cclaw-cli 0.48.31 → 0.48.33

package/dist/content/stages/plan.js

@@ -100,6 +100,11 @@ export const PLAN = {
             "WAIT_FOR_CONFIRM present and unresolved until user approves",
             "artifact ready for TDD execution",
             "acceptance mapping complete"
+        ],
+        platformNotes: [
+            "Per-task verification commands must be runnable on Windows PowerShell, macOS bash/zsh, and Linux bash. Prefer `npm run <script>` / `pnpm <script>` / `pytest -k <name>` over raw shell one-liners so the command portability is handled by the script runner.",
+            "If a task command needs globbing, wrap the glob in single quotes on POSIX and escape as needed on PowerShell (`'src/**/*.ts'` vs `\"src/**/*.ts\"`). Note the quoting variant when the task is expected to run in mixed-OS CI.",
+            "Environment variables referenced from tasks must be named in uppercase with underscores (`CCLAW_PROJECT_ROOT`) and set via a cross-shell wrapper (e.g. `cross-env` for Node tasks) — do not inline `KEY=value cmd` style that fails in PowerShell/cmd.exe."
         ]
     },
     artifactRules: {

package/dist/content/stages/review.js

@@ -100,6 +100,11 @@ export const REVIEW = {
             "all review sections evaluated",
             "critical blockers resolved",
             "ship readiness explicitly stated"
+        ],
+        platformNotes: [
+            "When citing file locations in findings, use repo-relative forward-slash paths with a line number (`src/foo/bar.ts:42`). Avoid IDE-generated hyperlinks that embed absolute machine-specific paths.",
+            "Line-range or diff-range references must match `git diff --unified=0` output format so reviewers on any OS can reproduce the range locally without GUI tooling.",
+            "Commands in remediation suggestions must be portable (`npm run lint`, `pytest -x path/to/test`) — if a platform-specific command is required, tag the note explicitly (`# PowerShell only`, `# macOS only`)."
         ]
     },
     artifactRules: {

package/dist/content/stages/schema-types.d.ts

@@ -50,6 +50,14 @@ export interface StagePhilosophy {
 export interface StageExecutionModel {
     interactionProtocol: string[];
     process: string[];
+    /**
+     * Optional custom mermaid `flowchart` body (without the fenced `mermaid`
+     * code block) that overrides the auto-generated linear flowchart in the
+     * rendered `## Process` section. Use for stages whose state machine is
+     * non-linear (loops, conditional branches) — otherwise leave unset and
+     * let the renderer derive a simple `A --> B --> C` chart from `process`.
+     */
+    processFlow?: string;
     checklist: string[];
     requiredGates: StageGate[];
     requiredEvidence: string[];
@@ -58,6 +66,13 @@ export interface StageExecutionModel {
     researchPlaybooks?: string[];
     blockers: string[];
     exitCriteria: string[];
+    /**
+     * Optional platform-specific notes (Windows/macOS/Linux path separators,
+     * PowerShell vs cmd, harness-specific tool names). Rendered under
+     * "## Platform Notes" when present. Omit when the stage is
+     * platform-agnostic.
+     */
+    platformNotes?: string[];
 }
 export interface StageArtifactRules {
     artifactFile: string;
@@ -70,10 +85,18 @@ export interface StageReviewLens {
     outputs: string[];
     reviewSections: ReviewSection[];
     mandatoryDelegations: string[];
+    reviewLoop?: StageReviewLoop;
 }
 export interface StageReviewLensInput {
     outputs: string[];
     reviewSections: ReviewSection[];
+    reviewLoop?: StageReviewLoop;
+}
+export interface StageReviewLoop {
+    stage: "scope" | "design";
+    checklist: string[];
+    maxIterations: number;
+    targetScore: number;
 }
 export interface StageSchema {
     schemaShape: "v2";
@@ -100,6 +123,10 @@ export interface StageSchema {
     whenNotToUse: string[];
     interactionProtocol: string[];
     process: string[];
+    /** See {@link StageExecutionModel.processFlow}. */
+    processFlow?: string;
+    /** See {@link StageExecutionModel.platformNotes}. */
+    platformNotes?: string[];
     requiredGates: StageGate[];
     requiredEvidence: string[];
     inputs: string[];
@@ -130,6 +157,8 @@ export interface StageSchema {
     trivialOverrideSections?: string[];
     /** Agent names that MUST be dispatched (or waived) before stage transition — derived from mandatory auto-subagent rows. */
     mandatoryDelegations: string[];
+    /** Optional shared outside-voice loop config for scope/design stages. */
+    reviewLoop?: StageReviewLoop;
 }
 export type StageSchemaLegacyInput = Omit<StageSchema, "schemaShape" | "philosophy" | "executionModel" | "artifactRules" | "reviewLens" | "mandatoryDelegations" | "complexityTier"> & {
     schemaShape?: "legacy";

