create-quiver 0.10.0 → 0.12.1

package/specs/quiver-v25-ai-first-lifecycle-orchestrator/slices/slice-04-agent-profiles-adapters/CLOSURE_BRIEF.md

@@ -0,0 +1,34 @@
+# CLOSURE BRIEF - slice-04: Agent profiles and provider adapters
+
+## Summary of Work
+
+Implemented the provider-agnostic agent profile contract for planner, executor, reviewer, and doctor roles, plus prompt-only rendering and redacted provider output.
+
+## Validation Against Acceptance Criteria
+
+- [x] Profiles inspected.
+- [x] Prompt-only mode verified.
+- [x] Missing CLI guidance verified.
+- [x] Structured output captured.
+- [x] Tests run.
+
+## Relevant Changes
+
+- Added `doctor` as a supported agent profile and removed `researcher` from the public profile contract.
+- Added `--print-prompt` support for planner/reviewer AI commands without invoking provider CLIs.
+- Redacted likely secrets from provider stdout, stderr, and error messages.
+- Added focused tests for doctor profiles, prompt-only output, missing CLI guidance, structured provider metadata, and log redaction.
+- Updated README and command reference docs.
+
+## Pending
+
+No pending implementation for this slice.
+
+## Remaining Risks
+
+- Redaction remains best-effort and should not be treated as a reason to intentionally print secrets.
+- `ai doctor` does not yet use the doctor provider profile; it remains a local readiness/GitHub preflight command.
+
+## Future Recommendations
+
+- Consider a later explicit `ai doctor --provider` diagnostic flow if the doctor profile should run model-assisted diagnosis.

package/specs/quiver-v25-ai-first-lifecycle-orchestrator/slices/slice-04-agent-profiles-adapters/EXECUTION_BRIEF.md

@@ -0,0 +1,54 @@
+# EXECUTION BRIEF - slice-04: Agent profiles and provider adapters
+
+## Context
+
+Quiver must support different model CLIs without hard-coding one provider or storing credentials.
+
+## Objective
+
+Add provider-agnostic agent profiles and adapter contracts, starting with prompt-only behavior.
+
+## Scope
+
+- Planner, executor, reviewer, and doctor profiles.
+- Provider command validation.
+- Prompt rendering.
+- Structured command result capture.
+- Secret redaction for adapter logs.
+
+## Acceptance Criteria
+
+- Profiles store labels and command preferences, not secrets.
+- Prompt-only mode works.
+- Missing CLI errors are actionable.
+- Execution metadata is captured when execution is enabled.
+
+## Technical Plan Summary
+
+Build a small adapter layer around external CLIs with explicit input/output contracts and dry-run-safe behavior.
+
+## Suggested Execution Steps
+
+1. Inspect current agent profile code.
+2. Define adapter interface.
+3. Implement prompt-only path.
+4. Add CLI availability checks.
+5. Add structured result capture and redaction.
+6. Test with mocked provider commands.
+
+## Restrictions
+
+- Do not require real provider credentials in tests.
+- Do not store API keys or tokens.
+
+## Risks
+
+- Provider CLIs differ substantially. Keep the first contract conservative.
+
+## Completion Checklist
+
+- [ ] Profile schema updated.
+- [ ] Prompt-only path validated.
+- [ ] Missing CLI errors tested.
+- [ ] Redaction tested.
+- [ ] Evidence appended.

package/specs/quiver-v25-ai-first-lifecycle-orchestrator/slices/slice-04-agent-profiles-adapters/slice.json

