@deftai/directive-content 0.113.2 → 0.115.0

package/scm/github.md

@@ -38,6 +38,37 @@ Rules that apply to every `gh` invocation, regardless of context.
 - ⊗ Construct Markdown-rich `gh api -f body="..."` or `gh issue comment --body "..."` commands when the body contains backticks, dollar signs, quotes, or fenced code blocks -- use a body file and the `scm:body:*` wrapper instead
 - ⊗ Write `--body-file` temp files inside the worktree or repository directory -- always use the OS temp directory (`$env:TEMP` on PowerShell, `$TMPDIR` or `/tmp` on Unix)
 
+### Explicit PR bodies skip the GitHub template (#4293)
+
+`gh pr create --body-file`, `--body`, and `--fill` replace `.github/PULL_REQUEST_TEMPLATE.md`. They do not fill it. Leftover-complete and `swarm:finalize-cohort` openers use an explicit body.
+
+- ! Compose the template `Documentation impact` block (`change_class` / `surfaces` / quoted `rationale`) into the body-file or `--body` payload before create or edit
+- ! Run `task verify:docs-impact -- --body-file <path>` on that same file object, then pass `--body-file <path>` to `gh pr create` / `gh pr edit`. The verified bytes are the uploaded bytes
+- ⊗ Treat a custom `--body-file` as an exemption from the declaration
+- ⊗ Verify file A then upload file B
+- ⊗ Treat naming the check in pre-pr / review-cycle / swarm story-open as the leftover-complete or finalize-cohort remedy
+- ⊗ Invent `scm:pr:create` -- `tasks/scm.yml` has `body:pr:edit` only; create stays `gh pr create --body-file`
+
+**PowerShell (PR open, leftover-complete included):**
+
+```powershell
+$bodyFile = [System.IO.Path]::GetTempFileName()
+# $content already includes the template Documentation impact block
+[System.IO.File]::WriteAllText($bodyFile, $content, [System.Text.UTF8Encoding]::new($false))
+task verify:docs-impact -- --body-file $bodyFile
+gh pr create --title "feat: example" --body-file $bodyFile
+```
+
+**Unix (bash/zsh):**
+
+```bash
+bodyFile=$(mktemp)
+# $content already includes the template Documentation impact block
+printf "%s" "$content" > "$bodyFile"
+task verify:docs-impact -- --body-file "$bodyFile"
+gh pr create --title "feat: example" --body-file "$bodyFile"
+```
+
 ## Safe Markdown Body Posting (#1555)
 
 Use `task scm:body:*` (`scm:body:comment:create`, `scm:body:issue:edit`, `scm:body:pr:edit`) whenever an agent needs to post or edit Markdown-rich GitHub text. The helper accepts `--body-file <path>`, wraps the body as JSON, calls `gh api --input -` with explicit UTF-8 encoding, and prints the live read-back object returned by `gh`.
@@ -214,6 +245,7 @@ On Windows PowerShell (5.1 and often `pwsh` when commands are not routed through
 $bodyFile = [System.IO.Path]::GetTempFileName()
 [System.IO.File]::WriteAllText($bodyFile, $prBody, [System.Text.UTF8Encoding]::new($false))
 git commit -F $bodyFile
+task verify:docs-impact -- --body-file $bodyFile
 gh pr create --title "feat: example" --body-file $bodyFile
 ```
 

package/skills/deft-directive-build/SKILL.md

@@ -277,7 +277,9 @@ Multi-iteration implement-fix and pre-PR polish loops MUST carry **both** a succ
 |------------|--------------|----------------------|
 | Implement / quality fix (tests, lint, typecheck, coverage, AC) | Affected/static gates green for the change; AC met | **max 5** fix iterations **or** **3** consecutive identical outcomes (same failing command + same primary error class) with no material code/config change |
 | Pre-PR polish (`deft-directive-pre-pr` Read-Write-Lint-Diff) | Full pass with zero further edits | **max 3** polish passes **or** **2** consecutive no-diff / same-diff outcomes |
-| Full `task check` re-run after a red merge chokepoint | `task check` green | Counts toward the implement/quality fix envelope above (do not open a separate unbounded check-retry loop) |
+| Full `task check` re-run | `task check` green after a red merge chokepoint or a new commit | Counts toward the implement/quality fix envelope above (do not open a separate unbounded check-retry loop) |
+
+- ! Re-run full `task check` only after a red merge chokepoint or a new commit.
 
 **On failure stop:**
 

package/skills/deft-directive-design-critique/SKILL.md

@@ -45,12 +45,13 @@ Comment lead (model then role): Stop 3 — Critic envelope.
 Operator-gated loop. Successor lean. Parent-side substantiation. Operator verbs. Dual stop. Halt line. Bind after accepted synthesis.
 After this round's siblings are posted: successor lean, then verbs.
 Auto-stamp after operator confirm; not while same-round siblings outstanding.
+Yolo standing confirm of a posted all-accept map: Operator verbs. Confirm conjunct only. Not ingest.
 Walk / walk all. Auto-stamp when agents agree: Operator verbs.
 Parent chip write: scm:issue:design-critique-chip.
 Chip apply miss is non-blocking convenience; do not halt. Ingest waits on the completed-arc record, not a catalog chip.
 Run posture: parse closed tokens; else ask. Field lives in the contract.
 Grok-bot detect default and widget apply-set live in the contract. Consume parseOperatorRunPosture.
-Widgets: numbered Discuss and Back. Plain English first in main-chat.
+Widgets: numbered Discuss and Back. Plain English first in main-chat. `## Plain-language summary`.
 
 Each critic dispatch EXITs after posting.
 

package/skills/deft-directive-pre-pr/SKILL.md

@@ -64,11 +64,12 @@ task verify:plan-sequence -- --target-kind <entry-kind> --target <entry-id>
 
 > **Invariant:** every change MUST pass the full gate at least once before merge. Pre-PR is the merge chokepoint — NOT every iteration commit.
 
-- ! **Iteration lane (Phases 2–3 loop):** use affected/static gates on changed paths — `vitest run --coverage <paths>`, relevant `verify:*` on touched files, `task coverage:hotspots` — instead of full `task check` on every pre-PR iteration.
-- ! **Merge chokepoint (Phase 3 Lint exit + final confirm):** run full `task check` once before push/PR; Phase 3c targeted coverage precedes but does not replace the full gate.
+- ! **Iteration lane (Phases 2–3 loop):** use affected/static gates on changed paths — `vitest run --coverage <paths>`, relevant `verify:*` on touched files, `task coverage:hotspots` — instead of full `task check` on every pre-PR iteration. Phase 3 sits inside this loop, so it uses the iteration lane, not a full suite.
+- ! **Merge chokepoint (once after the loop, before push/PR):** run full `task check` once before push/PR. This is a #1704 revision of the prior Phase 3 Lint exit + final confirm (two runs, one inside every iteration). The escape-rate guard below still applies; do not invent a separate surface.
 - ! **Escape-rate safety (#1703 Tier-1):** before recommending fleet-wide fast-lane tightening, cite `#1703` Tier-1 telemetry (`helped/crud-metrics.jsonl`) and `task eval:health` — do NOT invent a separate escape-rate surface.
