cc-codeconductor 1.1.0 → 1.3.0

package/presets/opencode/README.md

@@ -17,30 +17,33 @@ providers)** models.
 | `claude-sonnet-4-6`         | Balanced, general purpose       | Default for most agents         |
 | `claude-haiku-4-5-20251001` | Fast, lightweight               | Task Coach, Docs, Repo Explorer |
 
-### OpenCode Go Models (Chinese Providers)
-
-| Model             | Provider        | Strength                         | Best For                   |
-| ----------------- | --------------- | -------------------------------- | -------------------------- |
-| `deepseek-v4-pro` | DeepSeek        | Reasoning complejo, arquitectura | Architect, orchestrator    |
-| `mimo-v2.5-pro`   | MiniMax         | Razonamiento matemático, código  | Implementer                |
-| `minimax-m2.7`    | MiniMax         | Tareas equilibradas              | Tester, orchestrator       |
-| `qwen3.6-plus`    | Qwen            | Rápido, eficiente                | Reviewer, Task Coach, Docs |
-| `kimi-k2-6`       | Kimi (Moonshot) | Conversación natural             | Task Coach, Docs           |
+### OpenCode Go models
+
+Canonical slugs live in `src/presets/models/*.yml`. Production roles never use
+`muse-spark-1.2-contributor` (trains on prompts). `ox-alpha-free` is an optional
+fallback, not a default.
+
+| Model | Best for |
+| --- | --- |
+| `deepseek-v4-pro` | Architect |
+| `kimi-k3` | Security reviewer |
+| `qwen3.8-max` | Reviewer, contract-builder |
+| `glm-5.3` | Devil, complexity-auditor |
+| `mimo-v2.5` | Implementer |
+| `minimax-m3` | Tester |
+| `qwen3.7-plus` | Orchestrator (TUI default) |
+| `gpt-5.6-luna` | Task coach, planner |
+| `deepseek-v4-flash` | Goal planner |
+| `longcat-2.0` | Repo explorer |
+| `hy3` | Docs |
 
 ---
 
 ## Agent Model Matrix
 
-| Agent             | Claude (Default)            | OpenCode Go (Recommended) | Alternative       |
-| ----------------- | --------------------------- | ------------------------- | ----------------- |
-| **Orchestrator**  | `claude-sonnet-4-6`         | `deepseek-v4-pro`         | `minimax-m2.7`    |
-| **Architect**     | `claude-opus-4-7`           | `deepseek-v4-pro`         | `mimo-v2.5-pro`   |
-| **Implementer**   | `claude-sonnet-4-6`         | `mimo-v2.5-pro`           | `minimax-m2.7`    |
-| **Tester**        | `claude-sonnet-4-6`         | `minimax-m2.7`            | `deepseek-v4-pro` |
-| **Reviewer**      | `claude-sonnet-4-6`         | `qwen3.6-plus`            | `minimax-m2.7`    |
-| **Task Coach**    | `claude-haiku-4-5-20251001` | `qwen3.6-plus`            | `kimi-k2.6`       |
-| **Docs**          | `claude-haiku-4-5-20251001` | `qwen3.6-plus`            | `kimi-k2.6`       |
-| **Repo Explorer** | `claude-haiku-4-5-20251001` | `qwen3.6-plus`            | `kimi-k2.6`       |
+See `skills/cc-update-preset-models/references/role-map.md`. OpenCode column
+examples: architect `deepseek-v4-pro`, implementer `mimo-v2.5`, tester
+`minimax-m3`, reviewer `qwen3.8-max`.
 
