-matter): AI writes code in seconds, then the work *sits* — the slow part is the invisible waiting between actions. The arc runs hook → interrupt cost (one interrupt = five transitions) → handoffs restart discovery → git history hides the waiting → make state observable (owner / blocker / latest decision / safe next step) → measure → change → verify → ground it in real mechanisms (board lifecycle, gate evidence trail, deterministic next-action resolver, provider-agnostic CI waits, post-merge reclaim/archive) → close (*"The next agent will write your code in seconds. The lever you control is how long it waits afterward, so measure that."*). Four captioned `mermaid` diagrams (interrupt-cost chain, handoff round trip, measurement loop, observable-state grounding) plus a "rendering on Medium" note. A standalone, CSP-safe `docs/articles/make-the-waiting-visible.html` renders the full article in the dark glass-card identity (navy gradient, glass cards, violet accent, blue kicker, system fonts) as an article layout (~65ch measure), drawing the four diagrams as inline CSS flow / inline SVG — no mermaid runtime, no font/CDN/remote assets. One editorial/storytelling review pass (lens from [docs/slides-story-review-loop.md](docs/slides-story-review-loop.md), adapted for prose) is recorded and applied in `docs/articles/make-the-waiting-visible-notes.md`.
+- **Slides content & storytelling review loop formalized as a bounded reviewer mode** (#929). A sibling of the [UI Designer + Vision Review Loop](docs/ui-designer-review-loop.md) behind `dev-loop` that judges a deck's *narrative*, not its pixels — "does it land?" rather than "does it look right?". `docs/slides-story-review-loop.md` defines the contract: the public entrypoint/dependency boundary (no new public name), a fail-closed REQUIRED INPUT BUNDLE (deck source path + slice-level acceptance criteria + a short storytelling brief + optional captured slide screenshots from the UI smoke harness), the public-audience REVIEW LENS (arc/hook/close, one message per slide + claim titles, sequencing/no-forward-refs, jargon translation, cut/merge/reorder), and a REQUIRED OUTPUT BUNDLE (findings + corrective actions + a single structured outcome `story_review_satisfied` | `needs_iteration`). The pure module `scripts/loop/slides-story-review-contract.mjs` exports the outcome constants plus a fail-closed input-bundle validator and a result-shape validator (no I/O); the prompt template lives at `skills/dev-loop/templates/slides-story-review.md`. The two inline applications already run over the decks are recorded as the first two runs (`docs/presentations/applied-dev-loops-review-notes.md` #926, `docs/presentations/process-observability-review-notes.md` #927). Cross-linked from the UI loop doc, the README, and `docs/index.md`.
+- **Process Observability deck refreshed + a self-contained shareable HTML render** (#927). `docs/presentations/process-observability-presentation.md` (Slidev) is restructured into one 9-slide public-audience story arc in the existing dark glass-card style: claim-style titles, one message per slide, jargon trimmed (`task state` / `pipeline latency` pills cut), the two overlapping handoff-cost slides merged, and a memorable close (*"Stop optimizing how fast you write code. Start measuring how long it waits."*) replacing the bare metric grid. A new grounding slide (`instrumented`) ties the abstract "observable state cuts delay" claim to what actually exists — the queue board lifecycle (owner / safe next step), the gate evidence trail (latest decision + findings), the deterministic next-action resolver, and "automate only where state supports safe continuation" via provider-agnostic CI waits and operator-induced post-merge worktree reclaim / long-done archival — in plain language, no version labels or raw identifiers. A standalone, CSP-safe `docs/presentations/process-observability.html` ships the full deck with all CSS inline (ported from `style.css`: navy gradient, glass cards, violet accent, blue kicker, mono pills) and the flowcharts rendered as inline CSS/HTML flow (no mermaid/CDN, no remote resources) — each slide is a stable-id `<section>` (`hero`, `interrupt-cost`, `handoff`, `blind-spot`, `observable-state`, `measurement-loop`, `instrumented`, `metrics`, `close`). A thin WebKit smoke spec (`test/playwright/observability-deck.spec.mjs` + `playwright.observability-deck.config.mjs`, `npm run test:playwright:obs-deck`) asserts every named section is present with no body horizontal overflow and captures the `hero`, `interrupt-cost`, `observable-state`, `measurement-loop`, `instrumented`, `metrics`, and `close` states under `test-results/ui-smoke/observability-deck/`; one designer/vision review pass (notes in `docs/presentations/process-observability-review-notes.md`, storytelling + visual sections) confirmed equal-height cards, legible flow diagrams, and a landing close with no corrective CSS required beyond the ported baseline.