package/dist/content/stages/scope.js

@@ -1,3 +1,4 @@
+import { REVIEW_LOOP_CHECKLISTS } from "../review-loop.js";
 // ---------------------------------------------------------------------------
 // SCOPE — reference: gstack CEO review
 // ---------------------------------------------------------------------------
@@ -43,7 +44,7 @@ export const SCOPE = {
     },
     executionModel: {
         checklist: [
-            "**Pre-Scope System Audit** — before premise challenge, gather reality snapshot: recent commits (`git log -30 --oneline`), current diff (`git diff --stat`), stash state (`git stash list`), and deferred debt markers (`rg -n 'TODO|FIXME|XXX|HACK'`). Record findings in scope artifact.",
+            "**Pre-Scope System Audit (opt-in)** — when `.cclaw/config.yaml::optInAudits.scopePreAudit` is true, before premise challenge gather reality snapshot: recent commits (`git log -30 --oneline`), current diff (`git diff --stat`), stash state (`git stash list`), and deferred debt markers (`rg -n 'TODO|FIXME|XXX|HACK'`). Record findings in scope artifact.",
             "**Assess complexity** — Read the brainstorm artifact. If project is simple (single component, clear architecture, personal/prototype), run light-touch scope: mode selection, 3-5 key in/out boundaries, deferred items. Skip Dream State Mapping and Temporal Interrogation. If project is complex (multi-component, team delivery, production), run the full checklist.",
             "**Prime Directives** — Zero silent failures. For each in-scope capability, name concrete failure modes, the exact error surface, and trace all four data-flow paths (happy, nil, empty, upstream error). Include interaction edge cases (double-click, navigate-away, stale state), observability commitments, and explicit deferred-item logging.",
             "**Premise Challenge** — Is this the right problem? What if we do nothing? What are we optimizing for?",
@@ -55,7 +56,8 @@ export const SCOPE = {
             "**Temporal Interrogation** — (complex projects only) simulate implementation timeline: HOUR 1 foundations, HOUR 2-3 core logic, HOUR 4-5 integration surprises, HOUR 6+ polish/tests. Decide what must be locked now vs safely deferred.",
             "**Mode Selection** — Present expand/selective/hold/reduce with recommendation and default heuristic: greenfield -> expand, feature enhancement -> selective, bugfix/hotfix/refactor -> hold, broad blast radius (>15 files or multi-team impact) -> reduce.",
             "**Mode-Specific Analysis** — After mode is selected, run the matching analysis: EXPAND (10x and delight opportunities), SELECTIVE (hold-scope rigor then cherry-picked expansions), HOLD (minimum-change-set hardening), REDUCE (ruthless cuts and follow-up split).",
-            "**Outside Voice + Spec Review Loop** — run an adversarial second-opinion pass on the scope artifact, reconcile findings, and iterate up to 3 cycles or until quality score >= 0.8.",
+            "**Plant-seed shelf (optional)** — when a deferred/out-of-scope idea still has upside, capture it as `.cclaw/seeds/SEED-<YYYY-MM-DD>-<slug>.md` with trigger_when and action instead of losing it in prose-only notes.",
+            "**Outside Voice + Spec Review Loop** — run an adversarial second-opinion pass on the scope artifact, reconcile findings, and iterate up to 3 cycles or until quality score >= 0.8. When `.cclaw/config.yaml::reviewLoop.externalSecondOpinion.enabled` is true, run an additional external-model pass and explicitly resolve score/finding disagreements.",
             "**Error and Rescue Registry** — For each capability: what breaks, how detected, what fallback."
         ],
         interactionProtocol: [
@@ -74,15 +76,16 @@ export const SCOPE = {
             "**STOP BEFORE ADVANCE.** Mandatory delegation `planner` must be marked completed or explicitly waived in `.cclaw/state/delegation-log.json`. Then close the stage via `node .cclaw/hooks/stage-complete.mjs scope` (do not hand-edit `.cclaw/state/flow-state.json`)."
         ],
         process: [
-            "Run pre-scope system audit (git log/diff/stash/debt markers).",
+            "When `.cclaw/config.yaml::optInAudits.scopePreAudit` is true, run pre-scope system audit (git log/diff/stash/debt markers).",
             "Run premise challenge and existing-solution leverage check.",
             "When mode is EXPAND/SELECTIVE, run brief landscape check before final scope lock.",
             "Calibrate quality bar against 2-3 strong existing modules/files.",
             "Produce 2-3 scope alternatives in a structured format (Name, Summary, Effort, Risk, Pros, Cons, Reuses) with minimum viable and ideal architecture options included.",
             "Choose scope mode with user approval.",
             "Run mode-specific analysis that matches the selected scope mode.",
+            "Optionally plant high-upside deferred ideas into `.cclaw/seeds/SEED-<YYYY-MM-DD>-<slug>.md` with trigger_when/action notes.",