 ---
 
@@ -93,7 +96,7 @@ The following paths are denied by default:
 
 ### Medium Tasks (Implementation, testing)
 
-**Recommended:** `mimo-v2.5-pro` or `minimax-m2.7` (OpenCode Go) or
+**Recommended:** `mimo-v2.5` or `minimax-m3` (OpenCode Go) or
 `claude-sonnet-4-6` (Claude)
 
 - Implementer code writing
@@ -165,7 +168,7 @@ export KIMI_API_KEY="your-key"
 | ------------------------------- | -------------------------------------------------- |
 | Complex reasoning, architecture | OpenCode Go (`deepseek-v4-pro`) or Claude (`opus`) |
 | Fast iteration, simple tasks    | OpenCode Go (`qwen3.6-plus`) or Claude (`haiku`)   |
-| Code implementation             | OpenCode Go (`mimo-v2.5-pro`) or Claude (`sonnet`) |
+| Code implementation             | OpenCode Go (`mimo-v2.5`) or Claude (`sonnet`) |
 | Budget constraints              | OpenCode Go (generally lower cost)                 |
 | Availability issues             | Switch to alternative from the matrix              |
 

package/presets/opencode/agents/architect.md

@@ -246,3 +246,9 @@ Rules under CCEP-1:
 - Never approve your own plan — the human approves before implementation starts.
 - If scope expands during design, flag it as a separate task, not an extension
   of the current one.
+
+## Composition
+
+Invoke via `/cc-feature`, `/cc-openspec`, or the orchestrator. Do not invoke
+other agents. If an OpenSpec change folder is active, invoke skill `openspec`
+and run `openspec analyze --output json` before implementation is allowed.

package/presets/opencode/agents/docs.md

@@ -77,7 +77,7 @@ CHANGELOG is mandatory for every implementation change. No exceptions.
 - `docs/**/*.md` — any markdown documentation file
 - `docs/adr/*.md` — Architecture Decision Records
 - `CHANGELOG.md` — always update for any implementation change
-- `.codeconductor/sessions/handoff.md` — `/cc:handoff` only; gitignored; redact secrets
+- `.codeconductor/sessions/handoff.md` — `/cc-handoff` only; gitignored; redact secrets
 - `openapi.yaml`, `openapi.json`, or any OpenAPI spec file
 - Any `*-api.yaml` or `*-api.json` file
 

package/presets/opencode/agents/implementer.md

@@ -215,3 +215,10 @@ tests. On **REJECT**, stop and escalate to the orchestrator.
 - Never modify the Technical Plan — if the plan is wrong, escalate to
   `architect` via the orchestrator.
 - Never commit without human confirmation.
+
+## Composition
+
+Invoke via `/cc-feature`, `/cc-fix`, `/cc-tdd-cycle`, or the orchestrator.
+Do not invoke other agents. Invoke skill `testing-tdd`.
+Deliverable is incomplete without a passing suite and runner TDD evidence when
+TDD is required. Scorecard: `bun run dev scorecard create --from-diff`.

package/presets/opencode/agents/reviewer.md

@@ -260,7 +260,7 @@ When the orchestrator requests evaluation, produce scores for all 8 criteria in
 2. Minimal diff (20%)
 3. Tests present and passing (15%)
 4. No regressions (15%)
-5. Code conventions (10%)
+5. Code conventions (5%)
 6. Documentation updated (5%)
 7. Context discipline (5%)
 8. Complexity diffusion / cc-gain (5%)
@@ -282,3 +282,9 @@ Invoke skill `evaluation` and run `scorecard record` with agent `reviewer`, mode
 - Never issue vague findings ("this could be better") — every finding must name
   the exact location and the specific required action.
 - Never run `git push` or `git commit`.
+
+## Composition
+
+Invoke via `/cc-review` or the orchestrator. Do not invoke other agents.
+Invoke skill `evaluation`. Deliverable is incomplete without
+`scorecard create --from-diff` and a recorded PASS|REVISE|REJECT.

package/presets/opencode/agents/tester.md

@@ -285,3 +285,9 @@ and confirmed failing; list the failing test paths in `artifacts` so the
 - Never mock real behavior that could be tested with an in-memory alternative.
 - Never declare coverage complete when any acceptance criterion lacks a test.
 - Never run `git push` or `git commit`.
+
+## Composition
+
+Invoke via `/cc-tdd-cycle`, `/cc-feature`, or the orchestrator (test before
+implement). Do not invoke other agents. Invoke skill `testing-tdd`.
+Deliverable is incomplete without `captureTddSuiteEvidence` when TDD is required.

package/presets/opencode/commands/cc-api-contract.md

@@ -20,6 +20,20 @@ Command: `api-contract` (fixed for this workflow — do not infer from user text
 