 - ~ **In-engine incrementality (#1713):** content-hash cache + runner-delegated affected selection are sibling work (#1713).
 - ⊗ Treat Phase 3c targeted coverage alone as PR-ready without full `task check` at the merge chokepoint.
+- ⊗ Run full `task check` inside every pre-PR Phase 3 iteration — Phase 3 is inside the loop; the chokepoint is once after exit.
 
 ## When to Use
 
@@ -90,6 +91,7 @@ Each iteration proceeds through all phases in order. Do NOT skip phases or reord
 - ~ If changed files include `xbrief/PROJECT-DEFINITION.xbrief.json`, a configured `codebase-map` provider artifact, or `.planning/codebase/MAP.md`, read the MAP and canonical metadata together. The MAP is orientation; `plan.architecture.codeStructure` and provider artifacts remain authoritative.
 - ! When adding a `!` or `⊗` rule that prohibits a specific command, pattern, or behavior, search the same file for any `~`, `≉`, or prose that recommends or permits the same command/pattern -- resolve all contradictions in the same commit before pushing
 - ! When strengthening a rule (e.g. upgrading `~` to `!`), grep for the term in the full file and verify no weaker-strength duplicate remains
+- ! When adding a weaker rule (`~`, `≉`, or prose that recommends or permits a command/pattern) beside an existing `!` or `⊗` for the same command/pattern in the same file, resolve the contradiction in the same commit — weaken-beside-a-MUST is the same-commit reconcile as adding or strengthening a MUST (#4135 / #4324)
 - ~ Note any inconsistencies, missing RFC2119 markers, stale cross-references, or incomplete sections
 - ~ Check that CHANGELOG.md entries match the actual changes made
 - ! If the change alters **user-visible behavior**, apply [coding/docs.md](../../coding/docs.md) (#447): update the matching user-facing surface (CHANGELOG when user/operator-visible, CLI help / commands.md for new or changed verbs/flags, README/getting-started for install/first-run, skill/strategy triggers when entry points change) in this PR
@@ -108,9 +110,9 @@ Each iteration proceeds through all phases in order. Do NOT skip phases or reord
 
 ### Phase 3 -- Lint
 
-! Run the merge-chokepoint gate and fix any failures (#1704).
+! Run the iteration-lane gates and fix any failures (#1704). Full `task check` is the post-loop merge chokepoint, not this per-iteration phase.
 
-- ! Run full `task check` (fmt + lint + typecheck + tests + coverage + verify:*) — the merge chokepoint, not every pre-PR iteration
+- ! Run affected/static gates on changed paths (`vitest run --coverage <paths>`, relevant `verify:*`, `task coverage:hotspots`) — not full `task check` inside the loop
 - ! Fix all failures before proceeding to Phase 3b
 - ~ If a lint fix requires changing a file, that counts as a change for the Loop phase
 
@@ -210,7 +212,10 @@ A red product verification may be resolved only by a product change or an indepe
 ! Exit when a complete Read-Write-Lint-Diff cycle produces **zero changes** -- no file edits in Write, no lint fixes in Lint, and no scope issues in Diff.
 
 After exiting:
-- ! Run `task check` one final time to confirm clean state
+- ! Run full `task check` once — the merge chokepoint. Phase 3 used the iteration lane, so this is the first full suite in the loop, not a second run.
+- ! If that post-loop `task check` is red and the fix changes files, restart from Phase 1 (Read). Do not only re-run the gate — those edits must pass Read and Diff (#4324).
+- ! After that restart exits with zero changes, run full `task check` once again (recovery after a red merge chokepoint).
+- ! Re-run the full gate only after a red merge chokepoint or a new commit.
 - ~ The branch is now ready for push and PR creation
 
 
@@ -233,6 +238,8 @@ Docs: `docs/decision-log.md`.
 - ⊗ Make out-of-scope fixes during Write -- this introduces scope creep that Diff will flag, forcing another iteration
 - ⊗ Ignore the iteration count -- more than 3 iterations usually indicates oscillating fixes or an unclear spec task
 - ⊗ Add a prohibition (`!` or `⊗`) without scanning the same file for conflicting softer-strength rules (`~`, `≉`) that reference the same term
+- ⊗ Add a weaker rule (`~` / `≉`) beside an existing `!` / `⊗` for the same command/pattern without reconciling in the same commit
+- ⊗ After a red post-loop `task check`, fix files and re-run only the gate without restarting Read-Write-Lint-Diff
 - ⊗ Skip `task pr:check-closing-keywords` (#737) before pushing a PR. Intent mode (#3015) also refuses bare/conditional real `Closes #N` without `--allow-close`. The negation-context substring match is the Layer 0 (prevention) gate that prevents the recurring auto-close of umbrella / staying-OPEN issues observed in #697 (closed #642), #401 (closed #642), #700 (closed #233), and #735 (closed #734) -- each incident required manual reopen and downstream cleanup. The lint's three-state exit (0 clean / 1 hits found / 2 config error) MUST be treated as a hard refusal: rewrite the PR body / commit messages until clean, OR pass `--allow-known-false-positives` ONLY for legitimately-quoted occurrences (test fixtures, documentation that discusses the trigger token literally). See `skills/deft-directive-swarm/SKILL.md` Phase 6 Step 1 for the corresponding Layer 3 (recovery) `pr:check-protected-issues` rule (#701)
 - ⊗ Invent remote PR/SHA/CI/review claims in handoff evidence without same-turn probe binding — invented-done (#3120)
 - ⊗ Fill remote ship/gate fields from memory when only local work completed; legal partial omits PR fields (#3120)

package/skills/deft-directive-review-cycle/SKILL.md

@@ -494,7 +494,7 @@ Workflow failover arming (Blacksmith cancelled → GH-hosted lane) is sibling is
 
 ! Auto-restart detection -- when the polling loop observes a NEW `startedAt` (Greptile dropped its prior check run and started a fresh one without any push from the agent, e.g. service-side restart), the agent MUST reset its elapsed-time clock to the new `startedAt` AND notify the user that an auto-restart was detected. Resetting the clock without notifying is forbidden -- the user needs to know the cycle effectively re-started.
 
-! **`INCOMPLETE_BUT_RATED` stall signature (#1259):** when a poll observes a parsed `Confidence Score: X/5` number BUT no terminal check-run (no `completed` status with a `{success, neutral}` conclusion) AND/OR no HEAD-matching `Last reviewed commit:` completion marker, classify the state as **`INCOMPLETE_BUT_RATED`** — Greptile has emitted a confidence rating against a review that has NOT terminally landed on the current HEAD. This is NOT an exit condition (the Step 6 fail-closed all-of resolves the missing fields to `unknown`); treat it as a stall signature and keep polling, escalating per the 10-minute threshold above if it persists. A confidence number is the single most common false-positive for a premature exit — `INCOMPLETE_BUT_RATED` names the trap so the agent does not mistake a rating for a verdict.
