cclaw-cli 0.51.21 → 0.51.23

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-next`.\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 or medium 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-next` = 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-next`.\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-next` = 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 function buildRulesJson(): Record<string, unknown>;

package/dist/content/templates.js

@@ -143,6 +143,7 @@ ${SEED_SHELF_SECTION}
 | Reversibility cost? |  |  |
 
 ## Dream State Mapping
+- Deep/optional only; omit for compact scope.
 - CURRENT STATE:
 - THIS PLAN:
 - 12-MONTH IDEAL:
@@ -156,6 +157,7 @@ ${SEED_SHELF_SECTION}
 | C (optional) |  |  |  |  |  |  |
 
 ## Temporal Interrogation
+- Deep/optional only; omit for compact scope.
 | Time slice | Likely decision pressure | Lock now or defer? | Reason |
 |---|---|---|---|
 | HOUR 1 (foundations) |  |  |  |
@@ -170,12 +172,11 @@ ${SEED_SHELF_SECTION}
 - [ ] SCOPE REDUCTION — strip to the smallest useful wedge when risk/blast radius is too high.
 
 ## Mode-Specific Analysis
-- **Selected mode:**
-- **Analysis:**
-  - (SCOPE EXPANSION: 10-star opportunities, delight features)
-  - (SELECTIVE EXPANSION: hold-scope baseline, cherry-picked expansions)
-  - (HOLD SCOPE: approved slice with maximum rigor)
-  - (SCOPE REDUCTION: ruthless cuts, follow-up split)
+| Selected mode | Rationale | Depth |
+|---|---|---|
+|  |  | default / deep |
+
+> Default path: one selected-mode row plus rationale. Deep/high-risk scope may expand below with mode-specific analysis.
 
 ## Requirements (stable IDs)
 | ID | Requirement (observable outcome) | Priority | Source (origin doc / prompt line) |
@@ -224,7 +225,7 @@ ${SEED_SHELF_SECTION}
 |---|---|---|---|---|
 | F-1 | premise_fit |  | accept/reject/defer |  |
 
-## Spec Review Loop
+## Scope Outside Voice Loop
 | Iteration | Quality Score | Findings | Stop decision |
 |---|---|---|---|
 | 1 | 0.00 | 0 | continue/stop |
@@ -288,8 +289,13 @@ ${SEED_SHELF_SECTION}
 
 # Design Artifact
 
+## Compact-First Scaffold
+- Default to the compact design spine unless risk requires Standard/Deep add-ons.
+- Compact required spine: Codebase Investigation, Architecture Boundaries, Architecture Diagram, Data Flow, Failure Mode Table, Test Strategy, and Completion Dashboard.
+- Mark optional Standard/Deep sections as \`Omitted - compact design\` when they do not apply; do not expand the scaffold just to fill empty tables.
+
 ## Upstream Handoff
-- Source artifacts: \`02-scope-<slug>.md\`, \`02a-research.md\` when present
+- Source artifacts: \`02-scope-<slug>.md\`, \`02a-research.md\` only when present for deep/high-risk research
 - Decisions carried forward:
 - Constraints carried forward:
 - Open questions:
@@ -308,12 +314,11 @@ ${SEED_SHELF_SECTION}
 | Layer 3 |  |  |
 
 ## Research Fleet Synthesis
-| Lens | Key findings | Design impact | Evidence |
+| Lens actually run | Key findings | Design impact | Evidence |
 |---|---|---|---|
-| stack-researcher |  |  |  |
-| features-researcher |  |  |  |
-| architecture-researcher |  |  |  |
-| pitfalls-researcher |  |  |  |
+| compact inline synthesis |  |  |  |
+
+> Default path: compact inline synthesis here. Deep/high-risk work may also write \`.cclaw/artifacts/02a-research.md\`.
 
 ## Architecture Boundaries
 | Component | Responsibility | Requirement Refs (R#) | Decision Refs (LD#hash) | Owner |
@@ -329,12 +334,14 @@ ${MARKDOWN_CODE_FENCE}
 ${MARKDOWN_CODE_FENCE}
 
 ## Data-Flow Shadow Paths
+- Standard/Deep add-on; omit when compact design does not need a shadow path.
 <!-- diagram: data-flow-shadow-paths -->
 | Path | Trigger | Fallback/Degrade behavior |
 |---|---|---|
 |  |  |  |
 
