cclaw-cli 0.51.30 → 0.55.2

package/dist/content/subagents.js

@@ -9,6 +9,7 @@ const SUBAGENT_AGENT_NAMES = [
     "researcher",
     "architect",
     "spec-validator",
+    "spec-document-reviewer",
     "slice-implementer",
     "performance-reviewer",
     "compatibility-reviewer",
@@ -16,6 +17,7 @@ const SUBAGENT_AGENT_NAMES = [
     "release-reviewer",
     "planner",
     "product-manager",
+    "product-strategist",
     "critic",
     "reviewer",
     "security-reviewer",
@@ -94,10 +96,10 @@ can enforce phase-appropriate write boundaries. Use separate workers only when t
 
 | Harness | Fallback | Delegation tool | Structured ask | Capability source |
 |---|---|---|---|---|
-| Claude | \`native\` | Task (named subagent_type) | AskUserQuestion | \`npx cclaw-cli doctor\` |
-| Cursor | \`generic-dispatch\` | Task (generic subagent_type: explore/generalPurpose/…) | AskQuestion | \`npx cclaw-cli doctor\` |
-| OpenCode | \`native\` | generated \`.opencode/agents/<agent>.md\` subagents via Task / \`@agent\` mention | \`question\` (permission-gated; \`permission.question: "allow"\`) | \`npx cclaw-cli doctor\` |
-| Codex | \`native\` | generated \`.codex/agents/<agent>.toml\` custom agents via native parallel subagent spawning | \`request_user_input\` (experimental; Plan / Collaboration mode) | \`npx cclaw-cli doctor\` |
+| Claude | \`native\` | Task (named subagent_type) | AskUserQuestion | \`npx cclaw-cli sync\` |
+| Cursor | \`generic-dispatch\` | Task (generic subagent_type: explore/generalPurpose/…) | AskQuestion | \`npx cclaw-cli sync\` |
+| OpenCode | \`native\` | generated \`.opencode/agents/<agent>.md\` subagents via Task / \`@agent\` mention | \`question\` (permission-gated; \`permission.question: "allow"\`) | \`npx cclaw-cli sync\` |
+| Codex | \`native\` | generated \`.codex/agents/<agent>.toml\` custom agents via native parallel subagent spawning | \`request_user_input\` (experimental; Plan / Collaboration mode) | \`npx cclaw-cli sync\` |
 
 **Dispatch rules driven by \`subagentFallback\`:**
 
@@ -144,9 +146,9 @@ Concrete per-stage rules so the controller does not have to guess which tier fit
 | Stage | Deep slot | Balanced slot(s) | Fast fan-out | Trigger to escalate |
 |---|---|---|---|---|
 | brainstorm | planner (only if ambiguity spans >1 module) | product-manager / critic when product value or premise is uncertain | run in-thread research playbooks | promote to \`balanced\` critic if the do-nothing path may beat the idea |
-| scope | planner (always) | product-manager / critic when mode changes user value or boundaries are soft | run \`research/git-history.md\` in-thread when churn is high | promote to \`balanced\` critic if scope mode is disputed |
+| scope | planner (always) | product-manager / product-strategist / critic when mode changes value, trajectory, or boundaries | run \`research/git-history.md\` in-thread when churn is high | promote to \`balanced\` critic if scope mode is disputed |
 | design | planner (always) | critic, security-reviewer, test-author when alternatives/trust/testability apply | run \`research/framework-docs-lookup.md\` + \`research/best-practices-lookup.md\` in-thread | escalate one specialist to \`deep\` only if a failure mode is Critical-severity |
-| spec | — | reviewer (if spec > 200 lines or multiple ACs) | — | escalate to \`deep\` only for spec ↔ design contradictions |
+| spec | — | spec-validator / spec-document-reviewer / reviewer (for long or high-risk specs) | — | escalate to \`deep\` only for spec ↔ design contradictions |
 | plan | planner (solo, always) | — | — | never fan out at plan stage; one owner for dependency graph |
 | tdd | — | ${formatAgentList(stageSummary("tdd").primaryAgents)} (per slice, carrying RED/GREEN/REFACTOR evidence) · reviewer (slice-local only when sliceReview triggers) | doc-updater (API surface changes) | escalate to \`deep\` only when a RED test cannot be expressed (design leak) |
 | review | — | ${formatAgentList(stageSummary("review").mandatoryAgents)} (both mandatory) | doc-updater for release-note drift checks | escalate a \`balanced\` reviewer to \`deep\` only when two reviewers disagree on severity |
@@ -642,6 +644,30 @@ Required output:
 - TESTABILITY_MAP: concrete test level and command/manual evidence per AC
 ${MARKDOWN_CODE_FENCE}
 
