@izkac/forgekit 0.3.12 → 0.3.13

package/vendor/skills/forge/references/test-strategy.md

@@ -1,68 +1,68 @@
-# Tiered test strategy (Forge)
-
-Forge uses **three tiers**. Do **not** run the full workspace or monorepo test suite on every subagent task — that is slow and redundant when verify runs tier 3 once.
-
-## Tier 1 — TDD loop (within a task)
-
-During red/green/refactor, run **scoped** commands only:
-
-```bash
-npm test -- path/to/scoped.test.ts -- path/to/changed.test.ts
-```
-
-- Single test file or pattern for the behavior under change
-- Mandatory for TDD integrity (watch fail → pass)
-- Fast; repeat as often as needed
-
-## Tier 2 — Task completion evidence
-
-Before marking a task complete, run the **narrowest** command that proves **this task's** changes work:
-
-- Changed test file(s) + directly related tests in the same module/directory
-- Subpath or pattern filter when the runner supports it
-- **Not** the full workspace suite by default
-
-Record in `.forge/sessions/<id>/tasks/<nn>-<slug>/test-evidence.md`. See [test-evidence.md](./test-evidence.md).
-
-### Tier 2 may be wider (still not full monorepo)
-
-- Shared contract or public API change → include contract tests in affected packages
-- Cross-module integration within one workspace → run the relevant integration test directory
-- Task brief explicitly requires full workspace coverage
-
-### Do not use full workspace at tier 2 when
-
-- Changes are localized (typical 1–3 file task)
-- Multiple tasks target the same workspace (full suite runs once at **verify**)
-- Motivation is only "to be safe" — that is **verify's** job
-
-## Tier 3 — Verify phase (coordinator, after all tasks)
-
-After all implement tasks complete:
-
-1. **Audit** every `tasks/<nn>-<slug>/test-evidence.md` (exit `0`, pass summary, reviewers approved).
-2. Run **one fresh full workspace test** for each affected workspace (plus downstream consumer workspaces when contracts changed).
-3. Save output to `.forge/sessions/<id>/verify-evidence.md`.
-
-Tier 3 is the integration/regression gate. Cite this file when claiming the implementation passes.
-
-Re-run tier 3 when: tier 3 failed, coordinator edited code after verify evidence, quality/final review flagged test gaps, or the user asks for a fresh run.
-
-**Do not** re-run tier 2 narrow commands at verify — audit them instead.
-
-## Quick reference
-
-| Tier | Who | When | Scope |
-| ---- | --- | ---- | ----- |
-| 1 | Implementer | Each TDD red/green/refactor | Single file or pattern |
-| 2 | Implementer | Task done | Narrowest proof for this task |
-| 3 | Coordinator | Verify phase (once) | Full affected workspace(s) |
-
-## Anti-patterns
-
-| Don't | Do instead |
-| ----- | ---------- |
-| Full workspace test on every task | Tier 1 scoped + tier 2 narrow evidence |
-| Skip tier 3 because tasks passed | Always run tier 3 once at verify |
-| Re-run tier 2 commands at verify | Audit task evidence; run tier 3 full suite |
-| Full monorepo test at tier 2 | Tier 2 narrow; tier 3 per workspace (or listed consumers) |
+# Tiered test strategy (Forge)
+
+Forge uses **three tiers**. Do **not** run the full workspace or monorepo test suite on every subagent task — that is slow and redundant when verify runs tier 3 once.
+
+## Tier 1 — TDD loop (within a task)
+
+During red/green/refactor, run **scoped** commands only:
+
+```bash
+npm test -- path/to/scoped.test.ts -- path/to/changed.test.ts
+```
+
+- Single test file or pattern for the behavior under change
+- Mandatory for TDD integrity (watch fail → pass)
+- Fast; repeat as often as needed
+
+## Tier 2 — Task completion evidence
+
+Before marking a task complete, run the **narrowest** command that proves **this task's** changes work:
+
+- Changed test file(s) + directly related tests in the same module/directory
+- Subpath or pattern filter when the runner supports it
+- **Not** the full workspace suite by default
+
+Record in `.forge/sessions/<id>/tasks/<nn>-<slug>/test-evidence.md`. See [test-evidence.md](./test-evidence.md).