+! **`INCOMPLETE_BUT_RATED` stall signature (#1259):** when a poll observes a parsed `Confidence Score: X/5` number BUT no terminal check-run (no `completed` status with a `{success, neutral}` conclusion) AND/OR no HEAD-matching `Last reviewed commit:` completion marker, classify the state as **`INCOMPLETE_BUT_RATED`** — Greptile has emitted a confidence rating against a review that has NOT terminally landed on the current HEAD. This is NOT an exit condition (the Step 6 fail-closed all-of resolves the missing fields to `unknown`); treat it as a stall signature and keep polling, escalating per the 10-minute threshold above if it persists. A confidence number is the single most common false-positive for a premature exit. Thin HTML `<!-- greptile_summary -->` with confidence and no Last reviewed commit is the #4289 named state, not INCOMPLETE_BUT_RATED — `INCOMPLETE_BUT_RATED` names the trap so the agent does not mistake a rating for a verdict.
 
 ⊗ Auto-retrigger Greptile (empty commits, force-pushes, agent-posted `@greptileai` comments, status-check rebuilds) without explicit user approval. The escalation menu's option 2 is the ONLY supported re-trigger path, and even that requires the user to pick it.
 
@@ -778,8 +778,13 @@ NOTES: <short>
     ```
 
    A missing or non-matching completion marker is `unknown`. See [`../../templates/swarm-greptile-poller-prompt.md`](../../templates/swarm-greptile-poller-prompt.md) `### Last reviewed commit:` for the canonical regex shared with the push-driven poller loop.
+
+   ! **Thin HTML named state (#4289):** when the rolling summary is `<!-- greptile_summary -->` with a parsed `Confidence Score` and **no** `Last reviewed commit:` line, this is **not** `INCOMPLETE_BUT_RATED` and **not** #1543 informal-clean. Pin HEAD via the already-fetched `Greptile Review` check-run on current HEAD (`completed` + `{success,neutral}`), not `pending_required`. Name both SHA gates (`evaluateCleanGate` and `evaluateGates` / `findLastReviewedCommitSha`). Body SHA remains preferred when present. Findings for this shape come from REST `pulls/<N>/comments` and/or check-run `N comments added` text -- vacuous `detect()` zeros are not a findings pin. Missing findings channel fail-closes (do not CLEAN on confidence plus terminal check-run plus detect zeros). Dirty REST P0/P1 is NEW_P0_P1 without a body SHA.
+
+   Open-PR symptom if that pin or findings channel is missing (do **not** bind every-consumer-until-TIMEOUT): blocking `pr:watch` waits to cap TIMEOUT; `--one-shot` is PENDING; `pr:merge-ready` is a parse fail; swarm poller STALLs; this Step 6 field stays `unknown`. Keep #4288 as the merged/closed terminal.
+
 4. ! **Confidence meets resolved min (#3095)** — the parsed `Confidence Score: X/5` is **>=** the resolved floor from `plan.policy.review.minGreptileConfidence` (inspect: `task policy:show --field=minGreptileConfidence`; `pr:watch` / `pr:merge-ready` share this SoT). Resolution order: typed project policy > framework dogfood detect (framework source → **5**) > consumer default (**4**, legacy confidence > 3 / 4/5+). A score below the floor, an unparsed confidence, or an absent confidence is `unknown`. Directive dogfood MUST NOT exit CLEAN on 4/5.
-5. ! **No P0/P1 findings** — the triple-tier (+ Tier 2.5) detector reports zero P0 and zero P1 findings (P2 issues are non-blocking style suggestions and do not gate the loop).
+5. ! **No P0/P1 findings** (thin HTML #4289: REST pulls comments and/or check-run comments-added, not rolling-summary detect zeros) — the triple-tier (+ Tier 2.5) detector reports zero P0 and zero P1 findings (P2 issues are non-blocking style suggestions and do not gate the loop).
 
 ! All five fields MUST hold on the SAME single fresh fetch. The agent MUST NOT assemble a "pass" by combining a terminal check-run observed on one poll with a confidence parsed on an earlier poll — the read is atomic per the SHA-pinned-AT-READ-TIME rule above.
 
@@ -962,6 +967,7 @@ task lifecycle:event -- emit plan:approved \
 - ⊗ Run a partial test suite instead of `task check` without documenting the pre-existing failure reason and open issue number in the PR body
 - ⊗ Create a PR without running `skills/deft-directive-pre-pr/SKILL.md` first -- the pre-PR quality loop catches issues before they reach the reviewer
 - ⊗ Activate Approach 3 (blocking `Start-Sleep` loop) without first warning the user that it will lock the conversation pane and receiving confirmation
+- ⊗ CLEAN thin HTML on parsed confidence plus a terminal Greptile Review check-run plus vacuous detect() zeros -- findings come from REST pulls comments and/or comments-added (#4289)
 - ⊗ Exit the review loop on a Greptile confidence number alone while the check run is non-terminal -- a confidence score is NOT a verdict without a terminal check-run (`completed` + `{success, neutral}`) AND a HEAD-matching `Last reviewed commit:` completion marker (#1259)
 - ⊗ Call `gh pr merge` on cached/earlier review state without an immediately-preceding pre-merge re-poll that re-satisfies the Step 6 fail-closed all-of (#1259)
 - ⊗ Merge on mechanical Ready-to-merge / green checks while bot comment prose records should-not-merge or confidence below `minGreptileConfidence` (#3225 / #3095)

package/skills/deft-directive-setup/SKILL.md

@@ -607,8 +607,7 @@ omit = [
       "Strategy": "Use {strategy name} for this project",
       "Quality": "Run task check before every commit. Achieve >= {coverage}% coverage overall + per-module. Store secrets in secrets/ dir.",
       "ProjectRules": "{Any rules the user specified, or 'No project-specific rules defined.'}",
-      "Branching": "{If confirmed trunk-based: 'Trunk-based workflow', otherwise 'Branch-based workflow (default)'}",
-      "DeftVersion": "0.20.0"
+      "Branching": "{If confirmed trunk-based: 'Trunk-based workflow', otherwise 'Branch-based workflow (default)'}"
     },
     "items": []
   }
@@ -871,8 +870,9 @@ Per [strategies/interview.md](../../strategies/interview.md#interview-rules-shar
 
 ## Anti-Patterns
 
-- ! When deft-directive-setup generates or updates USER.md or PROJECT-DEFINITION.xbrief.json, the `deft_version` field MUST be set to the current framework version
-- ⊗ Generate a USER.md or PROJECT-DEFINITION.xbrief.json without including the `deft_version` field
+- ! When deft-directive-setup generates or updates USER.md, the `deft_version` field MUST be set to the current framework version. USER.md is the only artifact that carries this field.
+- ⊗ Generate a USER.md without including the `deft_version` field
+- ⊗ Write `deft_version` or `DeftVersion` into PROJECT-DEFINITION, specification, or plan xBRIEFs. Those stamps are retired (HASH4271). Specification `deft_version` has no framework seeding path; Pass 1 absence locks are the only control.
 - ⊗ Explore codebase before Phase 1 questions
 - ⊗ Read framework files before first question
 - ⊗ Batch multiple questions into one message — ask one at a time, interview style

package/skills/deft-directive-swarm/references/core-phase-0.md

@@ -46,7 +46,7 @@ Cross-links: Phase 3 Worker-owns-lifecycle (`references/core-phase-3.md`), revie
 ### Headless cohort fast-path: low-ceremony launch (C1 / #1387)
 
 ! `task swarm:launch` claims this worktree occupancy lease (`intent: swarm`) before emitting the manifest (#3433). A live foreign occupant fails closed.