+`;
+}
+function specDocumentReviewerEnhancedBody() {
+    return `
+
+## Task Tool Delegation
+
+Use this payload for final spec-document quality checks before plan handoff:
+
+${MARKDOWN_CODE_FENCE}
+You are a spec-document-reviewer subagent.
+
+SPEC_ARTIFACT: {04-spec excerpt or full body}
+UPSTREAM_CONTEXT: {scope/design refs used by the spec}
+
+Required output:
+- DOCUMENT_VERDICT: PASS | PASS_WITH_GAPS | FAIL | BLOCKED
+- COMPLETENESS_CHECK: missing required sections or weakly populated rows
+- CONSISTENCY_CHECK: contradictions across ACs, assumptions, mapping, and approval
+- CLARITY_CHECK: ambiguity/placeholders/two-way wording to rewrite
+- SCOPE_FIT_CHECK: whether the artifact still maps to one coherent subsystem/plan slice
+- PATCH_RECOMMENDATIONS: minimal edits to make the spec plan-ready
+${MARKDOWN_CODE_FENCE}
+
 `;
 }
 function sliceImplementerEnhancedBody() {
@@ -808,6 +834,30 @@ Required output:
 - SCOPE_HANDOFF: one recommendation for hold/selective/expand/reduce
 ${MARKDOWN_CODE_FENCE}
 
+`;
+}
+function productStrategistEnhancedBody() {
+    return `
+
+## Task Tool Delegation
+
+Use this payload for expansion-mode scope strategy checks:
+
+${MARKDOWN_CODE_FENCE}
+You are a product-strategist subagent.
+
+SCOPE_MODE: {SCOPE EXPANSION|SELECTIVE EXPANSION}
+DECISION_CONTEXT: {scope contract excerpt + constraints + approved brainstorm direction}
+DEPTH: {standard|deep}
+
+Required output:
+- VISION_DELTA: 10x trajectory vs hold-scope baseline
+- EXPANSION_PROPOSALS: 2-3 concrete proposals with add/defer/skip recommendation
+- UPSIDE_AND_RISK: strategic upside, reversibility, and principal downside per proposal
+- TRAJECTORY_FIT: whether current architecture trajectory can absorb accepted expansions
+- FINAL_RECOMMENDATION: smallest high-leverage expansion set to lock now
+${MARKDOWN_CODE_FENCE}
+
 `;
 }
 function criticEnhancedBody() {
@@ -971,6 +1021,8 @@ export function enhancedAgentBody(agentName) {
             return architectEnhancedBody();
         case "spec-validator":
             return specValidatorEnhancedBody();
+        case "spec-document-reviewer":
+            return specDocumentReviewerEnhancedBody();
         case "slice-implementer":
             return sliceImplementerEnhancedBody();
         case "performance-reviewer":
@@ -985,6 +1037,8 @@ export function enhancedAgentBody(agentName) {
             return plannerEnhancedBody();
         case "product-manager":
             return productManagerEnhancedBody();
+        case "product-strategist":
+            return productStrategistEnhancedBody();
         case "critic":
             return criticEnhancedBody();
         case "reviewer":