+- **Applied dev-loops deck refreshed for v0.4.0 + a self-contained shareable HTML render** (#926). `docs/presentations/applied-dev-loops-presentation.md` (Slidev) gains two slides in the existing dark glass-card style: the **gate fan-out/fan-in sub-loop** (build-once neutral bundle = full diff + 1-hop import adjacency, size-guarded; independent per-angle reviewers seeded with the identical bundle; `consolidateFanin` merges per-angle verdicts; no fork primitive / no Workflow dependency; fail-closed `fanout_fanin` verdict with enforced severity counts) and **the coordination runtime owning the full lifecycle** (enforced human merge via `autonomy.humanMergeOnly`, managed `tmp/worktrees/dev-loops/` worktrees, provider-agnostic `watch-ci`, queue board as a deterministic adapter). A standalone, CSP-safe `docs/presentations/applied-dev-loops.html` ships the full deck with all CSS inline and the mermaid diagrams rendered as inline CSS flow (no font/mermaid CDN, no remote resources) — each slide is a stable-id `<section>` for UI smoke targeting. A thin WebKit smoke spec (`test/playwright/applied-deck.spec.mjs` + `playwright.applied-deck.config.mjs`, `npm run test:playwright:deck`) captures the `hero`, `core-idea`, `parallel-review`, `trust`, and `impact` named states under `test-results/ui-smoke/applied-deck/`; one designer/vision review pass (notes in `docs/presentations/applied-dev-loops-review-notes.md`) fixed ragged card heights via equal-height grid rows. A follow-up **public-audience storytelling pass** restructured the deck to one ~8-slide narrative arc (claim-style titles, jargon translated to plain language, raw enum/pill walls cut to at most one identifier per slide as evidence, mechanism→outcome close) while keeping the dark glass-card visual identity unchanged.
+
+### Fixed
+
+- **`promote-plan` pushes the head branch before opening the PR, and the plan-commit step is re-runnable (#969).** Promotion committed the plan doc and then called `create-pr.mjs` (`gh pr create --head <branch>`) without first pushing the fresh local branch; when `gh pr create` did not auto-push, promotion failed `pr_create_failed` AFTER the plan commit had landed — an unrecoverable partial state (plan committed, no PR, no `prNumber`) where a re-run re-entered PROMOTE, found nothing to stage, and dead-ended on `git_commit_failed`. Fix in `scripts/refine/promote-plan.mjs`: (1) the head branch is now pushed with `git push -u origin <branch>` BEFORE invoking `create-pr.mjs`, so a fresh branch exists on the remote and `gh pr create --head` succeeds; a push failure fails closed with the new `git_push_failed` reason + detail and makes ZERO `gh` mutation. (2) The plan-commit step is idempotent: when the plan doc is already committed at HEAD (`git diff --cached --quiet` reports nothing staged), the commit is skipped instead of failing `git_commit_failed`, so a prior partial run recovers on a plain re-run — it pushes, opens the draft PR, and writes the plan↔PR link. The existing fail-closed ready-gate (zero `gh`), `already_promoted` idempotency, and the post-PR-open `failAfterPrOpen` link-commit recovery are unchanged. `test/loop/promote-plan.test.mjs` gains a real bare `origin` remote in its setup and three asserts: the fresh-branch run pushes the branch (asserted via `git ls-remote`) then opens exactly one draft PR, a partial state (plan committed at HEAD on the head branch, no `prNumber`) recovers on re-run with one `gh` call, and a missing-remote push failure surfaces `git_push_failed` with a detail and zero `gh` calls.
+- **Decks fit the phone screen — no horizontal scroll, no clipped content (~390px)** (#937). On a phone the inline flow diagram (`.flow { min-width: max-content }`, `.node { white-space: nowrap }`) forced its grid track wider than the viewport, and `.slide { overflow: hidden }` then **clipped** anything taller than one screen (bullets cut mid-word, flow nodes sheared, slide bottoms cut off). The requirement is that content **fits** both dimensions, not that it scrolls inside a card. Fix (both `docs/presentations/applied-dev-loops.html` and `process-observability.html`): (1) **diagrams fit by stacking** — at `≤600px` `.flow` switches to `flex-direction: column` with arrows rotated to point down and `.node { white-space: normal }`, so the diagram lays out vertically and fits the width (measured 319px flow in a 319px card, zero internal scroll); `.flow { flex-wrap: wrap }` keeps desktop rows fitting too, and `.flow-scroll { overflow-x: auto }` is now a never-triggered last-resort safety, not the fix; (2) **no vertical clip** — `.slide { overflow: hidden }` → `overflow: visible` so a tall slide grows and the page scrolls between slides instead of cutting content off; (3) **mobile sizing** — a `≤600px` breakpoint reduces heading/card/padding scale and top-aligns slides so content fits comfortably. `min-width: 0` on grid children and `overflow-wrap: anywhere` on `p`/`li`/`code` are added. The CSS lives only in the HTML renders (the Slidev `*-presentation.md` sources carry none of it). The dark visual identity, section ids, CSP guard, and reduced-motion guard are unchanged. Both Playwright deck specs (`test/playwright/applied-deck.spec.mjs`, `observability-deck.spec.mjs`) are hardened to enforce **fit** at mobile (390×844): a settle barrier (`waitForLoadState("networkidle")` + `waitForFunction(innerWidth === 390)` + `document.fonts.ready`) removes the cold-start false-fail that measured desktop geometry; the horizontal check now fails if **any** element's `getBoundingClientRect().right > innerWidth + 1` (the `overflow-x:auto` scroller exemption is dropped — diagrams must fit) and asserts `document.scrollingElement.scrollWidth <= innerWidth + 1` (no horizontal page scroll); a vertical-clip check fails any section whose `clientHeight < scrollHeight` while `overflow-y` is `hidden`/`clip`; a guard-the-guard test confirms the fit check fails on a deliberately-wide element. Each spec also captures one mobile named state.
+
 ## 0.4.0
 