-! **Spawned mutating workers take their own worktree (#4066 / #4215).** Phase 2/3 must put each implement/review/fix leaf on a uniquely reserved linked worktree (C3 worktree-map or `git worktree add`) *before* occupancy claim. Sharing the primary checkout with a live occupant is refuse, not grant-across-host. Grok `spawn_subagent` implement dest is `cwd` only. `isolation=worktree` / `worktree_path` stay on reroot hosts (Cursor Task, Claude subagent, Codex). Do not advertise `worktree_path` for Grok. Terminal release is dispatcher-owned (`phase: terminal` + incarnation); do not tell a refused Grok child to steal master.
+! **Spawned mutating workers take their own worktree (#4066 / #4215 / #4295).** Phase 2/3 must put each implement/review/fix leaf on a uniquely reserved linked worktree (C3 worktree-map or `git worktree add`) *before* occupancy claim. Sharing the primary checkout with a live occupant is refuse, not grant-across-host. Grok `spawn_subagent` implement dest is `cwd` only. Claude/Codex keep `isolation=worktree` / path keys. Cursor Task has no local dest field — local implement is dest-placing (nursery inherit on a dest-rooted window, or dest-rooted `@cursor/sdk` `Agent.create` cwd), not Task dest keys. Do not advertise `worktree_path` for Grok or Task dest keys for Cursor. Terminal release is dispatcher-owned (`phase: terminal` + incarnation); do not tell a refused Grok child to steal master.
 ! When the operator supplies a **pre-approved cohort** via the **C1** `task swarm:launch` CLI, Phase 0 runs in headless / low-ceremony mode: the per-phase interactive approval gates (the Step 0c promote-fill prompts, the Step 0.5 lifecycle-bridge approval, and the Step 4/5 allocation approval) collapse into a SINGLE consent -- the `## Allocation context` token (#1378) carried in the dispatch envelope. The interactive promote-fill loop (Step 0a -- 0d below) is SKIPPED.
 ! The **C1** signature is `task swarm:launch -- --stories <ids|paths> [--group <label>] [--worktree-map <path>] [--base-branch <branch>] [--autonomous]`. `--stories` names the pre-approved story ids or xBRIEF paths; `--group` is an optional cohort label; `--worktree-map` points at the pre-created **C3** worktree-map JSON consumed in Phase 2; `--base-branch` overrides the default `master`; `--autonomous` runs without the interactive launch confirmation.
 ! The SINGLE consent is the #1378 `## Allocation context` token with `dispatch_kind: swarm-cohort` and a NON-NULL `allocation_plan_id` AND `batching_rationale` (the recognition contract in `templates/agent-prompt-preamble.md` § 2.5). That token IS the batched approval for the whole cohort -- the deterministic-question gates the interactive path runs (per [`../../contracts/deterministic-questions.md`](../../contracts/deterministic-questions.md)) are bypassed wholesale on the headless path, not asked once per phase.

package/skills/deft-directive-swarm/references/core-phase-3.md

@@ -33,7 +33,7 @@
 8. ! **Select launch path automatically** based on detection results — do NOT present static options:
    - **`start_agent` available** → Orchestrated launch (Step 2a) — preferred path, fully automated, no manual tab management
    - **`start_agent` unavailable, Warp detected** → Interactive Warp tabs (Step 2b) — full MCP, global rules, warm index; requires manual tab management
-   - **Cursor `Task` tool available (no `start_agent`, no `WARP_*`)** → Cursor sub-agent launch (Step 2e) via the `Task` tool with `run_in_background: true` (Tier 1 / Approach 1) — keeps the monitor pane interactive; descriptor is `cursor-composer` (interactive) or `cursor-cloud-agent` (cloud)
+   - **Cursor detected (no `start_agent`, no `WARP_*`)** → Cursor dest-placing launch (Step 2e / `host-cursor.md`) (Tier 1 / Approach 1). Local implement is nursery inherit on a dest-rooted window or dest-rooted `@cursor/sdk` `Agent.create({ local: { cwd } })` with reservation before exec — not Task dest keys. Descriptor is `cursor-composer` (interactive) or `cursor-cloud-agent` (cloud)
    - **Claude Code available (no `start_agent`, no `WARP_*`, no Cursor classification)** → Claude Code launch (Step 2g) via `Agent` with `run_in_background: true` (Tier 1 / Approach 1) — descriptor is `claude-code` (#3134)
    - **OpenClaw `sessions_spawn` available (no `start_agent`, no `WARP_*`, no Cursor, no Claude Code)** → OpenClaw launch (Step 2f) via `sessions_spawn` (Tier 1 / Approach 1) — descriptor is `openclaw`
    - **Grok Bot unique signals available (no earlier Tier-1 primitive)** → Grok Bot launch (Step 2h) via Task/executor/CloudAgent (Tier 1 / Approach 1) — descriptor is `grok-bot` (#4201)

package/skills/deft-directive-swarm/references/host-cursor.md

@@ -2,32 +2,52 @@
 
 Legend (RFC2119): !=MUST, ~=SHOULD, ≉=SHOULD NOT, ⊗=MUST NOT, ?=MAY.
 
-Descriptor: `cursor-composer` or `cursor-cloud-agent` (Cursor `Task` tool).
+Descriptor: `cursor-composer` or `cursor-cloud-agent`. Detect still probes the Cursor `Task` tool. Local implement launch is dest-placing Cursor, not dest-binding Task (#4295).
 
 Load this file only after detect selects Cursor. Do not load other host adapters.
 
-! **Windows + Cursor Task-tool console windows (#2563):** Local Cursor Task swarms on Windows are first-class. Shipped mitigations (do not regress): `windowsHide: true` (CREATE_NO_WINDOW) on engine/spawn paths, and warm-dist skip via `tasks/ts-build-fresh.cjs` so `engine:_ts-build` does not cold-rebuild when `packages/cli/dist` is current. See `templates/agent-prompt-preamble.md` §3.8. ! Default to **local** Task workers on Windows (not cloud-for-Windows). Parallel cohorts are allowed — do not force concurrency=1 for #2563. ⊗ Drop or weaken those mitigations without a replacement that keeps Windows local swarm workable.
+! **Windows + Cursor Task-tool console windows (#2563):** Local dest-placing Cursor swarms on Windows are first-class. Shipped mitigations (do not regress): `windowsHide: true` (CREATE_NO_WINDOW) on engine/spawn paths, and warm-dist skip via `tasks/ts-build-fresh.cjs` so `engine:_ts-build` does not cold-rebuild when `packages/cli/dist` is current. See `templates/agent-prompt-preamble.md` §3.8. ! Default to **local dest-placing Cursor** on Windows (not cloud-for-Windows; not dest-binding Task). Parallel cohorts are allowed — do not force concurrency=1 for #2563. ⊗ Drop or weaken those mitigations without a replacement that keeps Windows local swarm workable.