package/dist/content/templates.d.ts

@@ -1,4 +1,4 @@
 export declare const ARTIFACT_TEMPLATES: Record<string, string>;
 export declare const RULEBOOK_MARKDOWN = "# Cclaw Rulebook\n\n## MUST_ALWAYS\n- Follow flow order: brainstorm -> scope -> design -> spec -> plan -> tdd -> review -> ship\n- Require explicit user confirmation after plan before TDD\n- Keep evidence artifacts in `.cclaw/artifacts/`\n- Enforce RED before GREEN in TDD\n- Run two-layer review (spec_compliance and code_quality) before ship\n- Validate all inputs before processing \u2014 never trust external data without sanitization\n- Prefer immutable data patterns and pure functions where the language supports them\n- Follow existing repo conventions, patterns, and directory structure \u2014 match the codebase\n- Verify claims with fresh evidence: \"tests pass\" requires running tests in this message\n- Use conventional commits: `type(scope): description` (feat, fix, refactor, test, docs, chore)\n\n## MUST_NEVER\n- Skip RED phase and jump directly to GREEN in TDD\n- Ship with critical review findings\n- Start implementation during /brainstorm\n- Modify generated cclaw files manually when CLI can regenerate them\n- Commit `.cclaw/` or generated shim files\n- Expose secrets, tokens, API keys, or absolute system paths in agent output\n- Duplicate existing functionality without explicit justification \u2014 search before building\n- Bypass security checks, linting hooks, or type checking to \"move faster\"\n- Claim success (\"Done,\" \"All good,\" \"Tests pass\") without running verification in this message\n- Make changes outside the blast radius of the current task without user consent\n\n## DELEGATION\nWhen a task requires specialist knowledge (security audit, performance profiling, database review),\ndelegate to a specialized agent or skill if the harness supports it. The primary agent should:\n1. Identify the specialist domain\n2. Provide focused context (relevant files, the specific concern)\n3. Evaluate the specialist output before acting on it \u2014 do not blindly apply recommendations\n";
-export declare const CURSOR_WORKFLOW_RULE_MDC = "---\ndescription: cclaw workflow guardrails for Cursor agent sessions\nglobs:\n  - \"**/*\"\nalwaysApply: true\n---\n\n<!-- cclaw-managed-cursor-workflow-rule -->\n\n# Cclaw Workflow Guardrails\n\n## Activation Rule\n\nBefore responding to coding work:\n1. Read `.cclaw/state/flow-state.json`.\n2. Start with `/cc` or continue with `/cc`.\n3. If no software-stage flow applies, respond normally.\n\n## Stage Order\n\n`brainstorm -> scope -> design -> spec -> plan -> tdd -> review -> ship`\n\nTrack-specific skips are allowed only when `flow-state.track` + `skippedStages` explicitly say so.\n\n## Task Classification\n\n| Class | Route |\n|---|---|\n| non-trivial software work | `/cc <idea>` |\n| trivial software fix | `/cc <idea>` (quick track) |\n| bugfix with repro | `/cc <idea>` and enforce RED-first in tdd |\n| pure question / non-software | direct answer (no stage flow) |\n\n## Command Surface\n\n- `/cc` = entry and resume.\n- `/cc` = only progression path.\n- Knowledge capture and recall use the `learnings` skill when requested.\n\n## Verification Discipline\n\n- No completion claim without fresh command evidence in this turn.\n- Do not mark gates passed from memory.\n- Keep evidence in `.cclaw/artifacts/`; archive via `npx cclaw-cli archive`.\n\n## Delegation And Approvals\n\n- Machine-only checks in design/plan/tdd/review/ship should auto-dispatch when tooling supports it.\n- Ask for user input only at explicit approval gates (scope mode, plan approval, challenge resolution, ship finalization).\n- If harness capabilities are partial, record waiver reasons in delegation logs.\n\n## Routing Source Of Truth\n\n- Primary router: `.cclaw/skills/using-cclaw/SKILL.md`.\n- Stage behavior: current stage skill plus `.cclaw/state/flow-state.json`.\n- Preamble budget: keep role/status announcements brief and avoid repeating\n  them unless the stage or role changes.\n";
+export declare const CURSOR_WORKFLOW_RULE_MDC = "---\ndescription: cclaw workflow guardrails for Cursor agent sessions\nglobs:\n  - \"**/*\"\nalwaysApply: true\n---\n\n<!-- cclaw-managed-cursor-workflow-rule -->\n\n# Cclaw Workflow Guardrails\n\n## Activation Rule\n\nBefore responding to coding work:\n1. Read `.cclaw/state/flow-state.json`.\n2. Start with `/cc` or continue with `/cc`.\n3. If no software-stage flow applies, respond normally.\n\n## Stage Order\n\n`brainstorm -> scope -> design -> spec -> plan -> tdd -> review -> ship`\n\nTrack-specific skips are allowed only when `flow-state.track` + `skippedStages` explicitly say so.\n\n## Task Classification\n\n| Class | Route |\n|---|---|\n| non-trivial software work | `/cc <idea>` |\n| trivial software fix | `/cc <idea>` (quick track) |\n| bugfix with repro | `/cc <idea>` and enforce RED-first in tdd |\n| pure question / non-software | direct answer (no stage flow) |\n\n## Command Surface\n\n- `/cc` = entry and resume.\n- `/cc` = only progression path.\n- Knowledge capture and recall use the `learnings` skill when requested.\n\n## Verification Discipline\n\n- No completion claim without fresh command evidence in this turn.\n- Do not mark gates passed from memory.\n- Keep evidence in `.cclaw/artifacts/`; archive through closeout via `/cc` or cancel early via `node .cclaw/hooks/cancel-run.mjs`.\n\n## Delegation And Approvals\n\n- Machine-only checks in design/plan/tdd/review/ship should auto-dispatch when tooling supports it.\n- Ask for user input only at explicit approval gates (scope mode, plan approval, challenge resolution, ship finalization).\n- If harness capabilities are partial, record waiver reasons in delegation logs.\n\n## Routing Source Of Truth\n\n- Primary router: `.cclaw/skills/using-cclaw/SKILL.md`.\n- Stage behavior: current stage skill plus `.cclaw/state/flow-state.json`.\n- Preamble budget: keep role/status announcements brief and avoid repeating\n  them unless the stage or role changes.\n";
 export declare function buildRulesJson(): Record<string, unknown>;

package/dist/content/templates.js

@@ -56,26 +56,6 @@ export const ARTIFACT_TEMPLATES = {
 - **Direct path?** (yes/no + one-line justification)
 - **What if we do nothing?** (concrete consequence, not "nothing happens")
 
-## Forcing Questions
-> Minimum 3 questions; each answer MUST contain at least one *specific* token: a concrete name, a role, a number, a repo-relative path, an external link, or a verbatim quote. Vague answers fail the linter.
-
-| # | Forcing question | Specific answer | Decision impact | Q\\<n\\> decision |
-|---|---|---|---|---|
-| 1 |  |  |  | decision: |
-| 2 |  |  |  | decision: |
-| 3 |  |  |  | decision: |
-
-## Premise List
-> ≥2 premises. Each premise must be in the form \`P<n>: <statement> — agreed | disagreed | revised\`. \`revised\` rows must include the revised statement on the next line.
-
-- P1: <statement> — agreed | disagreed | revised
-- P2: <statement> — agreed | disagreed | revised
-
-## Anti-Sycophancy Stamp
-- **Forbidden response openers acknowledged:** yes (no "you're absolutely right", "great point", "absolutely!", etc.)
-- **Posture commitment:** push back with reasoning when premises feel weak; do not perform agreement.
-- **Evidence-that-would-change-the-recommendation:** (one line per premise, or \`- None.\`)
-
 ## How Might We
 - *How might we …?* — one line naming the user, the desired outcome, and the binding constraint.
 