 ### Added

package/README.md

@@ -274,3 +274,4 @@ CI splits into a small changed-files gate plus parallel `verify` and conditional
 - [UI Smoke Harness](./docs/ui-smoke-harness.md) — reusable local Playwright/WebKit smoke baseline
 - [UI Artifact Contract](./docs/ui-artifact-contract.md) — screenshot/state artifact contract and CI-promotion rules
 - [UI Designer Review Loop](./docs/ui-designer-review-loop.md) — designer + vision (`uiReviewMode: vision`) review loop contract
+- [Slides Story Review Loop](./docs/slides-story-review-loop.md) — bounded slides content & storytelling reviewer (narrative, not pixels)

package/agents/refiner.agent.md

@@ -46,6 +46,7 @@ For the active phase, require and produce:
 - Do not invent audit findings when no audit artifact was provided
 - when the phase includes watcher or predicate-driven behavior: explicit timeout semantics and negative-case expectations for non-target identities/events
 - when the phase relies on package-first shared helpers inside a source-loaded workspace: explicit integration expectations about whether local callers use published package imports or a thin source/workspace adapter during development
+- cross-check the phase's claims against the contracts and docs they reference (the autonomous docs-grill step, see ../docs/docs-grill-step.md): surface code-vs-doc drift, stale references, and contract-surface inaccuracies as refinement findings while the claims are still being verified
 
 ## Working style
 - Prefer parallel fresh-context fan-out/fan-in when it improves refinement quality or surfaces materially different variants.