@@ -0,0 +1,52 @@
+{
+  "slice_id": "slice-04-agent-profiles-adapters",
+  "ticket": "QUIVER-25-04",
+  "type": "feature",
+  "title": "Agent profiles and provider adapters",
+  "objective": "Create a safe provider-agnostic contract for planner, executor, reviewer, and doctor agents.",
+  "description": "Add profile and adapter behavior for Codex, Claude, Gemini, or custom commands, starting with prompt-only and dry-run-safe execution.",
+  "git": {
+    "branch_type": "feature",
+    "base_branch": "main",
+    "branch_slug": "agent-profiles-adapters",
+    "branch_name": "feature/QUIVER-25-04-agent-profiles-adapters"
+  },
+  "files": [
+    "src/create-quiver/**",
+    "docs/**",
+    "tests/**"
+  ],
+  "depends_on": [
+    "slice-02-run-state-phase-locks"
+  ],
+  "parallel_safe": "yes",
+  "parallel_safe_reason": "Can run beside onboarding docs if adapter files are isolated.",
+  "must": [
+    "Support planner, executor, reviewer, and doctor profiles.",
+    "Store provider/model labels but never credentials.",
+    "Support prompt-only output with `--print-prompt` or equivalent.",
+    "Validate provider CLI availability before execution.",
+    "Capture stdout, stderr, exit code, timeout, and sanitized logs.",
+    "Define fallback behavior when a CLI is missing."
+  ],
+  "not_included": [
+    "Approval gate implementation.",
+    "Spec generation.",
+    "Automatic multi-agent scheduling."
+  ],
+  "acceptance": [
+    "Profiles can be configured and inspected without secrets.",
+    "Prompt-only mode works without provider authentication.",
+    "Missing provider CLI produces actionable guidance.",
+    "Adapter output is captured in a structured way.",
+    "Logs redact likely secrets before writing."
+  ],
+  "tests": [
+    "node --test tests/**/*.test.js",
+    "git diff --check"
+  ],
+  "estimated_hours": 7,
+  "status": "completed",
+  "completed_at": "2026-05-22T16:21:54-0300",
+  "blocked_reason": null
+}

package/specs/quiver-v25-ai-first-lifecycle-orchestrator/slices/slice-05-approval-gates/CLOSURE_BRIEF.md

@@ -0,0 +1,34 @@
+# CLOSURE BRIEF - slice-05: Approval gates and planner iterations
+
+## Summary of Work
+
+Implemented versioned planner approval gates for acceptance and technical-plan phases. `ai revise` now creates a new draft version from human feedback without approving or advancing the phase, while `ai approve` requires an explicit current draft version. Technical-plan approval now requires a production-readiness review for the current draft before spec generation can proceed.
+
+## Validation Against Acceptance Criteria
+
+- [x] Criteria revision flow verified.
+- [x] Criteria approval verified.
+- [x] Plan revision flow verified.
+- [x] Plan approval verified.
+- [x] Premature spec generation blocked.
+
+## Relevant Changes
+
+- Added `npx create-quiver ai revise`.
+- Changed `ai approve` to approve saved draft versions only with `--version <n>`.
+- Blocked approval of missing or non-current draft versions.
+- Required current production-readiness review before technical-plan approval.
+- Updated flow guidance, README, generated docs, init templates, and command reference.
+- Added regression tests for revise, approval versioning, stale review blocking, and spec generation gates.
+
+## Pending
+
+None for this slice.
+
+## Remaining Risks
+
+- Technical-plan review checks are local state based; external provider quality still depends on the configured reviewer prompt and model.
+
+## Future Recommendations
+
+- Keep approval messages concise and cite the approved draft version.

package/specs/quiver-v25-ai-first-lifecycle-orchestrator/slices/slice-05-approval-gates/EXECUTION_BRIEF.md

@@ -0,0 +1,54 @@
+# EXECUTION BRIEF - slice-05: Approval gates and planner iterations
+
+## Context
+
+Human approval is central to Quiver. Planner output must be iterated and approved before later phases can run.
+
+## Objective
+
+Persist criteria and technical-plan drafts, revisions, and approvals with strict phase gates.
+
+## Scope
+
+- Acceptance criteria drafts.
+- Technical-plan drafts.
+- Revise commands.
+- Approval commands.
+- Approval state and phase blocking.
+
+## Acceptance Criteria
+
+- Drafts are versioned.
+- Revisions do not advance phase.
+- Approvals require concrete versions.
+- Spec generation is blocked before approval.
+- Approval history survives session changes.
+
+## Technical Plan Summary
+
+Use the run state from slice-02 and adapter prompts from slice-04 to manage planner drafts and approvals.
+
+## Suggested Execution Steps
+
+1. Define draft storage layout.
+2. Add criteria draft/revise/approve behavior.
+3. Add technical-plan draft/revise/review/approve behavior.
+4. Wire phase guards.
+5. Test invalid transitions.
+
+## Restrictions
+
+- Do not create spec/slice files in this slice.
+- Do not execute implementation agents.
+
+## Risks
+
+- Loose approval semantics can let users approve the wrong version. Require explicit version IDs.
+
+## Completion Checklist
+
+- [ ] Criteria gate tested.
+- [ ] Technical-plan gate tested.
+- [ ] Invalid transitions tested.
+- [ ] Approval state documented.
+- [ ] Evidence appended.