             "Walk through scope review sections one at a time.",
-            "Run outside-voice spec review loop (up to 3 iterations, quality score target >= 0.8).",
+            "Run outside-voice spec review loop (up to 3 iterations, quality score target >= 0.8). If configured, include external second opinion and reconcile deltas.",
             "Write explicit scope contract, discretion areas, and deferred items.",
             "Freeze non-negotiable boundaries as stable Locked Decisions (D-XX IDs).",
             "Produce scope summary plus completion dashboard (section status, critical gaps, resolved decisions, unresolved items or `None`)."
@@ -94,7 +97,7 @@ export const SCOPE = {
         ],
         requiredEvidence: [
             "Artifact written to `.cclaw/artifacts/02-scope-<slug>.md`.",
-            "Pre-Scope System Audit findings are captured (git log/diff/stash/debt markers).",
+            "When `.cclaw/config.yaml::optInAudits.scopePreAudit` is true, Pre-Scope System Audit findings are captured (git log/diff/stash/debt markers).",
             "In-scope and out-of-scope lists are explicit.",
             "Discretion areas are explicit (or marked as `None`).",
             "Selected mode and rationale are documented.",
@@ -102,7 +105,9 @@ export const SCOPE = {
             "Premise challenge findings documented.",
             "Outside Voice findings and dispositions are recorded (accept/reject/defer with rationale).",
             "Spec review loop summary includes iteration count and quality score trajectory.",
+            "When `.cclaw/config.yaml::reviewLoop.externalSecondOpinion.enabled` is true, external second-opinion disposition is captured.",
             "Deferred items list with one-line rationale for each.",
+            "When an upside deferred idea is parked, a seed file is created under `.cclaw/seeds/` and referenced in the artifact.",
             "Completion dashboard lists per-section status, critical/open gaps, decision count, and unresolved items (or `None`)."
         ],
         inputs: ["brainstorm artifact", "timeline constraints", "product priorities"],
@@ -128,6 +133,11 @@ export const SCOPE = {
             "locked decisions captured with stable D-XX IDs",
             "completion dashboard produced",
             "scope summary produced"
+        ],
+        platformNotes: [
+            "Scope contract paths must be repo-relative with forward slashes so they resolve identically on Windows, macOS, and Linux (`src/pkg/mod.ts`, NOT `src\\pkg\\mod.ts`).",
+            "When invoking `git log`/`git diff` for the Pre-Scope audit, wrap glob patterns in single quotes on POSIX shells and double quotes on PowerShell (`git log -- 'src/**/*.ts'` vs `git log -- \"src/**/*.ts\"`). Document the command with the quoting style you actually ran.",
+            "Do not hard-code machine-specific absolute paths (home dirs, drive letters) into the scope contract — keep boundaries repo-relative."
         ]
     },
     artifactRules: {
@@ -139,7 +149,7 @@ export const SCOPE = {
             traceabilityRule: "Every scope boundary must be traceable to a brainstorm decision. Every downstream design choice must stay within the scope contract."
         },
         artifactValidation: [
-            { section: "Pre-Scope System Audit", required: false, validationRule: "Must capture git log/diff/stash/debt-marker findings before premise challenge." },
+            { section: "Pre-Scope System Audit", required: false, validationRule: "When `.cclaw/config.yaml::optInAudits.scopePreAudit` is true: must capture git log -30, git diff --stat, git stash list, and debt-marker scan (TODO/FIXME/XXX/HACK) before premise challenge." },
             { section: "Prime Directives", required: false, validationRule: "For each scoped capability: named failure modes, explicit error surface, four data-flow paths, interaction edge cases, observability expectations, and deferred-item handling." },
             { section: "Premise Challenge", required: false, validationRule: "Must contain explicit answers to: right problem? direct path? what if nothing?" },
             { section: "Landscape Check", required: false, validationRule: "When mode is EXPAND/SELECTIVE, include at least one external reference insight and its impact on scope." },
@@ -163,6 +173,12 @@ export const SCOPE = {
     },
     reviewLens: {
         outputs: ["scope mode decision", "scope contract", "discretion areas list", "deferred scope list", "scope summary", "scope completion dashboard"],
+        reviewLoop: {
+            stage: "scope",
+            checklist: REVIEW_LOOP_CHECKLISTS.scope.map((dimension) => dimension.id),
+            maxIterations: 3,
+            targetScore: 0.8
+        },
         reviewSections: [
             {
                 title: "Scope Boundary Audit",