 
-### Step 2e: Cursor Launch (Task tool available) — #1877
+### Step 2e: Cursor Launch — dest-placing, not Task dest (#4295 / #1877)
 
-! When the platform descriptor is `cursor-composer` or `cursor-cloud-agent` (Cursor `Task` tool detected with Cursor signals, no `start_agent`, no `WARP_*`, no Claude Code classification, no OpenClaw `sessions_spawn`, no `spawn_subagent`), dispatch each worker via the Cursor `Task` tool with:
-0. Create `<worktree>/.deft-scratch/subagent-status/` before spawn if `swarm:launch` / `swarm:pre-dispatch` did not already (#3730). Include preamble § 10.5 (heartbeat + commit early) and poll with `task verify:subagent-alive -- --require-agent <agent-id>`.
-1. The canonical `templates/agent-prompt-preamble.md` content as the preamble (AGENTS.md read mandate, #810 xBRIEF gate, #798 PowerShell UTF-8, pre-PR + review-cycle mandates).
-2. The standard worktree prompt (STEP 1-6 from the Prompt Template below).
-3. The worktree path set to the agent's isolated git worktree.
-4. ! **`run_in_background: true`** for any worker or poller whose loop runs longer than a short task (~3 min) — implementation, fix, and review-cycle workers — so the monitor conversation pane stays interactive (#1880 Gap D). The parent is notified on completion.
-5. ! **Deliberate model routing (#1739):** pass the route's `resolved_model` (when non-null) as the Task tool's `model` argument — stamping the C2 manifest is prep; the recorded model MUST reach the actual spawn call.
+Cursor `Task` has no local dest field. Extra keys (`isolation`, `worktree_path`, `cwd`) are stripped before `preToolUse`. Hook `updated_input` dest fields do not re-root Task. A prompt token may select a trusted reservation record; it does not move the child. Comment 5611146439 stands.
 
-~ This is the first-class Cursor path. It is **Tier 1 → Approach 1** (a backgroundable sub-agent primitive), equivalent in tier to `start_agent` / `spawn_subagent` / OpenClaw `sessions_spawn`; it MUST NOT be downgraded to a `generic-terminal` blocking poll. Cursor pollers whose loop runs > ~3 min MUST honour the sub-agent heartbeat contract (`docs/subagent-heartbeat.md`, #1166), same as the `spawn_subagent` path.
+⊗ Dispatch local implement via Cursor `Task` dest keys, a worktree map in the prompt, or inherited primary cwd.
+⊗ Treat a prompt token as moving the child onto a worktree.
 
+! **Local implement (required):** dest-placing Cursor. The child process or window folder is an already-created linked worktree *before* it can write.
+
+Two local hatches, in this order:
+
+1. **Nursery inherit (interactive dest-rooted window).** Open Composer on the reserved linked worktree (not primary). One Task child may inherit that payload root: consult treats the window as dest, mints a unique reservation, and admits the child through `occupancy:grant`. While that grant is live, parent product writes in that tree are denied. A second Task in the same window is reservation-conflict. Nested Task from a dest-rooted session is still dest-missing unless *this* window is the dest. Nursery child `workspace_roots` equal the parent window.
+2. **Headless dest-rooted SDK.** `@cursor/sdk` `Agent.create({ local: { cwd: <reserved-worktree> } })` with `Agent.resume` / `agent.send` as the retain-capable sibling. Reservation before exec. Process-handle liveness (not parent-cwd `verify:subagent-alive`). Doctor check `cursor-sdk-auth` for the separate SDK login (`CURSOR_API_KEY`). The leaf implements in that session and does not nested-Task.
+
+! Create `<worktree>/.deft-scratch/subagent-status/` before spawn if `swarm:launch` / `swarm:pre-dispatch` did not already (#3730). Include preamble § 10.5 (heartbeat + commit early). Headless SDK liveness is the process handle; scratch heartbeats remain for dest-rooted Composer sessions.
+! Include the canonical `templates/agent-prompt-preamble.md` content as the preamble.
+! Include the standard worktree prompt (STEP 1-6 from the Prompt Template).
+! **`run_in_background: true`** (or SDK async create) for any worker or poller whose loop runs longer than a short task (~3 min).
+! **Deliberate model routing (#1739):** pass the route `resolved_model` (when non-null) into the actual spawn primitive.
+
+~ Cursor stays **Tier 1 → Approach 1**. Do not downgrade to `generic-terminal`. Spawn available means dest-placing dispatch exists, not Task-present (#3032 / #4295).
+? Cloud `Task` (`environment: "cloud"`) is the #4294 dest-consult carve-out. Complementary. Do not treat cloud as local dest-proof. Do not duplicate that hook slice here.
+
+### Fence-in-place (parked)
+
+Fence-in-place (bind a parent-cwd Task and rewrite writes into `.deft-scratch/worktrees/`) stays parked until measured:
+
+- Shell `updated_input.cwd`
+- Write/ApplyPatch path rewrite
+- Composer visibility of gitignored `.deft-scratch/worktrees/`
+
+The write gate does not cover Shell or MCP. Relative targets resolve against hook cwd. Bind is specifiable later (`EXACT_LIFECYCLE_VERBS` + identity rewrite; adding `spawn:bind`). ⊗ Sequence `best-of-n-runner` ahead of nursery. ⊗ Primary-window writes through `.deft-scratch/worktrees/`.
 
 ## Nested Task boundary
 
-! Cursor ownership split (#2797 / #2893) lives in `references/core-phase-3.md` Orchestrator dispatch doctrine — a Cursor `Task` implementation leaf MUST NOT nested-spawn a review-monitor Task.
+! Cursor ownership split (#2797 / #2893) lives in `references/core-phase-3.md` Orchestrator dispatch doctrine — a Cursor implement leaf MUST NOT nested-spawn a review-monitor Task. Dest-rooted SDK / nursery leaves implement in-session.
 
 ## Retained / continue-by-id (#3158)
 
 ! **Default one-shot after Task completion:** Cursor `Task` leaves that exit their tool loop are typically terminal — prefer **split-dispatch** for mid-scope user-approval gates (#954) unless the host surfaces an explicit continue/resume-by-agent-id for that Task.
-? When the host documents resume of the same Task / agent id with context intact, treat as **retain-capable** for message-later / steer-mid-flight and re-message instead of a full second Task spawn.
-! Liveness failures (`task verify:subagent-alive` exit `1` / `REDISPATCH_OK`) still authorize replacement re-dispatch — retain does not override the false-alive contract (#2824).
+? `@cursor/sdk` `Agent.resume` / `agent.send` is retain-capable for the headless dest-rooted path.
+! Liveness failures (`task verify:subagent-alive` exit `1` / `REDISPATCH_OK`) still authorize replacement re-dispatch for Composer heartbeats — retain does not override the false-alive contract (#2824). Headless SDK replacement keys on the process handle.
 ~ Stance: orchestration only (#3164).

package/skills/deft-directive-swarm/references/host-grok-build.md

@@ -55,6 +55,18 @@ If the leaf needs another agent, it stops and reports `BLOCKED`. The parent owns
 ! Heartbeat liveness on the Grok Build hybrid path is required — see `references/core-phase-4.md` Heartbeat liveness check (#1365) and `docs/subagent-heartbeat.md`.
 ! Poll coordination uses worktree state + `get_command_or_subagent_output` (not OpenClaw parent-announce).