package/specs/quiver-v25-ai-first-lifecycle-orchestrator/slices/slice-05-approval-gates/slice.json

@@ -0,0 +1,53 @@
+{
+  "slice_id": "slice-05-approval-gates",
+  "ticket": "QUIVER-25-05",
+  "type": "feature",
+  "title": "Approval gates and planner iterations",
+  "objective": "Implement human approval gates for criteria, technical plans, and production-readiness review.",
+  "description": "Persist versioned planner drafts, allow revisions, require explicit approval, and block later phases until approvals exist.",
+  "git": {
+    "branch_type": "feature",
+    "base_branch": "main",
+    "branch_slug": "approval-gates-planner-iterations",
+    "branch_name": "feature/QUIVER-25-05-approval-gates-planner-iterations"
+  },
+  "files": [
+    "src/create-quiver/**",
+    "docs/**",
+    "tests/**"
+  ],
+  "depends_on": [
+    "slice-02-run-state-phase-locks",
+    "slice-04-agent-profiles-adapters"
+  ],
+  "parallel_safe": "no",
+  "parallel_safe_reason": "Depends on shared state and agent profile contracts.",
+  "must": [
+    "Store versioned acceptance criteria drafts.",
+    "Store versioned technical-plan drafts.",
+    "Support revise commands that do not advance phase.",
+    "Approve only concrete versions.",
+    "Block spec generation until the plan is approved.",
+    "Record approval decisions and timestamps."
+  ],
+  "not_included": [
+    "Spec file generation.",
+    "Slice execution.",
+    "PR creation."
+  ],
+  "acceptance": [
+    "Criteria drafts can be generated, revised, and approved.",
+    "Technical-plan drafts can be generated, revised, reviewed, and approved.",
+    "Approving a missing or non-current version fails clearly.",
+    "Spec generation is blocked before plan approval.",
+    "Approvals are readable from state without chat memory."
+  ],
+  "tests": [
+    "node --test tests/**/*.test.js",
+    "git diff --check"
+  ],
+  "estimated_hours": 7,
+  "status": "completed",
+  "blocked_reason": null,
+  "completed_at": "2026-05-22T16:45:36-0300"
+}

package/specs/quiver-v25-ai-first-lifecycle-orchestrator/slices/slice-06-spec-slice-generator/CLOSURE_BRIEF.md

@@ -0,0 +1,33 @@
+# CLOSURE BRIEF - slice-06: Spec, slice, handoff, and PR body generation
+
+## Summary of Work
+
+Completed the generated artifact contract for specs and slices. Generated `slice.json` files now include dependency data, parallel safety, expected read paths, allowed write paths, and validation hints, and generated `EXECUTION_BRIEF.md` files expose the same execution contract for agents.
+
+## Validation Against Acceptance Criteria
+
+- [x] Blocked before plan approval.
+- [x] Dry-run verified.
+- [x] Spec artifacts generated.
+- [x] Slice handoffs generated.
+- [x] JSON validation passed.
+
+## Relevant Changes
+
+- Added generated `expected_read_paths`, `allowed_write_paths`, and `validation_hints` to `slice.json`.
+- Added validation that generated slice JSON contains those arrays.
+- Added the same read/write/validation sections to generated execution briefs.
+- Preserved backward compatibility by deriving `allowed_write_paths` from existing `files` and deriving `files` from explicit allowed write paths when needed.
+- Added fixture coverage for explicit and default scope declarations.
+
+## Pending
+
+None for this slice.
+
+## Remaining Risks
+
+- Generated scopes are only as accurate as the approved planner output and safe defaults.
+
+## Future Recommendations
+
+- Use generated artifacts to dogfood the next Quiver spec.

package/specs/quiver-v25-ai-first-lifecycle-orchestrator/slices/slice-06-spec-slice-generator/EXECUTION_BRIEF.md