 ## Error Flow Diagram
+- Standard/Deep add-on; omit when the Failure Mode Table is sufficient.
 
 <!-- diagram: error-flow -->
 
@@ -343,6 +350,7 @@ ${MARKDOWN_CODE_FENCE}
 ${MARKDOWN_CODE_FENCE}
 
 ## State Machine Diagram
+- Deep add-on; omit for compact design.
 
 <!-- diagram: state-machine -->
 
@@ -351,6 +359,7 @@ ${MARKDOWN_CODE_FENCE}
 ${MARKDOWN_CODE_FENCE}
 
 ## Rollback Flowchart
+- Deep add-on; omit for compact design.
 
 <!-- diagram: rollback-flowchart -->
 
@@ -359,6 +368,7 @@ ${MARKDOWN_CODE_FENCE}
 ${MARKDOWN_CODE_FENCE}
 
 ## Deployment Sequence Diagram
+- Deep add-on; omit for compact design.
 
 <!-- diagram: deployment-sequence -->
 
@@ -426,7 +436,7 @@ ${MARKDOWN_CODE_FENCE}
 |---|---|---|---|---|
 | F-1 | architecture_fit |  | accept/reject/defer |  |
 
-## Spec Review Loop
+## Design Outside Voice Loop
 | Iteration | Quality Score | Findings | Stop decision |
 |---|---|---|---|
 | 1 | 0.00 | 0 | continue/stop |
@@ -439,6 +449,7 @@ ${MARKDOWN_CODE_FENCE}
 - 
 
 ## Parallelization Strategy
+- Standard/Deep add-on when multi-module; omit for compact sequential work.
 - Parallel lanes:
 - Conflict risks:
 
@@ -448,11 +459,13 @@ ${MARKDOWN_CODE_FENCE}
 |  |  |  |
 
 ## Interface Contracts
+- Standard/Deep add-on when module boundaries or APIs change; omit for compact local changes.
 | Module | Produces | Consumes |
 |---|---|---|
 |  |  |  |
 
 ## Unresolved Decisions
+- Standard/Deep add-on; use \`None\` for compact design with no unresolved decisions.
 | Decision | Missing info | Owner | Default |
 |---|---|---|---|
 |  |  |  |  |
@@ -481,7 +494,7 @@ ${SEED_SHELF_SECTION}
 # Specification Artifact
 
 ## Upstream Handoff
-- Source artifacts: \`02-scope-<slug>.md\`, \`03-design-<slug>.md\`
+- Source artifacts: standard uses \`02-scope-<slug>.md\` + \`03-design-<slug>.md\`; medium uses \`01-brainstorm-<slug>.md\` when present; quick uses \`00-idea.md\` plus reproduction context.
 - Decisions carried forward:
 - Constraints carried forward:
 - Open questions:
@@ -492,9 +505,14 @@ ${SEED_SHELF_SECTION}
 |---|---|---|---|
 | AC-1 | R1 |  |  |
 
-> Every AC must reference at least one \`R#\` from \`02-scope.md\`. ACs are
-> stable (never renumber): dropped ACs stay with Priority \`DROPPED\`; new
-> ones append with the next free \`AC-#\`.
+> Standard ACs reference at least one \`R#\` from \`02-scope.md\`. Quick-track ACs may instead put \`Quick Reproduction Contract\` / bug-slice refs in the Requirement Ref column and \`N/A\` for Design Decision Ref. ACs are stable (never renumber): dropped ACs stay with Priority \`DROPPED\`; new ones append with the next free \`AC-#\`.
+
+## Quick Reproduction Contract
+> Required for quick bug-fix specs; use \`N/A\` for non-bugfix or standard/medium tracks. TDD turns this contract into the RED reproduction test.
+
+| Bug slice | Symptom | Repro steps | Expected RED test behavior | Linked acceptance criterion |
+|---|---|---|---|---|
+| QS-1 |  |  |  | AC-1 |
 
 ## Edge Cases
 | Criterion ID | Boundary case | Error case |
@@ -627,7 +645,7 @@ Execution rule: complete and verify each batch before starting the next batch.
 # TDD Artifact
 
 ## Upstream Handoff
-- Source artifacts: \`04-spec.md\`, \`05-plan.md\`
+- Source artifacts: \`04-spec.md\` plus the active track's upstream source item (plan slice on standard/medium, spec acceptance item or bug reproduction slice on quick).
 - Decisions carried forward:
 - Constraints carried forward:
 - Open questions:
@@ -654,9 +672,11 @@ Execution rule: complete and verify each batch before starting the next batch.
 | S-1 |  |  |  |
 
 ## Acceptance Mapping
-| Slice | Plan task ID | Spec criterion ID |
+| Slice | Source item ID | Spec criterion ID |
 |---|---|---|
-| S-1 | T-1 | AC-1 |
+| S-1 | SRC-1 | AC-1 |
+
+> Map each slice to the active track's source item: plan slice on standard/medium, or the \`Quick Reproduction Contract\` bug slice / spec acceptance item on quick.
 