+
+### Tier 2 may be wider (still not full monorepo)
+
+- Shared contract or public API change → include contract tests in affected packages
+- Cross-module integration within one workspace → run the relevant integration test directory
+- Task brief explicitly requires full workspace coverage
+
+### Do not use full workspace at tier 2 when
+
+- Changes are localized (typical 1–3 file task)
+- Multiple tasks target the same workspace (full suite runs once at **verify**)
+- Motivation is only "to be safe" — that is **verify's** job
+
+## Tier 3 — Verify phase (coordinator, after all tasks)
+
+After all implement tasks complete:
+
+1. **Audit** every `tasks/<nn>-<slug>/test-evidence.md` (exit `0`, pass summary, reviewers approved).
+2. Run **one fresh full workspace test** for each affected workspace (plus downstream consumer workspaces when contracts changed).
+3. Save output to `.forge/sessions/<id>/verify-evidence.md`.
+
+Tier 3 is the integration/regression gate. Cite this file when claiming the implementation passes.
+
+Re-run tier 3 when: tier 3 failed, coordinator edited code after verify evidence, quality/final review flagged test gaps, or the user asks for a fresh run.
+
+**Do not** re-run tier 2 narrow commands at verify — audit them instead.
+
+## Quick reference
+
+| Tier | Who | When | Scope |
+| ---- | --- | ---- | ----- |
+| 1 | Implementer | Each TDD red/green/refactor | Single file or pattern |
+| 2 | Implementer | Task done | Narrowest proof for this task |
+| 3 | Coordinator | Verify phase (once) | Full affected workspace(s) |
+
+## Anti-patterns
+
+| Don't | Do instead |
+| ----- | ---------- |
+| Full workspace test on every task | Tier 1 scoped + tier 2 narrow evidence |
+| Skip tier 3 because tasks passed | Always run tier 3 once at verify |
+| Re-run tier 2 commands at verify | Audit task evidence; run tier 3 full suite |
+| Full monorepo test at tier 2 | Tier 2 narrow; tier 3 per workspace (or listed consumers) |

package/vendor/skills/forge/skills/subagent-driven-development/SKILL.md

@@ -1,87 +1,87 @@
----
-name: subagent-driven-development
-description: Forge — subagent per task with a combined spec+quality review. Internal skill; read via forge orchestrator.
----
-
-# Subagent-Driven Development
-
-Execute a plan by dispatching a fresh implementer subagent per task, followed by one **task reviewer** subagent that checks spec compliance first, then code quality.
-
-**Why subagents:** isolated context per task. You craft exactly the instructions and context each one needs — they never inherit session history — and your own context stays free for coordination.
-
-**Core principle:** fresh implementer per task + combined review (spec gate, then quality) = high quality, fast iteration.
-
-## When to use
-
-Use when you have an implementation plan with mostly independent tasks and you're staying in this session. Tightly coupled tasks or no plan → execute manually or brainstorm first.
-
-## Pace
-
-Honor [../../references/pace.md](../../references/pace.md) (`forge prefs` / session `resolvedPace`). Reviewers may be deferred to **OpenSpec group** boundaries under `standard` (`per-group`), or skipped for low-risk work under `brisk`/`lite`; **never** skip review for money/auth/contracts/migrations (immediate per-task review). Tier-2 test evidence stays mandatory for behavior changes on every task.
-
-## Per-task loop
-
-1. Extract full task text + context from the plan (read the plan file **once**, up front; never make a subagent read it). Note which OpenSpec **group** (`##` section) the task belongs to.
-2. Dispatch **implementer** — [../../subagents/implementer-prompt.md](../../subagents/implementer-prompt.md). Answer any questions it asks before letting it proceed.
-3. Dispatch **reviewer** when pace requires it — [../../subagents/task-reviewer-prompt.md](../../subagents/task-reviewer-prompt.md) (spec compliance gates quality):
-   - `always` → after this task
-   - `per-group` → after the last task in the current `##` group (or immediately if high-risk); mid-group low-risk → self-check only
-   - `high-risk-only` / `never` → only high-risk (hard floor)
-   If skipped, write a pace self-check `task-review.md`. Group reviews write `group-review.md` covering every task in the section.
-4. Reviewer REJECTED → same implementer fixes → re-review. Repeat until APPROVED (cap at `review.maxRounds`). Never skip the re-review when a reviewer was dispatched.
-5. Save test evidence; mark task complete.
-6. After all tasks: proceed to verify/review phases (final reviewer subject to pace).
-
-## Task batching (small tasks only)
-
-Each dispatch pays a full fresh context. Batch **consecutive, small, same-area tasks** (doc edits, config values, wording, mechanical renames) into one implementer brief listing them all, reviewed together in one review pass.