@@ -0,0 +1,56 @@
+# EXECUTION BRIEF - slice-06: Spec, slice, handoff, and PR body generation
+
+## Context
+
+Once the technical plan is approved, Quiver must turn it into durable artifacts that agents and humans can execute.
+
+## Objective
+
+Generate complete spec/slice/handoff/PR artifacts from approved planner output.
+
+## Scope
+
+- Spec folder creation.
+- Mandatory `slice-00`.
+- Slice JSON and handoff generation.
+- `EXECUTION_PLAN.md`, `EVIDENCE_REPORT.md`, `STATUS.md`, and `pr.md`.
+- File scope declarations.
+- Dry-run previews.
+
+## Acceptance Criteria
+
+- Generation is blocked before plan approval.
+- Dry-run lists all proposed files.
+- Artifacts follow Quiver layout.
+- Every slice has handoffs and parseable JSON.
+- `slice-00` exists.
+
+## Technical Plan Summary
+
+Transform approved planner output into the canonical `specs/<spec-slug>/` structure and validate before writing.
+
+## Suggested Execution Steps
+
+1. Define artifact generator contracts.
+2. Add dry-run preview.
+3. Generate top-level files.
+4. Generate slices and handoffs.
+5. Validate JSON and required files.
+6. Test with fixture planner outputs.
+
+## Restrictions
+
+- Do not run execution agents.
+- Do not commit or open PRs.
+
+## Risks
+
+- Poor planner output may create unusable slices. Validate and fail with actionable messages.
+
+## Completion Checklist
+
+- [ ] Dry-run tested.
+- [ ] Generated files validated.
+- [ ] Mandatory slice-00 tested.
+- [ ] PR body tested.
+- [ ] Evidence appended.

package/specs/quiver-v25-ai-first-lifecycle-orchestrator/slices/slice-06-spec-slice-generator/slice.json

@@ -0,0 +1,55 @@
+{
+  "slice_id": "slice-06-spec-slice-generator",
+  "ticket": "QUIVER-25-06",
+  "type": "feature",
+  "title": "Spec, slice, handoff, and PR body generation",
+  "objective": "Generate versionable implementation artifacts only after the technical plan is approved.",
+  "description": "Create SPEC, STATUS, evidence, execution plan, PR body, mandatory slice-00, slice JSON, execution briefs, closure briefs, and file scope declarations from approved planner output.",
+  "git": {
+    "branch_type": "feature",
+    "base_branch": "main",
+    "branch_slug": "spec-slice-generator",
+    "branch_name": "feature/QUIVER-25-06-spec-slice-generator"
+  },
+  "files": [
+    "src/create-quiver/**",
+    "docs/**",
+    "specs/**",
+    "tests/**"
+  ],
+  "depends_on": [
+    "slice-05-approval-gates"
+  ],
+  "parallel_safe": "no",
+  "parallel_safe_reason": "Creates the artifact contract consumed by planning and execution slices.",
+  "must": [
+    "Generate top-level spec artifacts from approved plan only.",
+    "Always create mandatory slice-00.",
+    "Create slice.json, EXECUTION_BRIEF.md, and CLOSURE_BRIEF.md for each slice.",
+    "Create pr.md body.",
+    "Include depends_on, parallel safety, expected read paths, allowed write paths, and validation hints.",
+    "Support dry-run previews before writing."
+  ],
+  "not_included": [
+    "Executing slices.",
+    "Opening PRs.",
+    "Dashboard export."
+  ],
+  "acceptance": [
+    "Spec generation is blocked before plan approval.",
+    "Dry-run reports all files that would be written.",
+    "Written artifacts match Quiver spec layout.",
+    "slice-00 is present and must run first.",
+    "Every slice has handoffs and parseable JSON.",
+    "PR body is generated."
+  ],
+  "tests": [
+    "node --test tests/**/*.test.js",
+    "find <fixture-output> -name \"slice.json\" -print -exec node -e \"JSON.parse(require('fs').readFileSync(process.argv[1], 'utf8'))\" {} \\;",
+    "git diff --check"
+  ],
+  "estimated_hours": 8,
+  "status": "completed",
+  "blocked_reason": null,
+  "completed_at": "2026-05-22T16:49:40-0300"
+}

package/specs/quiver-v25-ai-first-lifecycle-orchestrator/slices/slice-07-slice-execution-planner/CLOSURE_BRIEF.md