package/package.json

@@ -25,10 +25,12 @@
     "test": "npm run test:assets && npm run test:extension && npm run test:scripts && npm run test:core && npm run test:docs",
     "test:assets": "node --test --test-reporter ./test/failure-summary-reporter.mjs test/dev-loop-init-phase-smoke.test.mjs test/contracts/*.test.mjs test/workflow-handoff-contract.test.mjs",
     "test:extension": "node --import tsx --test --test-reporter ./test/failure-summary-reporter.mjs test/extension-checks.test.mjs test/extension-post-merge-update.test.mjs test/extension-command-contract.test.mjs test/extension-package-contract.test.mjs test/extension-pi-adapter.test.mjs test/extension-claude-adapter.test.mjs test/dev-loops-core.test.mjs test/dev-loops-cli.test.mjs",
-    "test:scripts": "node --test --test-reporter ./test/failure-summary-reporter.mjs test/github/*.test.mjs test/loop/*.test.mjs test/docs/*.test.mjs test/projects/*.test.mjs",
+    "test:scripts": "node --test --test-reporter ./test/failure-summary-reporter.mjs test/github/*.test.mjs test/loop/*.test.mjs test/docs/*.test.mjs test/projects/*.test.mjs test/pages/*.test.mjs",
     "test:core": "node --test --test-reporter ./test/failure-summary-reporter.mjs packages/core/test/*.test.mjs",
     "test:dev-loop": "node --test --test-reporter ./test/failure-summary-reporter.mjs skills/dev-loop/scripts/dev-mode-context.test.mjs skills/dev-loop/scripts/render-template.test.mjs skills/dev-loop/scripts/post-gate-verdict-fallback.test.mjs",
     "test:playwright:viewer": "playwright test -c playwright.inspect-run-viewer.config.mjs",
+    "test:playwright:deep-dive": "playwright test -c playwright.dev-loops-deep-dive.config.mjs",
+    "test:playwright:intro-deck": "playwright test -c playwright.intro-deck.config.mjs",
     "smoke:headless": "node scripts/claude/headless-info-smoke.mjs",
     "test:docs": "node scripts/docs/validate-links.mjs && node scripts/docs/validate-no-duplicate-rules.mjs",
     "repo-wiki": "node scripts/repo-wiki.mjs",
@@ -68,7 +70,7 @@
   },
   "dependencies": {
     "mermaid": "11.15.0",
-    "@dev-loops/core": "^0.4.0"
+    "@dev-loops/core": "^0.5.0"
   },
   "repository": {
     "type": "git",
@@ -78,7 +80,7 @@
     "url": "https://github.com/mfittko/dev-loops/issues"
   },
   "homepage": "https://github.com/mfittko/dev-loops#readme",
-  "version": "0.4.0",
+  "version": "0.5.0",
   "files": [
     "cli/",
     "lib/",

package/scripts/github/capture-review-threads.mjs

@@ -11,6 +11,7 @@ import {
 import { parseArgs } from "node:util";
 import { parsePrNumber, requireTokenValue, runChild } from "../_cli-primitives.mjs";
 import { parseRepoSlug } from "@dev-loops/core/github/repo-slug";
+import { JQ_OUTPUT_PARSE_OPTIONS, JQ_OUTPUT_USAGE, emitResult } from "../lib/jq-output.mjs";
 export const REVIEW_THREADS_QUERY = [
   "query($owner: String!, $name: String!, $pr: Int!) {",
   "  repository(owner: $owner, name: $name) {",
@@ -45,9 +46,11 @@ Modes:
 Options:
   --output <path>   Write JSON output to file in addition to stdout
   --help, -h        Show this help
+${JQ_OUTPUT_USAGE}
 Exit codes:
   0   Success
   1   Error
+  2   Invalid --jq filter
 `;
 export function parseCaptureCliArgs(argv) {
   const { tokens } = parseArgs({
@@ -58,6 +61,7 @@ export function parseCaptureCliArgs(argv) {
       output: { type: "string" },
       repo: { type: "string" },
       pr: { type: "string" },
+      ...JQ_OUTPUT_PARSE_OPTIONS,
     },
     allowPositionals: true,
     strict: false,
@@ -69,6 +73,8 @@ export function parseCaptureCliArgs(argv) {
     repo: undefined,
     pr: undefined,
     help: false,
+    jq: undefined,
+    silent: false,
   };
   for (const token of tokens) {
     if (token.kind === "positional") {
@@ -97,6 +103,14 @@ export function parseCaptureCliArgs(argv) {
       options.pr = parsePrNumber(requireTokenValue(token));
       continue;
     }
+    if (token.name === "jq") {
+      options.jq = requireTokenValue(token);
+      continue;
+    }
+    if (token.name === "silent") {
+      options.silent = true;
+      continue;
+    }
     throw new Error(`Unknown argument: ${token.rawName}`);
   }
   const hasLiveArgs = options.repo !== undefined || options.pr !== undefined;
@@ -188,10 +202,14 @@ export async function runCli(
   if (options.outputPath) {
     await writeOutputFile(options.outputPath, payload);
   }
-  stdout.write(`${JSON.stringify(payload)}\n`);
+  return emitResult(payload, { jq: options.jq, silent: options.silent, stdout });
 }
 if (isDirectCliRun(import.meta.url)) {
-  runCli().catch((error) => {
+  runCli().then((code) => {
+    if (typeof code === "number") {
+      process.exitCode = code;
+    }
+  }).catch((error) => {
     process.stderr.write(`${formatCliError(error)}\n`);
     process.exitCode = 1;
   });