 ## Failure Analysis
 | Slice | Expected missing behavior | Actual failure reason |
@@ -673,7 +693,7 @@ Execution rule: complete and verify each batch before starting the next batch.
 - Behavior preserved:
 
 ## Traceability
-- Plan task IDs:
+- Source item IDs:
 - Spec criterion IDs:
 
 
@@ -709,7 +729,7 @@ Execution rule: complete and verify each batch before starting the next batch.
 # Review Artifact
 
 ## Upstream Handoff
-- Source artifacts: \`04-spec.md\`, \`05-plan.md\`, \`06-tdd.md\`
+- Source artifacts: \`04-spec.md\`, \`06-tdd.md\`, plus the active track's upstream source item when available.
 - Decisions carried forward:
 - Constraints carried forward:
 - Open questions:
@@ -723,7 +743,7 @@ Execution rule: complete and verify each batch before starting the next batch.
 ## Layer 2 Findings
 | ID | Severity | Category | Description | Status |
 |---|---|---|---|---|
-| R-1 | Critical/Important/Suggestion | correctness/security/performance/architecture |  | open/resolved |
+| R-1 | Critical/Important/Suggestion | correctness/security/performance/architecture/external-safety |  | open/resolved |
 - NO_CHANGE_ATTESTATION: <required when Category=security has no entries; explain why no security-relevant changes were detected>
 
 ## Incoming Feedback Queue
@@ -746,15 +766,15 @@ Execution rule: complete and verify each batch before starting the next batch.
 
 ## Completeness Snapshot
 - AC coverage: <N>/<M> (<percent>%)
-- Task coverage (tasks backed by ≥1 test slice): <N>/<M>
-- Slice coverage (slices linked to ≥1 AC): <N>/<M>
+- Source item coverage (source items backed by ≥1 test slice): <N>/<M> or \`N/A - direct spec/reproduction coverage\`
+- Slice coverage (slices linked to ≥1 AC or bug reproduction slice): <N>/<M>
 - Adversarial review: not triggered | pass | fail
 - Overall: complete | concerns | blocked
 
 ## Trace Matrix Check
-- Command: \`cclaw internal trace-matrix\`
+- Command: \`cclaw internal trace-matrix\` when the active track enforces it; otherwise record direct AC/reproduction-slice coverage.
 - Orphaned criteria: 0
-- Orphaned tasks: 0
+- Orphaned source items: 0 or \`N/A - direct spec/reproduction coverage\`
 - Orphaned tests: 0
 - Evidence ref:
 
@@ -955,7 +975,7 @@ Track-specific skips are allowed only when \`flow-state.track\` + \`skippedStage
 | Class | Route |
 |---|---|
 | non-trivial software work | \`/cc <idea>\` |
-| trivial software fix | \`/cc <idea>\` (quick or medium track) |
+| trivial software fix | \`/cc <idea>\` (quick track) |
 | bugfix with repro | \`/cc <idea>\` and enforce RED-first in tdd |
 | pure question / non-software | direct answer (no stage flow) |
 

package/dist/content/track-render-context.js

@@ -34,8 +34,15 @@ export function renderTrackTerminology(value, context) {
     }
     return value
         .replace(/\btask from the plan\b/giu, `${context.traceabilitySourceNoun} from the spec`)
+        .replace(/\bplan confirmation\b/giu, "spec approval")
+        .replace(/\bplan approval\b/giu, "spec approval")
+        .replace(/\bapproved plan slice\b/giu, `approved ${context.traceabilitySliceNoun}`)
+        .replace(/\bplanned slice\b/giu, context.traceabilitySliceNoun)
         .replace(/\bplan task ID\b/giu, context.traceabilityIdNoun)
         .replace(/\bplan task\b/giu, context.traceabilitySourceNoun)
+        .replace(/\bplan-task\b/giu, "acceptance-criterion")
+        .replace(/\btask coverage\b/giu, "source item coverage")
+        .replace(/\borphaned tasks\b/giu, "orphaned source items")
         .replace(/\bplan row\b/giu, "acceptance row")
         .replace(/\btraceable to plan slice\b/giu, `traceable to ${context.traceabilitySliceNoun}`)
         .replace(/\bplan slice\b/giu, context.traceabilitySliceNoun)

package/dist/content/view-command.js

@@ -37,9 +37,11 @@ ${conversationLanguagePolicyMarkdown()}
 For machine orchestration, emit one JSON envelope:
 