package/dist/content/stages/ship.js

@@ -92,6 +92,12 @@ export const SHIP = {
             "preflight completed",
             "rollback and release notes complete",
             "finalization action explicitly chosen and executed"
+        ],
+        platformNotes: [
+            "Release commands (`npm publish`, `git tag -s`, `gh release create`, `cargo publish`, `goreleaser`) behave the same across OSes, but signing keys differ: macOS Keychain, Windows credential store, Linux GPG agent. Verify the signing flow on the actual release machine before running the real publish.",
+            "Version tags must be pure ASCII and lowercase after an optional `v` prefix (`v1.2.3`, `v1.2.3-rc.1`). Avoid Unicode dashes and non-breaking spaces that sneak in via copy-paste from docs.",
+            "When the rollback plan references timestamps (CI run windows, DB snapshot IDs), pin them to UTC ISO-8601 so the plan reads identically across CI runners in different regions.",
+            "`gh release create` requires a repo-level `GH_TOKEN`/`GITHUB_TOKEN`; document whether it is sourced from the shell env, `.env`, or the OS keychain so another operator on a different OS can reproduce the release."
         ]
     },
     artifactRules: {

package/dist/content/stages/spec.js

@@ -40,6 +40,7 @@ export const SPEC = {
             "Capture edge cases — for each criterion, define at least one boundary condition and one error condition.",
             "Document constraints and assumptions — regulatory, system, integration, and performance boundaries. Surface implicit assumptions explicitly.",
             "Confirm testability — for each acceptance criterion, describe the test that would prove it. If untestable, rewrite the criterion.",
+            "Present acceptance criteria to the user in 3-5-item batches, pausing for explicit ACK between batches (see Interaction Protocol).",
             "Write spec artifact and request user approval — wait for explicit confirmation before proceeding."
         ],
         interactionProtocol: [
@@ -86,6 +87,11 @@ export const SPEC = {
             "required gates marked satisfied",
             "plan-ready acceptance mapping exists",
             "testability confirmed for all criteria"
+        ],
+        platformNotes: [
+            "Acceptance criteria that reference CLI commands must name the executable portably (`node`, `npm`, `pytest`) and avoid OS-specific shell features (`&&` is safe, `||` differs subtly between cmd.exe and POSIX — prefer explicit multi-step descriptions).",
+            "When a criterion specifies file-content expectations, use `LF` as the canonical newline and state any CRLF-on-Windows tolerance explicitly (most git-managed repos normalize via `.gitattributes`; the criterion should not implicitly depend on autocrlf).",
+            "Timezone-sensitive criteria (timestamps, retention windows) must pin UTC or note the source of truth — clocks differ across CI runners (GitHub macOS vs Linux image vs Windows image)."
         ]
     },
     artifactRules: {

package/dist/content/stages/tdd.js

@@ -103,6 +103,12 @@ export const TDD = {
             "REFACTOR evidence captured",
             "required gates marked satisfied",
             "traceability annotated"
+        ],
+        platformNotes: [
+            "Record the **exact** test command run (`npm test -- path/to/file`, `pytest path/to/file`, `go test ./...`) so RED/GREEN evidence is reproducible on any OS. Do not paraphrase to a shorter alias.",
+            "Line-ending drift (CRLF vs LF) can turn a passing test red on Windows if the repo mixes styles. When a GREEN flip happens only after whitespace normalization, record it as a refactor note, not a hidden fix.",
+            "When invoking a test file path from Windows PowerShell, use forward slashes (`npm test -- tests/foo.test.ts`) — backslashes trip globbing on `cross-env` and similar wrappers.",
+            "Flaky tests that only fail on one OS must be marked as such in the TDD artifact (OS + runner + one failing output snippet) — do not retry until green without evidence."
         ]
     },
     artifactRules: {

package/dist/content/start-command.js

@@ -47,13 +47,18 @@ This is the **recommended way to start** working with cclaw. Use \`/cc-next\` fo
 
    Record the chosen class in \`.cclaw/artifacts/00-idea.md\` on the \`Class:\` line. Do NOT silently treat a non-software task as software.
 