 ---
 
+## Step 0b — OpenSpec quality gates
+
+If `openspec status` reports an active change folder:
+
+1. Run: `npx cc-codeconductor openspec validate --output json`
+2. Run: `npx cc-codeconductor openspec analyze --output json`
+3. If analyze `stop` is true or any finding is CRITICAL, stop. Do not delegate to implementer.
+4. Next command spelling on this runner: `/cc-api-contract`
+
+Local development: `bun run dev <same argv>`. Published package: `npx cc-codeconductor`.
+
+---
+
+
 ## Step 1 — Task Card validation (task-coach)
 
 Invoke `task-coach` with the request above.
@@ -80,3 +94,7 @@ or docs/OpenAPI drift.
 
 Report the final Task Card, Technical Plan, implementation summary, test report,
 review report, docs updated, compatibility impact, and residual risks.
+
+## Next
+
+Run `/cc-review` on the diff before merging.

package/presets/opencode/commands/cc-ask.md

@@ -1,6 +1,6 @@
 ---
 description: >-
-  [cc: alias] Recommend one CodeConductor slash command from a natural-language
+  Recommend one CodeConductor slash command from a natural-language
   problem. Does not start that workflow.
 ---
 
@@ -9,7 +9,7 @@ description: >-
 Problem: $ARGUMENTS
 
 This is a **router**, not a delivery workflow. Recommend exactly one command.
-Do **not** run the recommended `/cc:` command unless the human confirms.
+Do **not** run the recommended `/cc-` command unless the human confirms.
 
 ---
 
@@ -31,13 +31,13 @@ The catalog is only:
 
 | Slash | When |
 | ----- | ---- |
-| `/cc:feature` | new behavior |
-| `/cc:fix` | bug / regression |
-| `/cc:refactor` | structure, no behavior change |
-| `/cc:review` | inspect a diff/PR |
-| `/cc:tdd-cycle` | red → green → refactor |
-| `/cc:backlog` | create or append BACKLOG.md / objectives |
-| `/cc:openspec` | BACKLOG / BC-xxx delivery |
+| `/cc-feature` | new behavior |
+| `/cc-fix` | bug / regression |
+| `/cc-refactor` | structure, no behavior change |
+| `/cc-review` | inspect a diff/PR |
+| `/cc-tdd-cycle` | red → green → refactor |
+| `/cc-backlog` | create or append BACKLOG.md / objectives |
+| `/cc-openspec` | BACKLOG / BC-xxx delivery |
 
 If the CLI is unavailable, apply the same catalog and the same priority:
 backlog (author) vs openspec (deliver) → tdd-cycle → review → refactor → fix → feature (default).
@@ -53,4 +53,4 @@ Print:
 
 **STOP.** Wait for the human. Do not invoke the recommended workflow.
 
-For a full preset inventory (skills, commands on disk), use `/cc:help` / `cc-help`.
+For a full preset inventory (skills, commands on disk), use `/cc-help` / `cc-help`.

package/presets/opencode/commands/cc-council.md

@@ -85,3 +85,7 @@ If APPROVED (no CRITICAL findings):
 ## Completion
 
 Deliver the complete Council Verdict. The feature is only complete when tests pass and the council explicitly approves the implementation according to the specification.
+
+## Next
+
+Approved: merge. Blocked: return the findings to `/cc-fix` or `/cc-feature`.

package/presets/opencode/commands/cc-db-migration.md

@@ -21,6 +21,20 @@ Command: `db-migration` (fixed for this workflow — do not infer from user text
 
 ---
 
+## Step 0b — OpenSpec quality gates
+
+If `openspec status` reports an active change folder:
+
+1. Run: `npx cc-codeconductor openspec validate --output json`
+2. Run: `npx cc-codeconductor openspec analyze --output json`
+3. If analyze `stop` is true or any finding is CRITICAL, stop. Do not delegate to implementer.
+4. Next command spelling on this runner: `/cc-db-migration`
+
+Local development: `bun run dev <same argv>`. Published package: `npx cc-codeconductor`.
+
+---
+
+
 ## Step 1 — Task Card validation (task-coach)
 
 Invoke `task-coach` with the request above.
@@ -80,3 +94,7 @@ undocumented deployment sequencing, or model/migration drift.
 