-
-**Never batch** tasks that touch money, auth, shared contracts, or migrations — those keep 1:1 dispatch. When in doubt, don't batch.
-
-## Model selection
-
-Subagent models use **two axes** — never invent host model IDs from memory:
-
-| Axis | Values | Default |
-| ---- | ------ | ------- |
-| Capability | `fast` · `standard` · `capable` | role-based (below) |
-| Billing | `included` · `metered` | **`included`** |
-
-**Billing `included` unless the user explicitly asks for API/metered models** (or has set `forge models -- metered`). Do **not** auto-switch to `metered` on failure.
-
-Capability by role:
-
-- Touches 1–2 files with a complete spec → `fast` (most tasks, when the plan is well-specified)
-- Multi-file integration, pattern matching, debugging → `standard`
-- Design judgment, broad codebase understanding, review → `capable`
-
-Before every Task/Agent dispatch, resolve:
-
-```bash
-forge resolve-model --tier <fast|standard|capable>
-```
-
-Honor the JSON: if `omitModel` is true, **omit** the host `model` parameter; otherwise pass `model` exactly as returned. Defaults live in `models.defaults.json`. Checkout overlay `.forge/models.local.json` exists only after `forge models included|metered` (bare `forge models` only prints).
-
-## Handling implementer status
-
-- **DONE** → proceed to review.
-- **DONE_WITH_CONCERNS** → read the concerns; correctness/scope concerns get addressed before review, observations are noted.
-- **NEEDS_CONTEXT** → provide the missing context, re-dispatch.
-- **BLOCKED** → context problem: add context, re-dispatch. Needs more reasoning: escalate **capability tier** within the current billing lane (still `included` by default). Task too large: split it. Plan wrong: escalate to the human. Never force the same model to retry unchanged; never flip to `metered` without an explicit user request.
-
-## Red flags — never
-
-- Start implementation on main/master without explicit user consent
-- Skip the review, proceed with unfixed spec gaps or Critical/Important issues, or accept "close enough" on spec compliance
-- Dispatch multiple implementer subagents in parallel (conflicts)
-- Make a subagent read the plan file, skip scene-setting context, or ignore its questions
-- Let implementer self-review replace the reviewer (both are needed)
-- Fix a failed task manually in coordinator context — dispatch a fix subagent with specific instructions
-
-## Integration
-
-- **Plan source:** OpenSpec (`openspec/changes/<name>/tasks.md`)
-- **Subagents must follow:** [references/tdd-core.md](../../references/tdd-core.md) (condensed TDD rules; full skill at [skills/test-driven-development](../test-driven-development/SKILL.md) when stuck)
-- **On blockers:** [skills/systematic-debugging](../systematic-debugging/SKILL.md) before guessing fixes
-- **After all tasks:** [phases/verify.md](../../phases/verify.md) → [phases/review.md](../../phases/review.md) → [phases/finish.md](../../phases/finish.md)
+---
+name: subagent-driven-development
+description: Forge — subagent per task with a combined spec+quality review. Internal skill; read via forge orchestrator.
+---
+
+# Subagent-Driven Development
+
+Execute a plan by dispatching a fresh implementer subagent per task, followed by one **task reviewer** subagent that checks spec compliance first, then code quality.
+
+**Why subagents:** isolated context per task. You craft exactly the instructions and context each one needs — they never inherit session history — and your own context stays free for coordination.
+
+**Core principle:** fresh implementer per task + combined review (spec gate, then quality) = high quality, fast iteration.
+
+## When to use
+
+Use when you have an implementation plan with mostly independent tasks and you're staying in this session. Tightly coupled tasks or no plan → execute manually or brainstorm first.
+
+## Pace
+
+Honor [../../references/pace.md](../../references/pace.md) (`forge prefs` / session `resolvedPace`). Reviewers may be deferred to **OpenSpec group** boundaries under `standard` (`per-group`), or skipped for low-risk work under `brisk`/`lite`; **never** skip review for money/auth/contracts/migrations (immediate per-task review). Tier-2 test evidence stays mandatory for behavior changes on every task.
+
+## Per-task loop
+
+1. Extract full task text + context from the plan (read the plan file **once**, up front; never make a subagent read it). Note which OpenSpec **group** (`##` section) the task belongs to.
+2. Dispatch **implementer** — [../../subagents/implementer-prompt.md](../../subagents/implementer-prompt.md). Answer any questions it asks before letting it proceed.