@@ -151,6 +131,11 @@ RECOMMENDATION: <approach letter — one-line rationale, traced to forcing-quest
 - **Approval:** pending
 - **Next-stage handoff:** On standard track, hand this to \`scope\`; on medium track, hand this directly to \`spec\`. Include upstream decisions used, drift, confidence, unresolved questions, risk hints, and non-goals.
 
+## Embedded Grill
+| Question | Recommended answer | Disposition (accept/refine/reject) |
+|---|---|---|
+|  |  |  |
+
 ## Not Doing
 - (3-5 things this brainstorm is *not* committing to — distinct from \`Deferred\`. These will not appear in scope unless the user explicitly opts in.)
 
@@ -175,6 +160,24 @@ ${SEED_SHELF_SECTION}
 - **Assumptions:**
 - **Open questions (or "None"):**
 
+## Victory Detector
+- Loop status file: \`.cclaw/state/early-loop.json\`
+- Iteration: <n>/<maxIterations>
+- Open concerns: <count>
+- Convergence guard: clear | tripped
+- Decision: iterate | ready_for_advance | escalate_for_human_override
+
+## Critic Pass
+After each producer iteration append one JSON line to \`.cclaw/state/early-loop-log.jsonl\`:
+
+${MARKDOWN_CODE_FENCE}json
+{"ts":"<ISO-8601>","runId":"<run-id>","stage":"brainstorm","iteration":1,"concerns":[{"id":"C-1","severity":"important","locator":"Approaches > Row 2","summary":"<specific concern>"}],"resolvedConcernIds":[]}
+${MARKDOWN_CODE_FENCE}
+
+- \`concerns\` lists unresolved concerns after this pass.
+- Keep concern IDs stable across iterations so convergence checks stay meaningful.
+- Include \`resolvedConcernIds\` when a previously open concern is explicitly closed.
+
 ## Learnings
 - None this stage.
 `,
@@ -211,13 +214,6 @@ ${SEED_SHELF_SECTION}
 | Existing-code leverage? |  |  |
 | Reversibility cost? |  |  |
 
-## Dream State Mapping
-- Deep/optional only; omit for compact scope.
-- CURRENT STATE:
-- THIS PLAN:
-- 12-MONTH IDEAL:
-- Alignment verdict:
-
 ## Implementation Alternatives
 | Option | Summary | Effort (S/M/L/XL) | Risk (Low/Med/High) | Pros | Cons | Reuses |
 |---|---|---|---|---|---|---|
@@ -227,27 +223,6 @@ ${SEED_SHELF_SECTION}
 
 RECOMMENDATION: <option letter — one-line rationale tying back to premise challenge and existing-code leverage>
 
-## Failure Modes Registry
-> Universal failure-mode shape — applies to CLI, library, infra, web, batch jobs.
-
-| Codepath | Failure mode | Rescued? (yes/no) | Test? (unit/integration/e2e) | User sees? (message/silent/N/A) | Logged? (level/none) | Q\\<n\\> decision |
-|---|---|---|---|---|---|---|
-|  |  |  |  |  |  | decision: |
-
-## Reversibility Rating
-- Score (1-5, 1 = one-way door / unrecoverable, 5 = trivially reversible):
-- Justification (cite a specific artifact/file or migration step):
-- Rollback plan reference:
-
-## Temporal Interrogation
-- Deep/optional only; omit for compact scope.
-| Time slice | Likely decision pressure | Lock now or defer? | Reason |
-|---|---|---|---|
-| HOUR 1 (foundations) |  |  |  |
-| HOUR 2-3 (core logic) |  |  |  |
-| HOUR 4-5 (integration) |  |  |  |
-| HOUR 6+ (polish/tests) |  |  |  |
-
 ## Scope Contract
 - **Selected mode:** HOLD SCOPE | SELECTIVE EXPANSION | SCOPE EXPANSION | SCOPE REDUCTION
 - **In scope:**
@@ -300,7 +275,7 @@ RECOMMENDATION: <option letter — one-line rationale tying back to premise chal
 |  |  |  |  |
 
 ## Reference Pull
-- Optional evidence from \`/Users/zuevrs/Downloads/references\`; list accepted/rejected ideas or \`Not needed - compact scope\`.
+- Optional evidence from \`<repo-relative references dir>\`; list accepted/rejected ideas or \`Not needed - compact scope\`.
 
 ## Ambitious Alternatives
 - Optional for SCOPE EXPANSION/SELECTIVE; list larger alternatives and disposition.
@@ -381,6 +356,24 @@ ${SEED_SHELF_SECTION}
 - Explicitly excluded:
 - Next-stage handoff: identify whether the next stage is \`design\` (standard track) or \`spec\` (medium track), and list the exact artifacts/decisions it must carry forward.
 