 Report migration files changed, model files changed, tests run, lock risk, data
 risk, operational sequencing, rollback/forward-fix notes, and residual risk.
+
+## Next
+
+Run `/cc-review` on the diff before merging.

package/presets/opencode/commands/cc-explore.md

@@ -26,10 +26,10 @@ If `graphify-out/graph.json` exists, run `graphify query "$ARGUMENTS"`. Invoke `
 
 ## Step 2 — Suggest next command (orchestrator)
 
-Recommend exactly one next slash command (`/cc:feature`, `/cc:fix`, `/cc:refactor`, `/cc:review`, `/cc:backlog`, `/cc:openspec`, `/cc:triage`, …) with a one-sentence rationale. Do not start that workflow.
+Recommend exactly one next slash command (`/cc-feature`, `/cc-fix`, `/cc-refactor`, `/cc-review`, `/cc-backlog`, `/cc-openspec`, `/cc-triage`, …) with a one-sentence rationale. Do not start that workflow.
 
 ---
 
 ## Completion
 
-Report the Repo Map and the recommended next `/cc:` command.
+Report the Repo Map and the recommended next `/cc-` command.

package/presets/opencode/commands/cc-feature.md

@@ -21,6 +21,24 @@ Command: `feature` (fixed for this workflow — do not infer from user text)
 
 ---
 
+## Step 0b — OpenSpec quality gates
+
+If `openspec status` reports an active change folder:
+
+1. Run: `npx cc-codeconductor openspec validate --output json`
+2. Run: `npx cc-codeconductor openspec analyze --output json`
+3. If analyze `stop` is true or any finding is CRITICAL, stop. Do not delegate to implementer.
+4. Next command spelling on this runner: `/cc-feature`
+
+Local development: `bun run dev <same argv>`. Published package: `npx cc-codeconductor`.
+
+Skills: `using-cc-skills`, `openspec`, `testing-tdd`, `evaluation`.
+Do not skip `openspec analyze` when a change folder is active.
+"I'll add tests later" is not allowed — tester before implementer.
+
+---
+
+
 ## Step 1 — Wayfinding (repo-explorer)
 
 If `graphify-out/graph.json` exists, run `graphify query "$ARGUMENTS"` (and
@@ -134,3 +152,7 @@ Report the following to the human:
 
 The feature is complete only when: all tests pass, no CRITICAL review findings
 remain, and documentation reflects the implemented behavior.
+
+## Next
+
+Run `/cc-review` on the diff before merging.

package/presets/opencode/commands/cc-fix.md

@@ -31,6 +31,20 @@ Command: `fix` (fixed for this workflow — do not infer from user text)
 
 ---
 
+## Step 0b — OpenSpec quality gates
+
+If `openspec status` reports an active change folder:
+
+1. Run: `npx cc-codeconductor openspec validate --output json`
+2. Run: `npx cc-codeconductor openspec analyze --output json`
+3. If analyze `stop` is true or any finding is CRITICAL, stop. Do not delegate to implementer.
+4. Next command spelling on this runner: `/cc-fix`
+
+Local development: `bun run dev <same argv>`. Published package: `npx cc-codeconductor`.
+
+---
+
+
 ## Step 1 — Wayfinding (repo-explorer)
 
 If `graphify-out/graph.json` exists, run `graphify query "$ARGUMENTS"` (and
@@ -148,3 +162,10 @@ resolved.
 Report: Task Card, Implementation Summary, regression test added, Review Report
 (if applicable). The fix is complete only when: the regression test passes, the
 full suite passes, and no CRITICAL review findings remain.
+
+Skills: `testing-tdd`, `evaluation`. Record `scorecard create --from-diff`.
+A small fix still needs a Task Card and a failing regression test first.
+
+## Next
+
+Run `/cc-review` on the diff before merging.

package/presets/opencode/commands/cc-handoff.md

@@ -29,7 +29,7 @@ errors instead of pasting dumps. If unresolved credential material remains, stop
 and wait for a human (CCEP `stopOnHighRisk`).
 