+3. Dispatch **reviewer** when pace requires it — [../../subagents/task-reviewer-prompt.md](../../subagents/task-reviewer-prompt.md) (spec compliance gates quality):
+   - `always` → after this task
+   - `per-group` → after the last task in the current `##` group (or immediately if high-risk); mid-group low-risk → self-check only
+   - `high-risk-only` / `never` → only high-risk (hard floor)
+   If skipped, write a pace self-check `task-review.md`. Group reviews write `group-review.md` covering every task in the section.
+4. Reviewer REJECTED → same implementer fixes → re-review. Repeat until APPROVED (cap at `review.maxRounds`). Never skip the re-review when a reviewer was dispatched.
+5. Save test evidence; mark task complete.
+6. After all tasks: proceed to verify/review phases (final reviewer subject to pace).
+
+## Task batching (small tasks only)
+
+Each dispatch pays a full fresh context. Batch **consecutive, small, same-area tasks** (doc edits, config values, wording, mechanical renames) into one implementer brief listing them all, reviewed together in one review pass.
+
+**Never batch** tasks that touch money, auth, shared contracts, or migrations — those keep 1:1 dispatch. When in doubt, don't batch.
+
+## Model selection
+
+Subagent models use **two axes** — never invent host model IDs from memory:
+
+| Axis | Values | Default |
+| ---- | ------ | ------- |
+| Capability | `fast` · `standard` · `capable` | role-based (below) |
+| Billing | `included` · `metered` | **`included`** |
+
+**Billing `included` unless the user explicitly asks for API/metered models** (or has set `forge models -- metered`). Do **not** auto-switch to `metered` on failure.
+
+Capability by role:
+
+- Touches 1–2 files with a complete spec → `fast` (most tasks, when the plan is well-specified)
+- Multi-file integration, pattern matching, debugging → `standard`
+- Design judgment, broad codebase understanding, review → `capable`
+
+Before every Task/Agent dispatch, resolve:
+
+```bash
+forge resolve-model --tier <fast|standard|capable>
+```
+
+Honor the JSON: if `omitModel` is true, **omit** the host `model` parameter; otherwise pass `model` exactly as returned. Defaults live in `models.defaults.json`. Checkout overlay `.forge/models.local.json` exists only after `forge models included|metered` (bare `forge models` only prints).
+
+## Handling implementer status
+
+- **DONE** → proceed to review.
+- **DONE_WITH_CONCERNS** → read the concerns; correctness/scope concerns get addressed before review, observations are noted.
+- **NEEDS_CONTEXT** → provide the missing context, re-dispatch.
+- **BLOCKED** → context problem: add context, re-dispatch. Needs more reasoning: escalate **capability tier** within the current billing lane (still `included` by default). Task too large: split it. Plan wrong: escalate to the human. Never force the same model to retry unchanged; never flip to `metered` without an explicit user request.
+
+## Red flags — never
+
+- Start implementation on main/master without explicit user consent
+- Skip the review, proceed with unfixed spec gaps or Critical/Important issues, or accept "close enough" on spec compliance
+- Dispatch multiple implementer subagents in parallel (conflicts)
+- Make a subagent read the plan file, skip scene-setting context, or ignore its questions
+- Let implementer self-review replace the reviewer (both are needed)
+- Fix a failed task manually in coordinator context — dispatch a fix subagent with specific instructions
+
+## Integration
+
+- **Plan source:** OpenSpec (`openspec/changes/<name>/tasks.md`)
+- **Subagents must follow:** [references/tdd-core.md](../../references/tdd-core.md) (condensed TDD rules; full skill at [skills/test-driven-development](../test-driven-development/SKILL.md) when stuck)
+- **On blockers:** [skills/systematic-debugging](../systematic-debugging/SKILL.md) before guessing fixes
+- **After all tasks:** [phases/verify.md](../../phases/verify.md) → [phases/review.md](../../phases/review.md) → [phases/finish.md](../../phases/finish.md)

package/vendor/skills/forge/subagents/final-reviewer-prompt.md

@@ -1,56 +1,56 @@
-# Final implementation reviewer
-
-Review the **entire** Forge session implementation against the plan **and**
-the capability specs. See [references/runtime-integrity.md](../references/runtime-integrity.md).