+## Victory Detector
+- Loop status file: \`.cclaw/state/early-loop.json\`
+- Iteration: <n>/<maxIterations>
+- Open concerns: <count>
+- Convergence guard: clear | tripped
+- Decision: iterate | ready_for_advance | escalate_for_human_override
+
+## Critic Pass
+After each producer iteration append one JSON line to \`.cclaw/state/early-loop-log.jsonl\`:
+
+${MARKDOWN_CODE_FENCE}json
+{"ts":"<ISO-8601>","runId":"<run-id>","stage":"scope","iteration":1,"concerns":[{"id":"C-1","severity":"important","locator":"Scope Contract > In Scope","summary":"<specific concern>"}],"resolvedConcernIds":[]}
+${MARKDOWN_CODE_FENCE}
+
+- \`concerns\` lists unresolved concerns after this pass.
+- Keep concern IDs stable across iterations so convergence checks stay meaningful.
+- Include \`resolvedConcernIds\` when a previously open concern is explicitly closed.
+
 ## Learnings
 - None this stage.
 `,
@@ -492,31 +485,17 @@ ${MARKDOWN_CODE_FENCE}
 (failure detection -> rescue action -> user-visible outcome)
 ${MARKDOWN_CODE_FENCE}
 
-## State Machine Diagram
-- Deep add-on; omit for compact design.
+## Deep Diagram Add-on
+- Deep add-on only.
+- type: state-machine | rollback | deployment-sequence | none
+- Choose exactly one marker below when type is not \`none\`.
 
 <!-- diagram: state-machine -->
-
-${MARKDOWN_CODE_FENCE}
-(state transitions for the critical flow lifecycle)
-${MARKDOWN_CODE_FENCE}
-
-## Rollback Flowchart
-- Deep add-on; omit for compact design.
-
 <!-- diagram: rollback-flowchart -->
-
-${MARKDOWN_CODE_FENCE}
-(trigger -> rollback actions -> verification)
-${MARKDOWN_CODE_FENCE}
-
-## Deployment Sequence Diagram
-- Deep add-on; omit for compact design.
-
 <!-- diagram: deployment-sequence -->
 
 ${MARKDOWN_CODE_FENCE}
-(rollout order, guard checks, and verification sequence)
+(deep add-on diagram: state lifecycle OR rollback flow OR deployment sequence)
 ${MARKDOWN_CODE_FENCE}
 
 ## Stale Diagram Audit
@@ -571,32 +550,6 @@ ${MARKDOWN_CODE_FENCE}
 |---|---|---|---|
 |  |  |  | covered/gap |
 
-## ASCII Coverage Diagram
-
-<!-- diagram: ascii-coverage -->
-
-${MARKDOWN_CODE_FENCE}
-entry-point
-  ├── happy path           [★★★]
-  ├── empty input          [★★]
-  ├── error path           [★]
-  ├── concurrency edge     [GAP]
-  ├── slow-network edge    [→E2E]
-  └── perf regression      [→EVAL]
-${MARKDOWN_CODE_FENCE}
-
-> Required marker tokens (at least one each present where applicable): \`[★★★]\` / \`[★★]\` / \`[★]\` / \`[GAP]\` / \`[→E2E]\` / \`[→EVAL]\`. The diagram is the single source of truth for coverage; gaps must be traced into Plan or Spec.
-
-## Regression Iron Rule
-- Iron rule acknowledged: yes — any diff that changes existing behavior gets a regression test added to the plan, no exceptions.
-- Detected behavior changes (or \`- None.\`):
-- Regression test handoff (Plan task ID or \`- None.\`):
-
-## Calibrated Findings
-> Format: \`[P1|P2|P3] (confidence: <n>/10) <repo-relative-path>[:<line>] — <one-line description>\`. Findings with confidence \`< 7\` are suppressed unless severity is \`P1\`.
-
-- (or \`- None this stage.\`)
-
 ## Performance Budget
 | Critical path | Metric | Target | Measurement method |
 |---|---|---|---|
@@ -629,6 +582,9 @@ ${MARKDOWN_CODE_FENCE}
 - Test/performance expectations:
 - Unresolved questions (or \`None\`):
 
+## Long-Term Trajectory
+- (1-3 lines: what comes after this ships, and whether this architecture can absorb that path without major rework. For tactical changes use \`None - tactical change only\`.)
+
 ## Outside Voice Findings
 | ID | Dimension | Finding | Disposition | Rationale |
 |---|---|---|---|---|
@@ -689,8 +645,23 @@ ${SEED_SHELF_SECTION}
 
 **Decisions made:** 0 | **Unresolved:** 0
 
-## Learning Capture Hint
-For meaningful design work, replace the Learnings sentinel with 1-3 JSON learning bullets, for example: \`- {"type":"lesson","trigger":"when design chooses a risky fallback path","action":"record the switch trigger and rollback signal in Spec Handoff","confidence":"medium","domain":"architecture","stage":"design"}\`
+## Victory Detector
+- Loop status file: \`.cclaw/state/early-loop.json\`
+- Iteration: <n>/<maxIterations>
+- Open concerns: <count>
+- Convergence guard: clear | tripped
+- Decision: iterate | ready_for_advance | escalate_for_human_override
+
+## Critic Pass
+After each producer iteration append one JSON line to \`.cclaw/state/early-loop-log.jsonl\`:
+
+${MARKDOWN_CODE_FENCE}json
+{"ts":"<ISO-8601>","runId":"<run-id>","stage":"design","iteration":1,"concerns":[{"id":"C-1","severity":"important","locator":"Engineering Lock > Row 1","summary":"<specific concern>"}],"resolvedConcernIds":[]}
+${MARKDOWN_CODE_FENCE}
+
+- \`concerns\` lists unresolved concerns after this pass.
+- Keep concern IDs stable across iterations so convergence checks stay meaningful.
+- Include \`resolvedConcernIds\` when a previously open concern is explicitly closed.
 