 \`\`\`json
-{"version":"1","kind":"stage-output","stage":"non-flow","payload":{"command":"/cc-view","subcommand":"status","summary":"<short>"},"emittedAt":"<ISO-8601>"}
+{"version":"1","kind":"stage-output","stage":"non-flow","payload":{"command":"/cc-view <status|tree|diff>","subcommand":"<status|tree|diff>","summary":"<short>"},"emittedAt":"<ISO-8601>"}
 \`\`\`
 
+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\`
 

package/dist/delegation.d.ts

@@ -6,8 +6,8 @@ export type DelegationStatus = "scheduled" | "completed" | "failed" | "waived";
  * How a delegation was actually fulfilled. Advisory — mirrors the harness
  * `subagentFallback` that was in effect when the entry was recorded.
  *
- * - `isolated`         — Claude-style isolated subagent worker.
- * - `generic-dispatch` — Cursor-style Task dispatch mapped to a named role.
+ * - `isolated`         — native isolated subagent worker (Claude/OpenCode/Codex).
+ * - `generic-dispatch` — generic Task/Subagent dispatch mapped to a named role.
  * - `role-switch`      — performed in-session with explicit role announce.
  * - `harness-waiver`   — auto-waived due to missing dispatch capability.
  */

package/dist/delegation.js

@@ -18,6 +18,13 @@ function delegationLockPath(projectRoot) {
 function createSpanId() {
     return `dspan-${Date.now().toString(36)}-${Math.random().toString(36).slice(2, 10)}`;
 }
+function activeHarnessSubagentFallback() {
+    const activeHarness = process.env.CCLAW_ACTIVE_HARNESS;
+    if (!activeHarness)
+        return undefined;
+    return HARNESS_ADAPTERS[activeHarness]
+        ?.capabilities.subagentFallback;
+}
 async function resolveReviewDiffBase(projectRoot) {
     let head = "";
     try {
@@ -103,6 +110,9 @@ async function detectReviewTriggers(projectRoot) {
         return empty;
     }
 }
+function hasValidWaiverReason(value) {
+    return typeof value === "string" && value.trim().length > 0;
+}
 function isDelegationTokenUsage(value) {
     if (!value || typeof value !== "object" || Array.isArray(value))
         return false;
@@ -130,6 +140,7 @@ function isDelegationEntry(value) {
             Number.isFinite(o.retryCount) &&
             Number.isInteger(o.retryCount) &&
             o.retryCount >= 0);
+    const waiverOk = o.status !== "waived" || hasValidWaiverReason(o.waiverReason);
     return (typeof o.stage === "string" &&
         typeof o.agent === "string" &&
         modeOk &&
@@ -141,6 +152,7 @@ function isDelegationEntry(value) {
         (o.endTs === undefined || typeof o.endTs === "string") &&
         (o.taskId === undefined || typeof o.taskId === "string") &&
         (o.waiverReason === undefined || typeof o.waiverReason === "string") &&
+        waiverOk &&
         (o.runId === undefined || typeof o.runId === "string") &&
         (o.fulfillmentMode === undefined ||
             o.fulfillmentMode === "isolated" ||
@@ -165,8 +177,10 @@ function parseLedger(raw, runId) {
         for (const item of entriesRaw) {
             if (isDelegationEntry(item)) {
                 const ts = item.startTs ?? item.ts ?? new Date().toISOString();
-                const inferredFulfillmentMode = item.fulfillmentMode
-                    ?? (item.status === "completed" ? "isolated" : undefined);
+                const isLegacyCompletion = item.fulfillmentMode === undefined &&
+                    item.schemaVersion === undefined &&
+                    item.status === "completed";
+                const inferredFulfillmentMode = item.fulfillmentMode ?? (isLegacyCompletion ? "isolated" : undefined);
                 entries.push({
                     ...item,
                     spanId: item.spanId ?? createSpanId(),
@@ -205,6 +219,9 @@ export async function appendDelegation(projectRoot, entry) {
         const filePath = delegationLogPath(projectRoot);
         const prior = await readDelegationLedger(projectRoot);
         const startTs = entry.startTs ?? entry.ts ?? new Date().toISOString();
+        if (entry.status === "waived" && !hasValidWaiverReason(entry.waiverReason)) {
+            throw new Error("waived delegation entries require a non-empty waiverReason");
+        }
         const stamped = { ...entry, runId: entry.runId ?? activeRunId };
         stamped.spanId = entry.spanId ?? createSpanId();
         stamped.startTs = startTs;
@@ -219,10 +236,16 @@ export async function appendDelegation(projectRoot, entry) {