-
-## Plan source
-
-{PLAN_REFERENCE}
-
-## Capability specs (source of truth)
-
-{CAPABILITY_SPECS_REFERENCE}
-
-## Tasks completed
-
-{TASK_SUMMARY}
-
-## Scope
-
-All files changed in this session (use git diff or explicit list):
-
-{CHANGED_FILES}
-
-## Spec-to-runtime trace (required)
-
-For **each** requirement in the change's capability specs, name the
-**production caller** (worker job kind, HTTP endpoint, CLI command, scheduled
-job, …) that invokes the implementing code. Cross-check against `spine.json`
-(`forge spine check`) — every capability row must be wired, not library-only.
-
-- Library-only / no production caller → **`NOT READY`**
-- Stub handler / false success / enqueueable-but-unhandled kind → **`NOT READY`**
-- Job kind on the product surface that is neither wired end-to-end nor deleted → **`NOT READY`**
-- UI/API reads a collection or artifact nothing in the production path writes → **`NOT READY`**
-- Missing E2E fixture path with no explicit `BLOCKED` in `verify-evidence.md` → **`NOT READY`**
-
-## Product-loop acceptance (required — executed, not described)
-
-`forge e2e check` must be green: `e2e.json` steps drive the **closed loop**
-(produce → consume → decision changes output) and `e2e-results.json` records a
-green, current run (steps hash matches). A single job slice (e.g. ingest→file)
-or a library-level E2E does **not** count as platform E2E. Read the steps —
-would they pass against a stubbed handler? If yes, they prove nothing.
-
-- No green, current e2e run and no `BLOCKED` in `verify-evidence.md` → **`NOT READY`**
-- E2E steps assert no domain side effects (would pass on a stub) → **`NOT READY`**
-- `e2e.json` `notApplicable` without a reason no command could overcome → **`NOT READY`**
-- `BLOCKED` present → **`NOT READY`** (honest, but not READY)
-- Unresolved deferrals in `forge defer list` → **`NOT READY`**
-
-## Verdict
-
-- **READY** — every capability has a runtime owner, product loop evidenced, tests evidence real outcomes, no critical gaps
-- **NOT READY** — list blockers (prefer runtime-integrity and missing wiring first)
-
-Do not approve if tests were not run, **`test-evidence.md` is missing**, tasks remain unchecked, or any claimed capability is library-only / stubbed / false-succeeding. Task checkboxes at 100% do **not** override a broken spine.
+# Final implementation reviewer
+
+Review the **entire** Forge session implementation against the plan **and**
+the capability specs. See [references/runtime-integrity.md](../references/runtime-integrity.md).
+
+## Plan source
+
+{PLAN_REFERENCE}
+
+## Capability specs (source of truth)
+
+{CAPABILITY_SPECS_REFERENCE}
+
+## Tasks completed
+
+{TASK_SUMMARY}
+
+## Scope
+
+All files changed in this session (use git diff or explicit list):
+
+{CHANGED_FILES}
+
+## Spec-to-runtime trace (required)
+
+For **each** requirement in the change's capability specs, name the
+**production caller** (worker job kind, HTTP endpoint, CLI command, scheduled
+job, …) that invokes the implementing code. Cross-check against `spine.json`
+(`forge spine check`) — every capability row must be wired, not library-only.
+
+- Library-only / no production caller → **`NOT READY`**
+- Stub handler / false success / enqueueable-but-unhandled kind → **`NOT READY`**
+- Job kind on the product surface that is neither wired end-to-end nor deleted → **`NOT READY`**
+- UI/API reads a collection or artifact nothing in the production path writes → **`NOT READY`**
+- Missing E2E fixture path with no explicit `BLOCKED` in `verify-evidence.md` → **`NOT READY`**
+
+## Product-loop acceptance (required — executed, not described)
+
+`forge e2e check` must be green: `e2e.json` steps drive the **closed loop**
+(produce → consume → decision changes output) and `e2e-results.json` records a
+green, current run (steps hash matches). A single job slice (e.g. ingest→file)
+or a library-level E2E does **not** count as platform E2E. Read the steps —
+would they pass against a stubbed handler? If yes, they prove nothing.
+
+- No green, current e2e run and no `BLOCKED` in `verify-evidence.md` → **`NOT READY`**
+- E2E steps assert no domain side effects (would pass on a stub) → **`NOT READY`**
+- `e2e.json` `notApplicable` without a reason no command could overcome → **`NOT READY`**
+- `BLOCKED` present → **`NOT READY`** (honest, but not READY)
+- Unresolved deferrals in `forge defer list` → **`NOT READY`**
+
+## Verdict
+
+- **READY** — every capability has a runtime owner, product loop evidenced, tests evidence real outcomes, no critical gaps
+- **NOT READY** — list blockers (prefer runtime-integrity and missing wiring first)
+
+Do not approve if tests were not run, **`test-evidence.md` is missing**, tasks remain unchecked, or any claimed capability is library-only / stubbed / false-succeeding. Task checkboxes at 100% do **not** override a broken spine.