 
+## Parent-steer inbox (#4286)
+
+! Grok-build leaves still need a parent-writable steer path because this host has no child prompt and no live `resume_from`. Directive owns that path. Do not wait for an xAI input field.
+
+! Inbox: `<worktree>/.deft-scratch/subagent-steer/<agent-id>.json` (not heartbeat JSON). Child reads on each pollable slice and acks apply-once via `<agent-id>.ack.json`. `task verify:subagent-steer` is the parent-visible unread flag (`STEER_PENDING`). It is not `REDISPATCH_OK`.
+
+! Tool-loop duty: no blocking wait longer than the heartbeat/steer poll interval when the leaf must remain steerable; between slices, read the inbox and rewrite heartbeat. A scratch path does not interrupt a blocked tool.
+
+⊗ Replace split-dispatch for mid-scope approval gates with this inbox.
+⊗ Invent OpenClaw `sessions_yield` or live `resume_from` on this host.
+⊗ Drop a second JSON schema into `.deft-scratch/subagent-status/`.
+
 ## Retained / continue-by-id (#3158)
 
 ! **Default one-shot:** `spawn_subagent` workers that finish their tool loop are observed terminal (`succeeded` / failed); the `agent_id` is not a general message-later inbox. Mid-scope user-approval gates MUST use **split-dispatch** (#954) unless this host later documents continue-by-agent-id.

package/tasks/scm.yml

@@ -93,7 +93,7 @@ tasks:
           ENGINE_CMD: 'scm issue edit {{.CLI_ARGS}}'
 
   issue:design-critique-chip:
-    desc: "[#3642] Exclusive design-critique catalog chip -- task scm:issue:design-critique-chip -- --issue N --chip triage-ready|mechanism-shaped|recut-needed [--repo OWNER/NAME]"
+    desc: "[#3642] Exclusive design-critique catalog chip -- task scm:issue:design-critique-chip -- --issue N --chip mechanism-shaped|in-progress|ingest-ready [--repo OWNER/NAME]"
     dir: '{{.USER_WORKING_DIR}}'
     deps:
       - task: :engine:_ts-build

package/tasks/verify.yml

@@ -703,3 +703,13 @@ tasks:
       - task: :engine:invoke
         vars:
           ENGINE_CMD: 'verify-subagent-alive {{.CLI_ARGS}}'
+
+  subagent-steer:
+    desc: "Parent-visible unread steer flag (#4286): exit 1 is STEER_PENDING, not REDISPATCH_OK. Three-state exit (0 none / 1 unread / 2 config)."
+    dir: '{{.USER_WORKING_DIR}}'
+    deps:
+      - task: :engine:_ts-build
+    cmds:
+      - task: :engine:invoke
+        vars:
+          ENGINE_CMD: 'verify-subagent-steer {{.CLI_ARGS}}'

package/templates/agent-prompt-preamble.md

@@ -97,7 +97,7 @@ Populate `selected_backend` OR `routing_policy` (or both when the operator sets
 **Role-boundary expectations (all providers):** the same boundaries apply whether the worker runs on Composer, Grok Build, Cursor/cloud, Claude Code, OpenClaw, or a future adapter:
 
 - ! `leaf-implementation` workers implement scoped xBRIEF work in their assigned worktree only -- gates (`task check`, file-scope audit, Greptile review cycle) are model-agnostic and MUST still pass.
-- ! **Spawned mutating workers take their own worktree (#4066).** Implement-class spawn must carry `isolation=worktree` or a linked `worktree_path`/`cwd` before occupancy claim. Sharing the primary checkout with a live occupant is refuse, not `occupancy:grant` across hosts. Master/primary occupancy is the exception (`release-cut`, `policy-restore`, operator-directed default-branch work). `--read-only` never claims. On DONE/terminal the dispatcher compare-and-releases the recorded child tree (incarnation + parent-id); do not steal the parent's lease to recover. Grok `spawn_subagent` cannot rewrite PreToolUse input -- pass cwd to the reserved worktree.
+- ! **Spawned mutating workers take their own worktree (#4066 / #4295).** Implement-class spawn must dest-place before occupancy claim. Grok `spawn_subagent` dest is `cwd` only (cannot rewrite PreToolUse input). Claude/Codex keep `isolation=worktree` or a linked path. Cursor Task has no local dest field — dest-placing is a dest-rooted window (nursery inherit) or `@cursor/sdk` `Agent.create({ local: { cwd } })`; do not advertise Task dest keys. Sharing the primary checkout with a live occupant is refuse, not `occupancy:grant` across hosts. Master/primary occupancy is the exception (`release-cut`, `policy-restore`, operator-directed default-branch work). `--read-only` never claims. On DONE/terminal the dispatcher compare-and-releases the recorded child tree (incarnation + parent-id); do not steal the parent's lease to recover.
 - ! `orchestrator`, `review-monitor`, and `merge-release` roles MUST run on strong or review-capable agents; dispatchers MUST NOT route these roles to cheap leaf backends.
 - ⊗ Route a cheap leaf backend onto the merge cascade, Phase 5->6 release gate, conflict-resolution rebase, or review-cycle merge-ready decision -- these are irreversible-damage surfaces that stay on the strong tier regardless of provider.
 
@@ -238,7 +238,7 @@ On Windows, Cursor Task-tool local subagents historically opened a visible `cmd.
 
 **Directive rule for orchestrators on Windows:**
 
-- ! Use **local** Cursor Task swarm workers as the default dispatch path — same as other platforms. Do not route to cloud solely because the host is Windows.
+- ! Use **local dest-placing Cursor** as the default dispatch path — same as other platforms (nursery inherit or dest-rooted `@cursor/sdk` `Agent.create`; not dest-binding Task). Do not route to cloud solely because the host is Windows.
 - ! Parallel local cohorts are allowed; do not force concurrency=1 because of #2563.
 - ~ Prefer the normal warm `task` / `dist/bin.js` path; avoid unnecessary `DEFT_FORCE_TS_BUILD=1` across a parallel cohort.
 - ⊗ Drop or weaken the #2563 `windowsHide` / warm-dist mitigations without a replacement that keeps Windows local swarm workable.
@@ -566,6 +566,13 @@ The parent monitor watches the heartbeat file directly (three-state exit 0 ok /
 
 ! **Recovery dispatch vs stale ritual (#3730 / #2992):** Before a replacement spawn, run `task verify:session-ritual -- --tier=gated`. If it fails, run `task session:start --rearm --session-id=<same>` (or the `deft` form) rather than a full cold ceremony, unless re-arm is ineligible. A ritual deny MUST name `session:start --rearm`. Do not treat a bare deny as a dead end, and do not re-arm under a new session id (that reclaims occupancy and blocks the parent's own writes).
 
+! **Parent-steer inbox (#4286):** Grok Build `spawn_subagent` has no child prompt and no live `resume_from`. That host gap does not invent retain, OpenClaw `sessions_yield`, or a replacement for split-dispatch mid-scope approval gates. Grok-build implementation leaves whose tool loop exceeds ~3 min MUST still have a parent-writable steer path because the host has no other channel.