-2. **Phase 1 — Origin-document discovery.** Before asking the user for context, scan for existing requirements/plan artifacts and merge them into initial context:
+2. **Phase 0.5 — Seed shelf recall.** Before routing, scan \`${RUNTIME_ROOT}/seeds/SEED-*.md\` and match each seed's \`trigger_when\` tokens against the prompt text (substring match is enough). If any match:
+   - Surface up to 3 matches (file + title + one-line action) as \`Seed recalls\`.
+   - Ask whether to apply now, keep as reference, or ignore for this run.
+   - If applied/reference, append selected seeds to \`00-idea.md\` under \`Discovered context\` so downstream stages keep the trace.
+
+3. **Phase 1 — Origin-document discovery.** Before asking the user for context, scan for existing requirements/plan artifacts and merge them into initial context:
    - \`.cclaw/artifacts/00-idea.md\` if it already exists (resumed flow).
    - Common origin locations: \`docs/prd/**\`, \`docs/rfcs/**\`, \`docs/adr/**\`, \`docs/design/**\`, \`specs/**\`, \`prd/**\`, \`rfc/**\`, \`design/**\`, root-level \`PRD.md\` / \`SPEC.md\` / \`DESIGN.md\` / \`REQUIREMENTS.md\` / \`ROADMAP.md\`.
    - Summarize each discovered doc in \`00-idea.md\` under a \`Discovered context\` section with path + 1-line summary.
    - If an origin doc contradicts the prompt, surface the conflict to the user before routing.
 
-3. **Phase 2 — Tech-stack + version detection.** Sniff the repo for stack + language versions and record under \`Stack:\`:
+4. **Phase 2 — Tech-stack + version detection.** Sniff the repo for stack + language versions and record under \`Stack:\`:
    - Node: \`package.json\` \`engines\` / \`volta\` / \`packageManager\` / \`devDependencies\`.
    - Python: \`pyproject.toml\` / \`requirements*.txt\` / \`.python-version\`.
    - Go: \`go.mod\` (module + Go version).
@@ -63,9 +68,9 @@ This is the **recommended way to start** working with cclaw. Use \`/cc-next\` fo
    - CI: \`.github/workflows\`, \`.gitlab-ci.yml\`.
    Skip detection quietly if no markers are found — do NOT invent a stack.
 
-4. Read \`${flowPath}\`.
-5. If flow already has completed stages, warn the user that starting a new tracked flow will reset progress. Ask for confirmation before proceeding.
-6. **Track heuristic** — classify the idea text and **recommend** a track (the user can override before any state mutation):
+5. Read \`${flowPath}\`.
+6. If flow already has completed stages, warn the user that starting a new tracked flow will reset progress. Ask for confirmation before proceeding.
+7. **Track heuristic** — classify the idea text and **recommend** a track (the user can override before any state mutation):
    - First, load \`${RUNTIME_ROOT}/config.yaml\`. If \`trackHeuristics\` is defined, apply those per-track vocabulary hints (\`fallback\`, \`tracks.<id>.{triggers,veto}\`) on top of the built-in defaults. Evaluation order is always \`standard -> medium -> quick\` (narrow-to-broad).
    - **quick** (\`spec → tdd → review → ship\`) — single-purpose work where the spec is essentially already known.
      Triggers (case-insensitive substring or close variant): \`bug\`, \`bugfix\`, \`fix\`, \`hotfix\`, \`patch\`, \`typo\`, \`regression\`, \`copy change\`, \`rename\`, \`bump\`, \`upgrade dep\`, \`config tweak\`, \`docs only\`, \`comment\`, \`lint\`, \`format\`, \`small\`, \`tiny\`, \`one-liner\`, \`revert\`.
@@ -74,17 +79,17 @@ This is the **recommended way to start** working with cclaw. Use \`/cc-next\` fo
    - **standard** (full 8 stages — default fallback) — anything that introduces a new capability with architecture uncertainty, touches many modules, or has unclear scope.
      Triggers: \`new feature\`, \`refactor\`, \`migration\`, \`platform\`, \`architecture\`, \`schema\`, \`integrate\`, \`workflow\`, \`onboarding\`, or any prompt that does not match quick/medium confidently.
    - When triggers conflict, prefer **standard** over **medium**, and **medium** over **quick**.
-7. Present the recommendation as a single decision with explicit options:
+8. Present the recommendation as a single decision with explicit options:
    > \`Recommended track: <quick|medium|standard>\` because \`<one-line reason citing matched triggers>\`.
    > Override? (A) keep \`<recommended>\`  (B) switch track  (C) cancel.
    If the harness's native ask tool is available (\`AskUserQuestion\` / \`AskQuestion\` / \`question\` / \`request_user_input\`), send exactly ONE question; on schema error, fall back to a plain-text lettered list.
-8. Persist the chosen track to \`${flowPath}\` (\`track\` field). Compute \`skippedStages\` from the track and write that too. Use the **first stage of the chosen track** as \`currentStage\` (quick → \`spec\`, medium/standard → \`brainstorm\`, trivial fast-path → \`design\` or \`spec\` per Phase 0).
-9. Write the prompt to \`.cclaw/artifacts/00-idea.md\` with the following header lines: \`Class:\` (from Phase 0), \`Track:\` (chosen track + matched heuristic), \`Stack:\` (from Phase 2 detection, or \`unknown\`), and a \`Discovered context\` section if Phase 1 found origin docs.
-10. Load the **first-stage skill for the chosen track** and its command file:
+9. Persist the chosen track to \`${flowPath}\` (\`track\` field). Compute \`skippedStages\` from the track and write that too. Use the **first stage of the chosen track** as \`currentStage\` (quick → \`spec\`, medium/standard → \`brainstorm\`, trivial fast-path → \`design\` or \`spec\` per Phase 0).
+10. Write the prompt to \`.cclaw/artifacts/00-idea.md\` with the following header lines: \`Class:\` (from Phase 0), \`Track:\` (chosen track + matched heuristic), \`Stack:\` (from Phase 2 detection, or \`unknown\`), and a \`Discovered context\` section if Phase 1/seed recall found references.
+11. Load the **first-stage skill for the chosen track** and its command file:
     - quick → \`.cclaw/skills/specification-authoring/SKILL.md\` + \`.cclaw/commands/spec.md\`
     - medium/standard → \`.cclaw/skills/brainstorming/SKILL.md\` + \`.cclaw/commands/brainstorm.md\`
     - trivial fast-path → design or spec skill per Phase 0 decision.
-11. Execute that stage with the prompt + Phase 1/Phase 2 context as initial input.
+12. Execute that stage with the prompt + Phase 1/Phase 2 + seed context as initial input.
 