package/vendor/skills/forge/subagents/implementer-prompt.md

@@ -1,38 +1,38 @@
-# Implementer subagent brief
-
-You are the **implementer** for one Forge task. You receive only this brief — no chat history.
-
-## Task
-
-{TASK_TITLE}
-
-{TASK_BODY}
-
-## Files
-
-{FILE_LIST}
-
-## Requirements / spec excerpt
-
-{SPEC_EXCERPT}
-
-## Forge constraints
-
-- Read and follow [references/tdd-core.md](./references/tdd-core.md) before any production code (full TDD skill only if stuck).
-- **No** `git commit` or `git push` unless the user explicitly asked in the current message.
-- **Testing tiers:** tier 1 = scoped test file/pattern per red/green cycle; tier 2 = narrowest command proving this task. **Do not** run the full workspace suite unless this task touches shared contracts or the brief says so — that is tier 3 and runs once at verify.
-- Include **command, exit code, and pass/fail summary** for tier 2 in your report (coordinator saves to `test-evidence.md`).
-- Minimal diff; match existing style; trace ecosystem consumers if contracts change.
-
-## Runtime integrity (hard)
-
-Read [references/runtime-integrity.md](../references/runtime-integrity.md). Honor these without exception:
-
-- **No stub-with-success.** Do not implement a handler that only logs / bumps progress / marks succeeded. Unwired kinds must fail closed or not be exposed.
-- **Name the runtime caller.** If you add library code, report which production path (job kind, endpoint, CLI) invokes it. If nothing calls it yet, end with `DONE_WITH_CONCERNS` and say so — do not pretend the capability is done.
-- **Tests must fail on a no-op.** Assert domain side effects, not ceremony.
-- If this brief tells you a stub is OK or to “wire later,” **reject the brief**: reply `NEEDS_CONTEXT` / `BLOCKED` and ask the coordinator to restore full scope or get user approval to shrink it.
-
-## Report status
-
-End with one of: `DONE`, `DONE_WITH_CONCERNS`, `NEEDS_CONTEXT`, `BLOCKED` plus details.
+# Implementer subagent brief
+
+You are the **implementer** for one Forge task. You receive only this brief — no chat history.
+
+## Task
+
+{TASK_TITLE}
+
+{TASK_BODY}
+
+## Files
+
+{FILE_LIST}
+
+## Requirements / spec excerpt
+
+{SPEC_EXCERPT}
+
+## Forge constraints
+
+- Read and follow [references/tdd-core.md](./references/tdd-core.md) before any production code (full TDD skill only if stuck).
+- **No** `git commit` or `git push` unless the user explicitly asked in the current message.
+- **Testing tiers:** tier 1 = scoped test file/pattern per red/green cycle; tier 2 = narrowest command proving this task. **Do not** run the full workspace suite unless this task touches shared contracts or the brief says so — that is tier 3 and runs once at verify.
+- Include **command, exit code, and pass/fail summary** for tier 2 in your report (coordinator saves to `test-evidence.md`).
+- Minimal diff; match existing style; trace ecosystem consumers if contracts change.
+
+## Runtime integrity (hard)
+
+Read [references/runtime-integrity.md](../references/runtime-integrity.md). Honor these without exception:
+
+- **No stub-with-success.** Do not implement a handler that only logs / bumps progress / marks succeeded. Unwired kinds must fail closed or not be exposed.
+- **Name the runtime caller.** If you add library code, report which production path (job kind, endpoint, CLI) invokes it. If nothing calls it yet, end with `DONE_WITH_CONCERNS` and say so — do not pretend the capability is done.
+- **Tests must fail on a no-op.** Assert domain side effects, not ceremony.
+- If this brief tells you a stub is OK or to “wire later,” **reject the brief**: reply `NEEDS_CONTEXT` / `BLOCKED` and ask the coordinator to restore full scope or get user approval to shrink it.
+
+## Report status
+
+End with one of: `DONE`, `DONE_WITH_CONCERNS`, `NEEDS_CONTEXT`, `BLOCKED` plus details.