 Include: goal, Task Card status (no secret fields), files touched, test
-pass/fail (not log dumps), open questions, and the next `/cc:` command.
+pass/fail (not log dumps), open questions, and the next `/cc-` command.
 
 Do not edit source or tests.
 

package/presets/opencode/commands/cc-iterative.md

@@ -21,6 +21,20 @@ Command: `iterative` (fixed for this workflow — do not infer from user text)
 
 ---
 
+## Step 0b — OpenSpec quality gates
+
+If `openspec status` reports an active change folder:
+
+1. Run: `npx cc-codeconductor openspec validate --output json`
+2. Run: `npx cc-codeconductor openspec analyze --output json`
+3. If analyze `stop` is true or any finding is CRITICAL, stop. Do not delegate to implementer.
+4. Next command spelling on this runner: `/cc-iterative`
+
+Local development: `bun run dev <same argv>`. Published package: `npx cc-codeconductor`.
+
+---
+
+
 ## Step 1 — Wayfinding (repo-explorer)
 
 If `graphify-out/graph.json` exists, run `graphify query "$ARGUMENTS"` (and

package/presets/opencode/commands/cc-openspec.md

@@ -29,6 +29,20 @@ Command: `openspec` (fixed for this workflow — do not infer from user text)
 
 ---
 
+## Step 0b — OpenSpec quality gates
+
+If `openspec status` reports an active change folder:
+
+1. Run: `npx cc-codeconductor openspec validate --output json`
+2. Run: `npx cc-codeconductor openspec analyze --output json`
+3. If analyze `stop` is true or any finding is CRITICAL, stop. Do not delegate to implementer.
+4. Next command spelling on this runner: `/cc-openspec`
+
+Local development: `bun run dev <same argv>`. Published package: `npx cc-codeconductor`.
+
+---
+
+
 ## Step 1 — Scan
 
 Run `npx cc-codeconductor openspec scan`. Report new, modified, and closed items.

package/presets/opencode/commands/cc-prototype.md

@@ -34,4 +34,4 @@ Invoke `implementer` in a Git worktree that is not a merge candidate. Label the
 
 ## Completion
 
-Report bounds, worktree path, and what was learned. Recommend `/cc:feature` or `/cc:fix` if the spike should become real work.
+Report bounds, worktree path, and what was learned. Recommend `/cc-feature` or `/cc-fix` if the spike should become real work.

package/presets/opencode/commands/cc-refactor.md

@@ -28,7 +28,7 @@ behavioral consequences.
 If coverage is insufficient:
 
 1. **STOP**. Report the coverage gap to the human.
-2. Suggest invoking `/cc:test-plan` first to establish coverage.
+2. Suggest invoking `/cc-test-plan` first to establish coverage.
 3. Do not proceed with the refactor until coverage is confirmed.
 
 ---
@@ -158,3 +158,7 @@ and after, Review Report (if applicable).
 The refactor is complete only when: all pre-existing tests still pass, the
 implementation matches the approved plan exactly, and no CRITICAL review
 findings remain.
+
+## Next
+
+Run `/cc-review` on the diff before merging.

package/presets/opencode/commands/cc-review.md

@@ -137,7 +137,7 @@ If any CRITICAL findings exist:
 - The Review Report status is **BLOCKED**
 - Report all CRITICAL findings to the human
 - Do not proceed until each CRITICAL finding is resolved
-- After resolution, invoke `/cc:review` again on the same target
+- After resolution, invoke `/cc-review` again on the same target
 
 If no CRITICAL findings exist:
 
@@ -151,3 +151,7 @@ If no CRITICAL findings exist:
 
 Deliver the complete Review Report. Never summarize or omit findings. Every
 finding must include a location and an actionable description.
+
+## Next
+
+Approved: merge. Blocked: return the findings to `/cc-fix` or `/cc-feature`.

package/presets/opencode/commands/cc-scorecard.md

@@ -19,6 +19,8 @@ Command: `scorecard` (fixed for this workflow — do not infer from user text)
 