+
+- Inbox path: `<worktree>/.deft-scratch/subagent-steer/<agent-id>.json` (sibling of heartbeat, not inside `subagent-status/`). Closed schema `deft.subagent.steer.v1`: bound to `agent_id`, `steer_id` apply-once ack, `expires_at`, `writer_kind` `occupancy-owner` | `dispatching-parent`. Ack file: `<agent-id>.ack.json` in the same steer dir. Distinct from the child-owned heartbeat JSON.
+- Tool-loop duty: no blocking wait longer than the heartbeat/steer poll interval (2-3 min) when the leaf must remain steerable. Between slices, read the inbox, apply unread once, rewrite heartbeat. A scratch path does not interrupt a blocked tool. Long `task check` recovery stays REDISPATCH_OK / split-dispatch — not "poll while blocked".
+- Parent-visible unread flag: `task verify:subagent-steer` exit `1` prints `STEER_PENDING`. ⊗ Treat unread steer as missing heartbeat. ⊗ Print `REDISPATCH_OK` from the steer gate. ⊗ Put a second JSON schema in `.deft-scratch/subagent-status/` — `sweepScratchDirs` only reads top-level heartbeat `<agent-id>.json` and skips `deft.subagent.steer*` schema.
+- ⊗ Invent message-later / live `resume_from` on grok-build. ⊗ Treat steer `text` as a new dispatch envelope or constitution edit. Depth: `docs/subagent-heartbeat.md`.
+
 ## 10.55 Rule Authority and Thin Fail-Closed (#3313)
 
 Principle: `main.md` `## Rule Authority [AXIOM]`.

package/templates/agents-entry.md

@@ -31,7 +31,7 @@ Deft is installed in .deft/core/. Full guidelines: .deft/core/main.md
 
 ! Writes: `./xbrief/` (`PROJECT-DEFINITION.xbrief.json`, `plan.xbrief.json`, `specification.xbrief.json`) as `"xBRIEFInfo"` `"version": "0.8"`. Legacy `vbrief/`; `deft migrate:xbrief`.
 ! Completed xBRIEFs are record of *what is*, zero authority over *what to build next* (#3383). Current contract = active xBRIEF + human operator live instruction.
-⊗ Treat a completed xBRIEF as the next-build contract. ⊗ Emit `"version": "0.6"` on new writes.
+⊗ Treat a completed xBRIEF as the next-build contract. ⊗ Emit `"version": "0.6"` on new writes. ⊗ No xBRIEF DeftVersion stamps (#4271).
 
 ## Unmanaged project header (#2065)
 
@@ -86,7 +86,7 @@ Deft is installed in .deft/core/. Full guidelines: .deft/core/main.md
 
 ## Through-merge worker dispatch (#3032)
 
-! On **through merge** / **drive to merge** / land-ship / **drive-to: merge-ready** story intent: parent MUST dispatch a `drive-to: merge-ready` worker (worktree, preflight, pre-pr, review-cycle, merge/`scope:complete`) via the **swarm/solo-worker launch path** even if **cohort size is 1** — parent MUST NOT implement as the leaf. Depth: swarm Phase 0 + skill-pin-policy (#3032 / #1880 Gap C).
+! On **through merge** / **drive to merge** / land-ship / **drive-to: merge-ready** story intent: parent MUST dispatch a `drive-to: merge-ready` worker (worktree, preflight, pre-pr, review-cycle, merge/`scope:complete`) via the **swarm/solo-worker launch path** even if **cohort size is 1** — parent MUST NOT implement as the leaf. Depth: swarm Phase 0 + skill-pin-policy (#3032 / #1880 Gap C). Spawn dest (#4066 / #4295): dest-place implement workers before occupancy. Cursor local is dest-placing (nursery inherit or dest-rooted `@cursor/sdk` `Agent.create`); ⊗ Task dest keys.
 ⊗ Parent conversation implements or babysits product fix/CI loops for drive-to:merge-ready work when background subagent/worktree dispatch is available (#3032).
 ! After leaf announce: tool-first / yield / one short non-repeated answer; ⊗ N>2 near-identical zero-tool (FC14 / #3131). Machine: `evaluateParentTurnShape` (`parent-turn-shape`). Depth: preamble §11 + `docs/openclaw-agent-host.md`.
 
@@ -109,6 +109,10 @@ Deft is installed in .deft/core/. Full guidelines: .deft/core/main.md
 
 ! Mid-scope gates: **split-dispatch** when `agent_id` is terminal; retain-capable hosts (continue-by-agent-id / message-later / steer-mid-flight) MAY re-message the live child. Retention = orchestration only — not constitution self-edit (#3164). Depth: preamble §10; `deft-directive-swarm`. Topology: #3155 nuclear-family. ⊗ Invent retain on one-shot hosts.
 
+## Parent-steer inbox (#4286)
+
+! Grok-build leaves whose tool loop exceeds ~3 min poll `.deft-scratch/subagent-steer/<agent-id>.json` on each pollable slice. `deft verify:subagent-steer` exit 1 is `STEER_PENDING`, not missing-heartbeat takeover. Depth: preamble §10.5; `docs/subagent-heartbeat.md`. ⊗ Replace split-dispatch mid-scope approval gates with this inbox. ⊗ Treat unread steer as REDISPATCH_OK.
+
 ## Review-surface precedence (#2308)
 
 ! Route PR shepherding / review work through `deft-directive-review-cycle` — `.deft/core/.agents/skills/deft-directive-review-cycle/SKILL.md`; host `babysit` / `bugbot` / `security-review` advisory-only (#2308 / #2261).
@@ -135,7 +139,7 @@ Deft is installed in .deft/core/. Full guidelines: .deft/core/main.md
 
 ## Windows PowerShell: multi-line git/gh bodies (#2646 / #2744)
 
-! Multi-line git commit / gh issue|pr|comment bodies: write UTF-8 (no BOM) to OS temp, then `git commit -F` / `gh --body-file` / `deft scm:body:* --body-file`. Issue-body RMW on win32: `deft scm:body:issue:fetch --out-file` then edit the file then `deft scm:body:issue:edit --body-file` (#2607 postcondition verify). ⊗ bash heredocs, `<<<`, inline multi-line `--body`, or PS capture-concat of `gh api --jq .body` (string[]/$OFS destroys bodies — #2087, #2741, #1492). Detail: `.deft/core/scm/github.md` § #2646 / #2744. `ghx` is read-only — mutations stay on live `gh`.
+! Multi-line git commit / gh issue|pr|comment bodies: write UTF-8 (no BOM) to OS temp, then `git commit -F` / `gh --body-file` / `deft scm:body:* --body-file`. Issue-body RMW on win32: `deft scm:body:issue:fetch --out-file` then edit the file then `deft scm:body:issue:edit --body-file` (#2607 postcondition verify). ⊗ bash heredocs, `<<<`, inline multi-line `--body`, or PS capture-concat of `gh api --jq .body` (string[]/$OFS destroys bodies — #2087, #2741, #1492). Detail: `.deft/core/scm/github.md` § #2646 / #2744. `ghx` is read-only — mutations stay on live `gh`. Explicit PR `--body-file` / `--body` / `--fill` skip the GitHub template (#4293): compose the template `Documentation impact` block, then `deft verify:docs-impact -- --body-file` on those same bytes (leftover-complete / finalize-cohort).