 ## Learnings
 - None this stage.
@@ -773,12 +744,6 @@ For meaningful design work, replace the Learnings sentinel with 1-3 JSON learnin
 |---|---|---|
 |  |  |  |
 
-## Testing Strategy
-- Behaviors covered (not implementation):
-- Integration vs. unit split (and why):
-- Real services vs. doubles (and why):
-- Coverage gaps with rationale (or \`- None.\`):
-
 ## Spec Self-Review
 > Inline pass; fix in place. If a check fails, do not move on without recording the fix.
 
@@ -791,13 +756,6 @@ For meaningful design work, replace the Learnings sentinel with 1-3 JSON learnin
 - Remaining concerns:
   - None
 
-## Reviewer Concerns (convergence guard)
-> Populate ONLY if the spec review loop did not converge after 3 iterations. Each row links a concern to the unresolved review pass.
-
-| ID | Concern | Reviewer / source | Disposition (open/accept/defer) | Rationale |
-|---|---|---|---|---|
-|  |  |  |  |  |
-
 ## Approval
 - Approved by:
 - Date:
@@ -906,23 +864,16 @@ Execution rule: complete and verify each batch before starting the next batch.
   - [ ] Step 4: run test, observe GREEN
   - [ ] Step 5: refactor + commit
 
-## High-Level Technical Design
-> "Directional guidance, not implementation specification." Choose the form that fits the work: pseudo-code grammar, mermaid sequence/state, data-flow ASCII, decision matrix. Skip if the plan is a pure rename/move.
-
-\`\`\`
-(pseudo-code, mermaid, ASCII data flow, or decision matrix)
-\`\`\`
+## Calibrated Findings
+> Recommended for high-risk or multi-batch plans. Use one line per finding, or \`None this stage\`.
+>
+> Canonical format:
+> \`- [P1|P2|P3] (confidence: <n>/10) <path>[:<line>] — <description>\`
 
-## Plan Self-Review
-- [ ] Spec coverage: every spec behavior maps to a unit/task
-- [ ] Placeholder scan (regex on full artifact, not only Task List)
-- [ ] Type/name consistency across units (signatures referenced match definitions)
-- [ ] No silent scope reduction
-- [ ] Confidence per unit recorded (1-10)
-- Patches applied:
-  - None
-- Remaining concerns:
-  - None
+## Regression Iron Rule
+- Iron rule acknowledged: yes
+- Critical regression guardrail:
+- Required pre-merge proof:
 
 ## Execution Handoff
 - **Posture chosen:** Subagent-Driven (recommended) | Inline executor
@@ -930,13 +881,13 @@ Execution rule: complete and verify each batch before starting the next batch.
 - **Subagent recipe (if Subagent-Driven):** \`<harness>\` -> \`<dispatch surface>\` -> \`<agent-definition path>\` (substitute neutral placeholders; full recipes in \`docs/harnesses.md\`)
 - **Inline recipe (if Inline executor):** TDD loop unit-by-unit with batch checkpoints
 
-## No-Placeholder Scan
-- Scanned tokens: \`TODO\`, \`TBD\`, \`FIXME\`, \`<fill-in>\`, \`<your-*-here>\`, \`xxx\`, bare ellipsis in task rows.
-- Hits: 0 (required for WAIT_FOR_CONFIRM to resolve).
-
-## No Scope Reduction Language Scan
-- Scanned phrases: \`v1\`, \`for now\`, \`later\`, \`temporary\`, \`placeholder\`, \`mock for now\`, \`hardcoded for now\`, \`will improve later\`.
-- Hits: 0 (required when Locked Decisions section is non-empty; use LD#hash anchors).
+## Plan Quality Scan
+- Placeholder scan:
+  - Scanned tokens: \`TODO\`, \`TBD\`, \`FIXME\`, \`<fill-in>\`, \`<your-*-here>\`, \`xxx\`, bare ellipsis in task rows.
+  - Hits: 0 (required for WAIT_FOR_CONFIRM to resolve).
+- Scope reduction language scan:
+  - Scanned phrases: \`v1\`, \`for now\`, \`later\`, \`temporary\`, \`placeholder\`, \`mock for now\`, \`hardcoded for now\`, \`will improve later\`.
+  - Hits: 0 (required when Locked Decisions section is non-empty; use LD#hash anchors).
 