             stamped.evidenceRefs = [];
         }
         if (stamped.status === "completed" && stamped.fulfillmentMode === undefined) {
-            const config = await readConfig(projectRoot).catch(() => null);
-            const harnesses = config?.harnesses ?? [];
-            const fallbacks = harnesses.map((h) => HARNESS_ADAPTERS[h].capabilities.subagentFallback);
-            stamped.fulfillmentMode = expectedFulfillmentMode(fallbacks);
+            const activeFallback = activeHarnessSubagentFallback();
+            if (activeFallback) {
+                stamped.fulfillmentMode = expectedFulfillmentMode([activeFallback]);
+            }
+            else {
+                const config = await readConfig(projectRoot).catch(() => null);
+                const harnesses = config?.harnesses ?? [];
+                const fallbacks = harnesses.map((h) => HARNESS_ADAPTERS[h].capabilities.subagentFallback);
+                stamped.fulfillmentMode = expectedFulfillmentMode(fallbacks);
+            }
         }
         // Idempotency: if a caller (or a retried hook) tries to append a row
         // with a spanId that already exists in the ledger, treat it as a no-op
@@ -256,10 +279,11 @@ export function expectedFulfillmentMode(fallbacks) {
     return "harness-waiver";
 }
 export async function checkMandatoryDelegations(projectRoot, stage, options = {}) {
-    const mandatory = stageSchema(stage).mandatoryDelegations;
-    const { activeRunId } = await readFlowState(projectRoot, {
+    const flowState = await readFlowState(projectRoot, {
         repairFeatureSystem: options.repairFeatureSystem
     });
+    const mandatory = stageSchema(stage, flowState.track).mandatoryDelegations;
+    const { activeRunId } = flowState;
     const ledger = await readDelegationLedger(projectRoot);
     const forStage = ledger.entries.filter((e) => e.stage === stage);
     const forRun = forStage.filter((e) => e.runId === activeRunId);
@@ -271,8 +295,9 @@ export async function checkMandatoryDelegations(projectRoot, stage, options = {}
     const missingEvidence = [];
     const config = await readConfig(projectRoot).catch(() => null);
     const harnesses = config?.harnesses ?? [];
-    const fallbacks = harnesses.map((h) => HARNESS_ADAPTERS[h].capabilities.subagentFallback);
-    const expectedMode = expectedFulfillmentMode(fallbacks);
+    const configuredFallbacks = harnesses.map((h) => HARNESS_ADAPTERS[h].capabilities.subagentFallback);
+    const activeFallback = activeHarnessSubagentFallback();
+    const expectedMode = expectedFulfillmentMode(activeFallback ? [activeFallback] : configuredFallbacks);
     for (const agent of mandatory) {
         const rows = forRun.filter((e) => e.agent === agent);
         const completedRows = rows.filter((e) => e.status === "completed");
@@ -287,9 +312,11 @@ export async function checkMandatoryDelegations(projectRoot, stage, options = {}
         if (hasWaived) {
             waived.push(agent);
         }
-        // Evidence is required for any non-isolated completion mode. Legacy rows
-        // without fulfillmentMode are inferred to `isolated` during parse.
-        const evidenceRequired = completedRows.some((e) => (e.fulfillmentMode ?? "isolated") !== "isolated");
+        // Evidence is required for non-isolated completions and for explicit
+        // degraded role-switch rows. Native OpenCode/Codex/Claude isolated
+        // dispatch is accepted as true subagent work; role-switch remains a
+        // fallback that must point at artifact evidence.
+        const evidenceRequired = expectedMode !== "isolated" || completedRows.some((e) => (e.fulfillmentMode ?? "isolated") !== "isolated");
         if (hasCompleted &&
             evidenceRequired &&
             !completedRows.some((e) => Array.isArray(e.evidenceRefs) && e.evidenceRefs.length > 0)) {

package/dist/doctor-registry.js

@@ -39,7 +39,7 @@ const RULES = [
         }
     },
     {
-        test: /^(dir:|command:|utility_command:|skill:|utility_skill:|agent:|harness_tool_ref:|harness_ref:|stage_examples_ref:|doctor_ref:)/,
+        test: /^(dir:|command:|utility_command:|stage_command:|skill:|utility_skill:|agent:|harness_tool_ref:|harness_ref:|stage_examples_ref:|doctor_ref:)/,
         metadata: {
             severity: "error",
             summary: "Generated runtime surface presence check.",