@@ -0,0 +1,33 @@
+# CLOSURE BRIEF - slice-07: Slice execution planning and parallel safety
+
+## Summary of Work
+
+Connected execution planning to the generated write-scope contract. Slice graph reads now use `allowed_write_paths` as the authoritative write scope when present, execution plans expose read/write/validation metadata for downstream agents, and human output labels dependency levels as waves with parallel-safety rationale.
+
+## Validation Against Acceptance Criteria
+
+- [x] Wave 0 verified.
+- [x] Parallel grouping verified.
+- [x] Conflict blocking verified.
+- [x] JSON output verified.
+- [x] Tests run.
+
+## Relevant Changes
+
+- `readAllSlices` now resolves write scope from `allowed_write_paths` before falling back to `files`.
+- Execution plan JSON includes `expected_read_paths`, `allowed_write_paths`, and `validation_hints` for each slice.
+- Human execution plan output now uses `Wave <n>` and prints each slice's parallel-safety declaration.
+- Added conflict tests for slices that only declare `allowed_write_paths`.
+- Added CLI JSON coverage for `ai execute-plan --json`.
+
+## Pending
+
+None for this slice.
+
+## Remaining Risks
+
+- Parallel safety still assumes declared write paths are accurate; later validation should flag incomplete scopes earlier.
+
+## Future Recommendations
+
+- Use this planner before any delegated multi-agent execution.

package/specs/quiver-v25-ai-first-lifecycle-orchestrator/slices/slice-07-slice-execution-planner/EXECUTION_BRIEF.md

@@ -0,0 +1,54 @@
+# EXECUTION BRIEF - slice-07: Slice execution planning and parallel safety
+
+## Context
+
+After specs and slices exist, Quiver must tell the human and agents what can run now, what must wait, and what can safely run in parallel.
+
+## Objective
+
+Generate execution plans with waves, dependencies, file conflicts, and JSON output.
+
+## Scope
+
+- Execution waves.
+- Dependency validation.
+- Parallel-safety rules.
+- File scope conflict checks.
+- Human and machine-readable output.
+
+## Acceptance Criteria
+
+- `slice-00` is always Wave 0.
+- Parallel waves exclude conflicting slices.
+- Blocked slices explain why.
+- JSON output is available.
+- Existing planning commands remain compatible.
+
+## Technical Plan Summary
+
+Extend current slice graph/planning machinery with file-scope-aware waves and clearer output.
+
+## Suggested Execution Steps
+
+1. Inspect current slice graph utilities.
+2. Add file conflict handling based on declared paths.
+3. Add wave output.
+4. Add JSON output tests.
+5. Validate with fixtures.
+
+## Restrictions
+
+- Do not execute slices.
+- Do not create commits.
+
+## Risks
+
+- False parallel safety can cause agents to edit the same files. Prefer conservative blocking.
+
+## Completion Checklist
+
+- [ ] Wave planning tested.
+- [ ] Conflict detection tested.
+- [ ] JSON output tested.
+- [ ] Docs updated.
+- [ ] Evidence appended.

package/specs/quiver-v25-ai-first-lifecycle-orchestrator/slices/slice-07-slice-execution-planner/slice.json

@@ -0,0 +1,52 @@
+{
+  "slice_id": "slice-07-slice-execution-planner",
+  "ticket": "QUIVER-25-07",
+  "type": "feature",
+  "title": "Slice execution planning and parallel safety",
+  "objective": "Generate a reliable execution plan that shows required order, parallel waves, dependencies, risks, and file conflicts.",
+  "description": "Turn generated slices into execution waves while enforcing slice-00 first, dependency validation, parallel safety, and conflict detection.",
+  "git": {
+    "branch_type": "feature",
+    "base_branch": "main",
+    "branch_slug": "slice-execution-planner",
+    "branch_name": "feature/QUIVER-25-07-slice-execution-planner"
+  },
+  "files": [
+    "src/create-quiver/**",
+    "docs/**",
+    "tests/**"
+  ],
+  "depends_on": [
+    "slice-06-spec-slice-generator"
+  ],
+  "parallel_safe": "no",
+  "parallel_safe_reason": "Uses generated artifact contracts and feeds executor behavior.",
+  "must": [
+    "Generate sequential and parallel execution waves.",
+    "Force slice-00 to execute first.",
+    "Detect dependency cycles and missing dependencies.",
+    "Detect overlapping allowed write paths between parallel slices.",
+    "Explain why a slice is blocked or parallel-safe.",
+    "Emit human and JSON output."
+  ],
+  "not_included": [
+    "Executing agents.",
+    "Creating commits.",
+    "Opening PRs."
+  ],
+  "acceptance": [
+    "Execution plan shows Wave 0 for slice-00.",
+    "Parallel-ready slices are grouped only when dependencies and file scopes allow it.",
+    "Unsafe parallel execution is explained.",
+    "JSON output is suitable for downstream agents or dashboards.",
+    "Existing `plan`, `graph`, and `next` behavior remains compatible."
+  ],
+  "tests": [
+    "node --test tests/**/*.test.js",
+    "git diff --check"
+  ],
+  "estimated_hours": 6,
+  "status": "completed",
+  "blocked_reason": null,
+  "completed_at": "2026-05-22T16:52:33-0300"
+}

package/specs/quiver-v25-ai-first-lifecycle-orchestrator/slices/slice-08-controlled-slice-execution/CLOSURE_BRIEF.md