 
 ## Contextual guardrails (runtime-detect lazy-load)
 

package/templates/design-critique-brief.md

@@ -12,7 +12,7 @@ Dispatch envelope skeleton for one critic or synthesis pass. Fill the fields. Re
 - Critic role (fresh | resume):
 - Id ceiling (GitHub comment id, inclusive):
 - SHA at dispatch:
-- Run posture (`arc-mode: direct` | `arc-mode: checkout`):
+- Run posture (`arc-mode: no-ingest` | `arc-mode: checkout`):
 - Target (work issue or umbrella):
 - Audit targets (marker ids, comma-separated, or `none`; ids only, no parent rationale):
 - Seat families (N≥3: three claimed families before spawn):

package/templates/swarm-greptile-poller-prompt.md

@@ -94,6 +94,23 @@ m = re.search(
 last_reviewed_sha = m.group("sha") if m else None
 ```
 
+### Thin HTML named state (#4289)
+
+When the body is `<!-- greptile_summary -->` with a parsed Confidence Score and **no** `Last reviewed commit:` line, this is **not** informal-clean and **not** INCOMPLETE_BUT_RATED. Do not scrape the HTML for a SHA.
+
+Pin SHA currency to the already-fetched `Greptile Review` check-run on current HEAD (`completed` + `{success, neutral}`), not `pending_required`. Body SHA still wins when present. Findings come from REST `pulls/<N>/comments` and/or check-run `N comments added` text -- vacuous `detect()` zeros are not a findings pin. Missing findings channel fail-closes (`clean_gate_holdout=findings_channel`). Dirty REST P0/P1 is NEW_P0/P1 without a body SHA.
+
+If that pin or findings channel is missing, restate the open-PR symptom **per surface** (do not bind every-consumer-until-TIMEOUT): blocking `pr:watch` waits to cap TIMEOUT; `--one-shot` is PENDING; `pr:merge-ready` is a parse fail; this poller STALLs; review-cycle Step 6 stays `unknown`. Keep #4288 as the merged/closed terminal.
+
+```python
+thin_html = ("<!-- greptile_summary -->" in body and confidence is not None and last_reviewed_sha is None)
+if thin_html and greptile_terminal:
+    last_reviewed_sha = head_sha
+# Do NOT set findings_channel_present=True here. Derive it from paginated REST
+# pulls comments and/or check-run comments-added in the CLEAN-gate call site.
+```
+
+
 A regex that requires the SHA inline after `Last reviewed commit:` will NEVER match Greptile's actual output -- the poller will fall through every iteration and run to its `{poll_cap_minutes}`-minute cap (Agent D, post-#721 swarm; #727 comment 2 Bug 1).
 
 ### P0/P1 findings detection (TRIPLE-TIER -- #910)
@@ -378,9 +395,14 @@ def evaluate_clean_gate(
     errored,
     terminal_check_run,
     min_confidence=4,
+    findings_channel_present=True,
 ):
     """Return (is_clean, clean_gate_holdout) per the (6)-condition AND gate.
 
+    findings_channel_present defaults True only for markdown Last-reviewed
+    bodies. Thin HTML MUST pass a derived value (paginated REST pulls
+    comments and/or check-run comments-added). Do not CLEAN on the default.
+
     clean_gate_holdout names the FIRST failing condition (in 1/2/3/4/5/6
     order) or None when all six pass. The order is the operative
     contract -- callers MUST NOT reorder the checks or the holdout will
@@ -416,6 +438,8 @@ def evaluate_clean_gate(
         return False, "errored"
     if not terminal_check_run:
         return False, "terminal_check_run"
+    if not findings_channel_present:
+        return False, "findings_channel"
     return True, None
 ```
 
@@ -434,6 +458,57 @@ greptile_terminal = (
 ```
 
 ```python
+import json
+import subprocess
+
+_COMMENTS_ADDED_RE = re.compile(r"(\d+)\s+comments?\s+added", re.I)
+_summary = None
+if greptile_run is not None:
+    _summary = (greptile_run.get("output") or {}).get("summary") or greptile_run.get("summary")
+_m = _COMMENTS_ADDED_RE.search(_summary or "")
+comments_added = int(_m.group(1)) if _m else None
+
+findings_channel_present = True
+if thin_html:
+    findings_channel_present = False
+    rest_fetched = False
+    rest_p0 = 0
+    rest_p1 = 0
+    proc = subprocess.run(
+        ["gh", "api", "--paginate", f"repos/{repo}/pulls/{pr_number}/comments?per_page=100"],
+        capture_output=True,
+        text=True,
+    )
+    items = []
+    if proc.returncode == 0:
+        raw = proc.stdout or "[]"
+        try:
+            parsed = json.loads(raw)
+            items = parsed if isinstance(parsed, list) else []
+            rest_fetched = isinstance(parsed, list)
+        except json.JSONDecodeError:
+            rest_fetched = False
+    for item in items:
+        user = (item or {}).get("user") or {}
+        if user.get("login") != "greptile-apps[bot]":
+            continue
+        commit_id = item.get("commit_id") or ""
+        if head_sha and commit_id and not (
+            str(head_sha).startswith(str(commit_id)) or str(commit_id).startswith(str(head_sha))
+        ):
+            continue
+        text = item.get("body") or ""
+        if '<img alt="P0"' in text:
+            rest_p0 += 1
+        if '<img alt="P1"' in text:
+            rest_p1 += 1
+    findings_channel_present = rest_fetched or comments_added is not None
+    if rest_fetched:
+        p0_count, p1_count = rest_p0, rest_p1
+        has_blocking = rest_p0 + rest_p1 > 0
+    elif comments_added is not None:
+        has_blocking = has_blocking or comments_added > 0
+
 is_clean, clean_gate_holdout = evaluate_clean_gate(
     last_reviewed_sha=last_reviewed_sha,
     head_sha=head_sha,
@@ -442,6 +517,7 @@ is_clean, clean_gate_holdout = evaluate_clean_gate(
     ci_failures=ci_failure_count,
     errored=errored,
     terminal_check_run=greptile_terminal,
+    findings_channel_present=findings_channel_present,
 )
 print(
     f"[poll {{i}}/{{cap}}] last_reviewed_sha={{last_reviewed_sha}} "
@@ -488,7 +564,7 @@ Send to parent:
 
 ### (2) NEW P0/P1 FINDINGS
 
-`last_reviewed_sha` matches HEAD AND `has_blocking` is True. Do NOT exit on P2 -- those are non-blocking style suggestions per `skills/deft-directive-review-cycle/SKILL.md`.
+`last_reviewed_sha` matches HEAD AND `has_blocking` is True. Thin HTML (#4289): `has_blocking` from REST pulls comments / comments-added, and NEW_P0/P1 does not require a body SHA when the Greptile Review check-run is terminal on HEAD. Do NOT exit on P2 -- those are non-blocking style suggestions per `skills/deft-directive-review-cycle/SKILL.md`.
 
 Address the findings per Phase 2 Step 2-3 of the review-cycle skill: read every finding, plan a single coherent batch, run `task check`, commit with message `fix: address Greptile review findings (batch)`, push. After the push, you MAY reset the **poll-wait timer** for the new HEAD (Greptile needs a fresh review window) — but you MUST NOT reset the dual-stop **fix-batch counter** (#2442).