 ## WAIT_FOR_CONFIRM
 - Status: pending
@@ -1028,18 +979,27 @@ Execution rule: complete and verify each batch before starting the next batch.
 |---|---|---|---|
 | S-1 |  |  |  |
 
-## Anti-Rationalization Checks
-- [ ] No "test passes immediately" — each new test was watched failing first
-- [ ] No "code before test" reuse from a prior session
-- [ ] No "tests after" backfill instead of RED-first
-- [ ] No "spirit not ritual" overrides
-- Notes (or \`- None this stage.\`):
-
 ## Verification Ladder
 | Slice | Tier reached | Evidence |
 |---|---|---|
 | S-1 |  |  |
 
+## TDD Blocker Taxonomy
+> Fill only when blocked. Choose one canonical class and capture the resume contract.
+>
+> Allowed classes: \`NO_SOURCE_CONTEXT\`, \`NO_TEST_SURFACE\`, \`NO_IMPLEMENTABLE_SLICE\`, \`RED_NOT_EXPRESSIBLE\`, \`NO_VCS_MODE\`.
+
+| Slice | Class | blockedBecause | missingInputs | recommendedRoute | nextCommand | resumeCriteria |
+|---|---|---|---|---|---|---|
+| S-1 |  |  |  |  |  |  |
+
+## Per-Slice Review
+> Only required when \`.cclaw/config.yaml::sliceReview.enabled\` triggers this slice; otherwise record \`not triggered\`.
+
+| Slice | Trigger fired | Spec-Compliance pass | Quality pass | Fulfillment mode (isolated/role-switch/not triggered) |
+|---|---|---|---|---|
+| S-1 |  |  |  |  |
+
 ## Coverage Targets
 | Code type | Target | Current | Command |
 |---|---|---|---|
@@ -1059,9 +1019,6 @@ Execution rule: complete and verify each batch before starting the next batch.
 |---|---|---|---|---|
 | S-1 |  |  |  |  |
 
-## Learning Capture Hint
-For meaningful TDD work, replace the Learnings sentinel with 1-3 JSON learning bullets, for example: \`- {"type":"pattern","trigger":"when a regression only fails after state rewind","action":"keep the RED fixture and add a cycle-log assertion before GREEN","confidence":"medium","domain":"testing","stage":"tdd"}\`
-
 ## Learnings
 - None this stage.
 `,
@@ -1076,15 +1033,13 @@ For meaningful TDD work, replace the Learnings sentinel with 1-3 JSON learning b
 - Open questions:
 - Drift from upstream (or \`None\`):
 
-## Self-Review First
+## Pre-Critic Self-Review
 - [ ] Build/lint/type-check/tests passed locally
 - [ ] Diff matches spec/plan (no scope creep)
 - [ ] Leftover prints / commented code / unused imports removed
 - [ ] Deletion test: each new module justifies its existence
 - Evidence (commands + result):
 - Patches applied (or \`- None.\`):
-
-## Frame the Review Request
 - **Goal:**
 - **Approach:**
 - **Risk areas:**
@@ -1176,7 +1131,7 @@ For meaningful TDD work, replace the Learnings sentinel with 1-3 JSON learning b
 - Overall: complete | concerns | blocked
 
 ## Trace Matrix Check
-- Command: \`cclaw internal trace-matrix\` when the active track enforces it; otherwise record direct AC/reproduction-slice coverage.
+- Command: \`npx cclaw-cli internal trace-matrix\` when the active track enforces it; otherwise record direct AC/reproduction-slice coverage.
 - Orphaned criteria: 0
 - Orphaned source items: 0 or \`N/A - direct spec/reproduction coverage\`
 - Orphaned tests: 0
@@ -1201,9 +1156,6 @@ For meaningful TDD work, replace the Learnings sentinel with 1-3 JSON learning b
 ## Final Verdict
 - APPROVED | APPROVED_WITH_CONCERNS | BLOCKED
 
-## Learning Capture Hint
-For meaningful review work, replace the Learnings sentinel with 1-3 JSON learning bullets, for example: \`- {"type":"lesson","trigger":"when security sweep finds no issues but touches trust boundaries","action":"record NO_SECURITY_IMPACT with inspected surfaces and rationale","confidence":"medium","domain":"security","stage":"review"}\`
-
 ## Learnings
 - None this stage.
 `,
@@ -1438,7 +1390,7 @@ Track-specific skips are allowed only when \`flow-state.track\` + \`skippedStage
 
 - No completion claim without fresh command evidence in this turn.
 - Do not mark gates passed from memory.
-- Keep evidence in \`.cclaw/artifacts/\`; archive via \`npx cclaw-cli archive\`.
+- Keep evidence in \`.cclaw/artifacts/\`; archive through closeout via \`/cc\` or cancel early via \`node .cclaw/hooks/cancel-run.mjs\`.
 
 ## Delegation And Approvals
 

package/dist/content/tree-command.js

@@ -31,7 +31,7 @@ Do not modify state in this command. It is a pure read/render operation.
    \`fulfillmentMode\` / \`evidenceRefs\`.
 3. Inspect \`${artifactsPath()}\` for per-stage artifact presence and stale copies,
    and for the retro artifact \`09-retro.md\`.
-4. Use \`npx cclaw-cli doctor --explain\` for harness capability status when needed.
+4. Use \`npx cclaw-cli sync\` for harness capability status when needed.
 5. Render one compact tree:
    - stage marker: passed/current/pending/skipped/stale,
    - gates summary,
@@ -47,7 +47,7 @@ Do not modify state in this command. It is a pure read/render operation.
      then \`next\`; the transient \`archived\` substate surfaces only if the
      archive step failed mid-run.
 7. Append a \`harnesses:\` branch. For each installed harness derive the tier
-   and fallback from cclaw capability metadata; use \`npx cclaw-cli doctor --explain\`
+   and fallback from cclaw capability metadata; use \`npx cclaw-cli sync\`
    for remediation details when needed.
 8. If rewind records exist in \`${rewindLogPath()}\`, include latest rewind note in footer.
 