 ### Reclassification on discovery
 
@@ -152,14 +157,15 @@ Do **not** silently discard an existing flow when the user provides a prompt. If
 ### Path A: \`/cc <prompt>\`
 
 1. **Task classification (Phase 0).** Decide whether the prompt is \`software-standard\`, \`software-trivial\`, \`software-bugfix\`, \`pure-question\`, or \`non-software\`. Non-software and pure-question exit immediately — answer directly, do not open a stage.
-2. **Origin-document discovery (Phase 1).** Scan for \`docs/prd/**\`, \`docs/rfcs/**\`, \`docs/adr/**\`, \`docs/design/**\`, \`specs/**\`, root-level \`PRD.md\` / \`SPEC.md\` / \`DESIGN.md\` / \`REQUIREMENTS.md\`. Summarize any hits in \`00-idea.md\` under \`Discovered context\`. Surface conflicts with the prompt before routing.
-3. **Stack detection (Phase 2).** Inspect \`package.json\` engines, \`pyproject.toml\`, \`go.mod\`, \`Cargo.toml\`, \`pom.xml\`, \`build.gradle*\`, \`Dockerfile\`, \`docker-compose*.yml\`, and CI configs. Record stack + versions on the \`Stack:\` line. Do not invent stack details.
-4. Read \`${flowPath}\`.
-5. If \`completedStages\` is non-empty:
+2. **Seed shelf recall (Phase 0.5).** Scan \`${RUNTIME_ROOT}/seeds/SEED-*.md\` and match \`trigger_when\` tokens against the prompt text. Surface up to 3 matching seeds with file/title/action and ask whether to apply or ignore. When applied, add them to \`00-idea.md\` under \`Discovered context\`.
+3. **Origin-document discovery (Phase 1).** Scan for \`docs/prd/**\`, \`docs/rfcs/**\`, \`docs/adr/**\`, \`docs/design/**\`, \`specs/**\`, root-level \`PRD.md\` / \`SPEC.md\` / \`DESIGN.md\` / \`REQUIREMENTS.md\`. Summarize any hits in \`00-idea.md\` under \`Discovered context\`. Surface conflicts with the prompt before routing.
+4. **Stack detection (Phase 2).** Inspect \`package.json\` engines, \`pyproject.toml\`, \`go.mod\`, \`Cargo.toml\`, \`pom.xml\`, \`build.gradle*\`, \`Dockerfile\`, \`docker-compose*.yml\`, and CI configs. Record stack + versions on the \`Stack:\` line. Do not invent stack details.
+5. Read \`${flowPath}\`.
+6. If \`completedStages\` is non-empty:
    - Inform: "You have an active flow at stage **{currentStage}** with {N} completed stages. Starting a new tracked flow will reset progress."
    - Ask: "Continue with reset? (A) Yes, start fresh (B) No, resume current flow"
    - If (B) → switch to Path B behavior.
-6. **Classify the idea** using the heuristic below and present a single track recommendation. Wait for explicit confirmation or override before mutating any state.
+7. **Classify the idea** using the heuristic below and present a single track recommendation. Wait for explicit confirmation or override before mutating any state.
    - If \`${RUNTIME_ROOT}/config.yaml\` defines \`trackHeuristics\`, apply those vocabulary hints (\`fallback\`, \`tracks.<id>.{triggers,veto}\`) on top of built-in defaults. Evaluation order is fixed: \`standard -> medium -> quick\`. (Honest note: this is advisory prose; the LLM applies it, not a Node-level router.)
 