 ---
 
+If a change folder exists, run `npx cc-codeconductor openspec analyze --output json` first. `--from-diff` overlays FR/SC coverage onto `acceptance` and TDD evidence onto `tests`.
+
 1. `scorecard create --task <id> --from-diff`
 2. Complete criteria (reviewer or human)
 3. `scorecard regression` (optional)

package/presets/opencode/commands/cc-security.md

@@ -177,3 +177,7 @@ resolved.
 Report: Task Card, Implementation Summary, regression test added, Review Report
 (if applicable). The task is complete only when: the regression test passes, the
 full suite passes, and no CRITICAL review findings remain.
+
+## Next
+
+Run `/cc-review` on the diff before merging.

package/presets/opencode/commands/cc-spec-mutation.md

@@ -0,0 +1,194 @@
+---
+description: >-
+  Spec-locked TDD with a mutation-testing gate — refine the intent
+  into an immutable Gherkin contract (SHA-256 frozen), implement under the
+  three laws of TDD, pass a judge audit, and merge only if every mutant dies.
+---
+
+# Spec-Mutation — Hard Spec → TDD → Judge → Mutation Gate
+
+Scope: $ARGUMENTS
+
+Describe what behavior you want to implement. Include:
+
+- The function, method, or feature to implement
+- The expected behavior (inputs, outputs, invariants, edge cases)
+- The allowed file scope (production files that may change)
+- The test command for the affected suite (e.g. `pytest tests/test_billing.py`)
+
+---
+
+## Step 0 — CCEP Bootstrap
+
+Command: `spec-mutation` (fixed for this workflow — do not infer from user text)
+
+1. Run: `npx cc-codeconductor ccep parse --command spec-mutation "$ARGUMENTS" --output json`
+2. Run: `npx cc-codeconductor ccep resolve --command spec-mutation "$ARGUMENTS" --output json`
+3. Run: `npx cc-codeconductor ccep profile spec-mutation --output json`
+4. After planner/intake JSON is available, run: `npx cc-codeconductor ccep evaluate --command spec-mutation --input <planner.json> --output json`. If `stop` is true, show questions or risks and wait for human input.
+5. Delegate to subagents using compiled CCEP prompts — never forward raw `$ARGUMENTS` to planners.
+   Canonical delivery order is test-before-implement whenever both phases apply.
+
+---
+
+## Step 0b — OpenSpec quality gates
+
+If `openspec status` reports an active change folder:
+
+1. Run: `npx cc-codeconductor openspec validate --output json`
+2. Run: `npx cc-codeconductor openspec analyze --output json`
+3. If analyze `stop` is true or any finding is CRITICAL, stop. Do not delegate to implementer.
+4. Next command spelling on this runner: `/cc-spec-mutation`
+
+Local development: `bun run dev <same argv>`. Published package: `npx cc-codeconductor`.
+
+---
+
+## Contract
+
+The Gherkin specification is the **immutable contract** of the system. No code
+merges unless it survives intentional source mutations. The loop is closed:
+
+```
+[Human + spec_partner] ──> [gherkin_author] ──> [Test Freeze: SHA-256]
+                                                       │
+┌──────────────────────────────────────────────────────┘
+▼
+[tdd_craftsman] <───────────────┐ (surviving mutant)
+   (Red-Green-Refactor)         │
+        │                       │
+        ▼                       │
+     [judge] ───────────> [mutation_testing] ───> [Safe Merge]
+```
+
+Role mapping onto Conductor Agents (AGENTS.md):
+
+| Workflow role      | Conductor Agent   | Deliverable                                |
+| ------------------ | ----------------- | ------------------------------------------ |
+| `craftsman_lead`   | `orchestrator`    | Routed Task Cards, per-stage scorecards    |
+| `spec_partner`     | `task-coach`      | Spec draft with invariants and boundaries  |
+| `gherkin_author`   | `contract-builder`| Strict `.feature` file (Given/When/Then)   |
+| `tdd_craftsman`    | `tester` → `implementer` | Failing test, then minimal production code |
+| `judge`            | `reviewer`        | Binary verdict (PASS/REJECT) with scoring  |
+| `mutation_testing` | `tester` (runner) | Killed/survived mutant report              |