package/dist/content/utility-skills.d.ts

@@ -18,13 +18,13 @@ export declare const LANGUAGE_RULE_PACK_FILES: {
 };
 /**
  * Folder (relative to runtime root) that holds every enabled language rule
- * pack. A single folder keeps discovery trivial for hooks and for `doctor`.
+ * pack. A single folder keeps discovery trivial for hooks and for `sync`.
  */
 export declare const LANGUAGE_RULE_PACK_DIR: readonly ["rules", "lang"];
 export declare const LANGUAGE_RULE_PACK_GENERATORS: Record<string, () => string>;
 /**
  * Legacy per-language folders under `.cclaw/skills/` used in v0.7.0. Listed
- * here so `cclaw sync` and `doctor` can surface drift and the installer can
+ * here so `cclaw sync` and `sync` can surface drift and the installer can
  * clean them up after the move to `.cclaw/rules/lang/`.
  */
 export declare const LEGACY_LANGUAGE_RULE_PACK_FOLDERS: readonly ["language-typescript", "language-python", "language-go"];

package/dist/content/utility-skills.js

@@ -211,7 +211,7 @@ export const LANGUAGE_RULE_PACK_FILES = {
 };
 /**
  * Folder (relative to runtime root) that holds every enabled language rule
- * pack. A single folder keeps discovery trivial for hooks and for `doctor`.
+ * pack. A single folder keeps discovery trivial for hooks and for `sync`.
  */
 export const LANGUAGE_RULE_PACK_DIR = ["rules", "lang"];
 export const LANGUAGE_RULE_PACK_GENERATORS = {
@@ -221,7 +221,7 @@ export const LANGUAGE_RULE_PACK_GENERATORS = {
 };
 /**
  * Legacy per-language folders under `.cclaw/skills/` used in v0.7.0. Listed
- * here so `cclaw sync` and `doctor` can surface drift and the installer can
+ * here so `cclaw sync` and `sync` can surface drift and the installer can
  * clean them up after the move to `.cclaw/rules/lang/`.
  */
 export const LEGACY_LANGUAGE_RULE_PACK_FOLDERS = [

package/dist/content/view-command.js

@@ -32,8 +32,10 @@ ${conversationLanguagePolicyMarkdown()}
    - \`diff\` -> use the **Diff Subcommand** section in \`${RUNTIME_ROOT}/skills/${VIEW_SKILL_FOLDER}/SKILL.md\`
 3. Unknown subcommand -> print supported values and stop.
 
-## Headless mode
+## Headless mode (CI/automation only)
 
+Headless envelopes are a machine-mode exception for CI/automation orchestration.
+In normal interactive runs, respond with concise read-only prose instead.
 For machine orchestration, emit one JSON envelope:
 
 \`\`\`json
@@ -43,7 +45,7 @@ For machine orchestration, emit one JSON envelope:
 Use the parsed/defaulted subcommand in both \`payload.command\` and \`payload.subcommand\`; do not collapse \`tree\` or \`diff\` responses to \`status\`.
 
 Validate envelopes with:
-\`cclaw internal envelope-validate --stdin\`
+\`npx cclaw-cli internal envelope-validate --stdin\`
 
 ## Primary skill
 

package/dist/delegation.d.ts

@@ -34,6 +34,7 @@ export interface DelegationTokenUsage {
     output: number;
     model: string;
 }
+export type DelegationWaiverAcceptedBy = "user-flag";
 export type DelegationEntry = {
     stage: string;
     agent: string;
@@ -56,6 +57,7 @@ export type DelegationEntry = {
      */
     taskId?: string;
     waiverReason?: string;
+    acceptedBy?: DelegationWaiverAcceptedBy;
     ts?: string;
     /**
      * Run id the entry belongs to. Older ledgers written before 0.5.17 may omit this;

package/dist/delegation.js

@@ -198,6 +198,7 @@ function isDelegationEntry(value) {
         (o.endTs === undefined || typeof o.endTs === "string") &&
         (o.taskId === undefined || typeof o.taskId === "string") &&
         (o.waiverReason === undefined || typeof o.waiverReason === "string") &&
+        (o.acceptedBy === undefined || o.acceptedBy === "user-flag") &&
         waiverOk &&
         (o.runId === undefined || typeof o.runId === "string") &&
         (o.fulfillmentMode === undefined ||
@@ -269,7 +270,7 @@ function parseLedger(raw, runId) {
                 // and entry, the entry has no fulfillmentMode, and there is no
                 // dispatch-surface or dispatch-id evidence on the row. We honor
                 // that by tagging fulfillmentMode = "legacy-inferred" so callers
-                // (stage-complete, doctor) can require an explicit `--rerecord`
+                // (stage-complete, sync/runtime checks) can require an explicit `--rerecord`
                 // before the row counts as proof-era.
                 const ledgerHasNoVersion = ledgerSchemaVersion === undefined || ledgerSchemaVersion === 1;
                 const entryHasNoVersion = item.schemaVersion === undefined || item.schemaVersion === 1;