    **Track heuristic** (lowercase substring match against the user prompt):
@@ -172,9 +178,9 @@ Do **not** silently discard an existing flow when the user provides a prompt. If
 
    - On conflict, prefer \`standard\` over \`medium\`, and \`medium\` over \`quick\`.
    - Always state the recommendation as a one-line reason citing matched triggers.
-7. Persist the chosen track in \`${flowPath}\` (\`track\` + \`skippedStages\`). Set \`currentStage\` to the first stage of the chosen track (\`quick\` → \`spec\`, \`medium\`/ \`standard\` → \`brainstorm\`, trivial fast-path → \`design\` or \`spec\`). Reset gate catalog.
-8. Write \`${RUNTIME_ROOT}/artifacts/00-idea.md\` with the user's prompt plus header lines: \`Class:\`, \`Track:\`, \`Stack:\`, and a \`Discovered context\` section from Phase 1.
-9. Load and execute the **first stage skill of the chosen track** (\`brainstorming\` for medium/standard, \`specification-authoring\` for quick) plus its matching command file.
+8. Persist the chosen track in \`${flowPath}\` (\`track\` + \`skippedStages\`). Set \`currentStage\` to the first stage of the chosen track (\`quick\` → \`spec\`, \`medium\`/ \`standard\` → \`brainstorm\`, trivial fast-path → \`design\` or \`spec\`). Reset gate catalog.
+9. Write \`${RUNTIME_ROOT}/artifacts/00-idea.md\` with the user's prompt plus header lines: \`Class:\`, \`Track:\`, \`Stack:\`, and a \`Discovered context\` section from Phase 0.5/1.
+10. Load and execute the **first stage skill of the chosen track** (\`brainstorming\` for medium/standard, \`specification-authoring\` for quick) plus its matching command file.
 
 ### Reclassification on discovery
 

package/dist/content/templates.js

@@ -28,17 +28,36 @@ inputs_hash: sha256:pending
 |---|---|---|---|
 | 1 |  |  |  |
 
+## Approach Tier
+- Tier: Lightweight | Standard | Deep
+- Why this tier:
+
+## Short-Circuit Decision
+- Status: bypassed
+- Why:
+- Scope handoff:
+
 ## Approaches
-| Approach | Architecture | Trade-offs | Recommendation |
-|---|---|---|---|
-| A |  |  |  |
-| B |  |  |  |
+| Approach | Role | Architecture | Trade-offs | Recommendation |
+|---|---|---|---|---|
+| A | baseline |  |  |  |
+| B | challenger: higher-upside |  |  |  |
+
+## Approach Reaction
+- Closest option:
+- Concerns:
+- What changed after reaction:
 
 ## Selected Direction
 - **Approach:**
 - **Rationale:**
 - **Approval:** pending
 
+## Seed Shelf Candidates (optional)
+| Seed file | Trigger when | Suggested action | Status (planted/deferred/ignored) |
+|---|---|---|---|
+| .cclaw/seeds/SEED-YYYY-MM-DD-<slug>.md |  |  |  |
+
 ## Design
 - **Architecture:**
 - **Key components:**
@@ -62,6 +81,14 @@ inputs_hash: sha256:pending
 