@@ -0,0 +1,39 @@
+# CLOSURE BRIEF - slice-08: Controlled slice execution and evidence
+
+## Summary of Work
+
+Implemented controlled slice execution and evidence closure. Direct `ai execute-slice` now uses the generated write-scope contract, blocks dirty or wrong worktrees, validates changed files against declared scope, writes closure/evidence/status artifacts after successful execution, redacts likely secrets from logs, and refuses to close no-op provider runs.
+
+## Validation Against Acceptance Criteria
+
+- [x] Prompt-only mode verified.
+- [x] Minimal context verified.
+- [x] Worktree preflight verified.
+- [x] Scope validation verified.
+- [x] Closure/evidence update verified.
+- [x] Secret redaction verified.
+- [x] No-op provider protection verified.
+
+## Relevant Changes
+
+- `resolveSliceContext` now prefers `allowed_write_paths` over legacy `files` and exposes generated read/validation metadata.
+- `validateScopeSnapshot` supports simple glob scopes and reports violations against the declared slice scope.
+- `ai execute-slice` validates the declared slice branch, clean tree, provider changes, scope, validations, and lifecycle artifacts.
+- Successful execution updates `CLOSURE_BRIEF.md`, `EVIDENCE_REPORT.md`, `COMMAND_LOG.md`, `STATUS.md`, and `slice.json`.
+- Provider and validation output are redacted before printing, returning, or writing evidence.
+- `ai execute-plan` passes an internal branch-check bypass because delegated worktrees are managed by the orchestrator.
+- README and generated command/workflow docs describe the controlled execution contract.
+
+## Pending
+
+None for this slice.
+
+## Remaining Risks
+
+- Scope glob support is intentionally lightweight and should stay covered by regression tests if Quiver adds richer path patterns.
+- Direct single-slice execution is stricter than orchestrated delegated execution by design; docs should keep that distinction visible.
+
+## Future Recommendations
+
+- Use the captured evidence as input for PR creation and review.
+- In `slice-10`, keep adding fixtures around path scopes, redaction, and worktree failures.

package/specs/quiver-v25-ai-first-lifecycle-orchestrator/slices/slice-08-controlled-slice-execution/EXECUTION_BRIEF.md

@@ -0,0 +1,56 @@
+# EXECUTION BRIEF - slice-08: Controlled slice execution and evidence
+
+## Context
+
+Executors should be cheap and focused. They need only the slice context, clear restrictions, and a closure contract.
+
+## Objective
+
+Enable controlled slice execution with minimal context, file scope validation, closure updates, evidence, and logs.
+
+## Scope
+
+- Executor prompt/context generation.
+- Prompt-only and execution modes.
+- Worktree and clean-tree checks.
+- File scope validation.
+- Closure/evidence/command-log updates.
+- Secret redaction.
+
+## Acceptance Criteria
+
+- Context excludes unrelated files by default.
+- Prompt-only mode works.
+- Dirty or wrong worktree blocks execution.
+- Out-of-scope writes are detected.
+- Closure and evidence are updated.
+
+## Technical Plan Summary
+
+Use execution-plan output and agent adapters to prepare focused executor handoffs and validate results before a slice can close.
+
+## Suggested Execution Steps
+
+1. Define executor context pack contract.
+2. Add prompt-only slice command.
+3. Add preflight checks.
+4. Add file scope diff validation.
+5. Add closure/evidence updates.
+6. Test manual and delegated paths with fixtures.
+
+## Restrictions
+
+- Do not auto-commit in this slice unless already required by existing command behavior.
+- Do not open PRs.
+
+## Risks
+
+- File scope validation can miss generated files if path rules are too strict. Report clearly and allow explicit review.
+
+## Completion Checklist
+
+- [ ] Context pack tested.
+- [ ] Preflight tested.
+- [ ] Scope validation tested.
+- [ ] Closure/evidence updated.
+- [ ] Evidence appended.