 # Scope Artifact
 
+## Pre-Scope System Audit
+| Check | Command | Findings |
+|---|---|---|
+| Recent commits | \`git log -30 --oneline\` |  |
+| Current diff | \`git diff --stat\` |  |
+| Stash state | \`git stash list\` |  |
+| Debt markers | \`rg -n "TODO|FIXME|XXX|HACK"\` |  |
+
 ## Prime Directives
 - Zero silent failures:
 - Every error has a name:
@@ -138,11 +165,28 @@ inputs_hash: sha256:pending
 |---|---|
 |  |  |
 
+## Seed Shelf Candidates (optional)
+| Seed file | Trigger when | Suggested action | Status (planted/deferred/ignored) |
+|---|---|---|---|
+| .cclaw/seeds/SEED-YYYY-MM-DD-<slug>.md |  |  |  |
+
 ## Error & Rescue Registry
 | Capability | Failure mode | Detection | Fallback |
 |---|---|---|---|
 |  |  |  |  |
 
+## Outside Voice Findings
+| ID | Dimension | Finding | Disposition | Rationale |
+|---|---|---|---|---|
+| F-1 | premise_fit |  | accept/reject/defer |  |
+
+## Spec Review Loop
+| Iteration | Quality Score | Findings | Stop decision |
+|---|---|---|---|
+| 1 | 0.00 | 0 | continue/stop |
+- Stop reason:
+- Unresolved concerns:
+
 ## Completion Dashboard
 - Checklist findings:
 - Resolved decisions count:
@@ -236,21 +280,55 @@ inputs_hash: sha256:pending
 
 ## Architecture Diagram
 
+<!-- diagram: architecture -->
+
 \`\`\`
 (ASCII, Mermaid, or tool-generated diagram showing component boundaries and data flow direction)
 \`\`\`
 
 ## Data-Flow Shadow Paths
+<!-- diagram: data-flow-shadow-paths -->
 | Path | Trigger | Fallback/Degrade behavior |
 |---|---|---|
 |  |  |  |
 
 ## Error Flow Diagram
 
+<!-- diagram: error-flow -->
+
 \`\`\`
 (failure detection -> rescue action -> user-visible outcome)
 \`\`\`
 
+## State Machine Diagram
+
+<!-- diagram: state-machine -->
+
+\`\`\`
+(state transitions for the critical flow lifecycle)
+\`\`\`
+
+## Rollback Flowchart
+
+<!-- diagram: rollback-flowchart -->
+
+\`\`\`
+(trigger -> rollback actions -> verification)
+\`\`\`
+
+## Deployment Sequence Diagram
+
+<!-- diagram: deployment-sequence -->
+
+\`\`\`
+(rollout order, guard checks, and verification sequence)
+\`\`\`
+
+## Stale Diagram Audit
+| File | Last modified | Diagram marker baseline | Status | Notes |
+|---|---|---|---|---|
+|  |  |  | clear/stale |  |
+
 ## What Already Exists
 | Sub-problem | Existing code/library | Layer | Reuse decision |
 |---|---|---|---|
@@ -262,6 +340,15 @@ inputs_hash: sha256:pending
 - Upstream error path:
 - Timeout/downstream path:
 
+### Interaction Edge Case Matrix
+| Edge case | Handled? | Design response | Deferred item (if not handled) |
+|---|---|---|---|
+| double-click | yes/no |  | None / D-XX |
+| nav-away-mid-request | yes/no |  | None / D-XX |
+| 10K-result dataset | yes/no |  | None / D-XX |
+| background-job abandonment | yes/no |  | None / D-XX |
+| zombie connection | yes/no |  | None / D-XX |
+
 ## Security & Threat Model
 | Boundary | Threat | Mitigation | Owner |
 |---|---|---|---|
@@ -292,6 +379,18 @@ inputs_hash: sha256:pending
 |---|---|---|
 |  |  |  |
 
+## Outside Voice Findings
+| ID | Dimension | Finding | Disposition | Rationale |
+|---|---|---|---|---|
+| F-1 | architecture_fit |  | accept/reject/defer |  |
+
+## Spec Review Loop
+| Iteration | Quality Score | Findings | Stop decision |
+|---|---|---|---|
+| 1 | 0.00 | 0 | continue/stop |
+- Stop reason:
+- Unresolved concerns:
+
 ## NOT in scope
 - 
 
@@ -314,6 +413,11 @@ inputs_hash: sha256:pending
 |---|---|---|---|
 |  |  |  |  |
 
+## Seed Shelf Candidates (optional)
+| Seed file | Trigger when | Suggested action | Status (planted/deferred/ignored) |
+|---|---|---|---|
+| .cclaw/seeds/SEED-YYYY-MM-DD-<slug>.md |  |  |  |
+
 ## Completion Dashboard
 | Review Section | Status | Issues |
 |---|---|---|