+
+## Stage 1 — Interactive refinement (`spec_partner` / task-coach)
+
+Do not jump to implementation. Apply Socratic questioning to the initial intent:
+
+- Preconditions, postconditions, and edge cases.
+- Invariant matrix: what must always be true.
+- Scope boundaries: explicit `Scope / Files` and `Scope / Out` for the Task Card.
+
+Stop gate: human confirms the draft before formalization.
+
+## Stage 2 — Hard Spec formalization (`gherkin_author` / contract-builder)
+
+Formalize the agreed draft into `specs/<task>.feature` using strict Gherkin:
+
+- No vague language ("must respond fast" is forbidden — use measurable Then steps).
+- Every scenario declares preconditions (`Given`), actions (`When`), and
+  observable states (`Then`).
+- Once the human approves the `.feature`, freeze it:
+
+```bash
+shasum -a 256 specs/<task>.feature tests/ > .codeconductor/tasks/<task_id>.lock
+```
+
+From this point `specs/` and `tests/` are **read-only** for implementation
+agents. Before every later gate, recompute the hash; any single-byte difference
+aborts the pipeline with scorecard 0 (Specification Gaming).
+
+## Stage 3 — TDD under the three laws (`tdd_craftsman`)
+
+Delegates to the `/cc-tdd-cycle` state machine (`tddCycleStateMachine` in
+`domain/loop`). Evidence must be captured with `captureTddSuiteEvidence` — do
+not hand-edit JSON under `.codeconductor/evidence/`.
+
+1. **Law 1 (RED):** no production code except to make a failing test pass. A
+   compile error from a missing interface counts as a failure.
+2. **Law 2:** write exactly one failing assertion or scenario at a time.
+3. **Law 3 (GREEN):** write only the minimal production code to pass. No
+   speculative code, no preventive heuristics, stdlib-first.
+
+Hard rule: any write attempt against `specs/` or `tests/` during GREEN/REFACTOR
+is a harness violation — stop execution and report.
+
+## Stage 4 — Judge audit (`judge` / reviewer)
+
+Deterministic gates before spending compute on mutation:
+
+- Clean compile / diagnostics exit code 0.
+- Traceability: every Gherkin step maps to an implemented test step.
+- Scope Gaming audit: `git diff --name-only` must match the Task Card
+  `Scope / Files` exactly. Relaxed types, weakened assertions, or out-of-scope
+  edits → REJECT with findings.
+
+Verdict is binary: PASS continues to the mutation gate; REJECT returns to the
+`implement` phase with the findings attached.
+
+## Stage 5 — Mutation gate (`mutation_testing`)
+
+Run the deterministic AST mutator shipped with this preset:
+
+```bash
+python3 presets/shared/mutation_runner.py \
+  --target <production_file.py> \
+  --test-command "<test command>" \
+  --spec-folder specs
+```
+
+The runner applies deterministic operator mutations (`>` → `<=`, `==` → `!=`,
+`is` → `is not`, …) one at a time, re-runs the suite per mutant, and restores
+the original source unconditionally (`finally` rollback).
+
+- **Mutant killed (tests fail):** the suite detects the corruption. Continue.
+- **Mutant survived (tests pass):** the tests are blind to this branch. The
+  runner writes `specs/handover.md` + appends to
+  `specs/implementation-summary.md` and exits with code **2**.
+
+Non-Python stacks: substitute Stryker (JS/TS), PITest (JVM), or Mutmut
+(Python full-suite) with the same contract — 100% kill rate or hands-off.
+
+### Hands-off protocol (exit code 2)
+
+1. Do NOT modify production code to "fix" a surviving mutant.
+2. Route back to `tdd_craftsman` with `specs/handover.md` as input: write the
+   missing failing test (Law 1 & 2) that asserts the mutated branch.
+3. Re-run stages 3–5.
+
+### Circuit breaker (max 3 loops)
+
+The orchestrator keeps a persistent counter per Task Card. If the
+`tdd_craftsman ↔ mutation_testing` loop does not reach a 100% kill rate after
+**3 iterations**:
+
+- Cancel active subagents (stop token/context spend).
+- `git checkout -- <scope>` rollback to the last clean state.
+- Scorecard: `STATUS = BLOCKED`; escalate to a human operator. The branch stays
+  frozen until human arbitration.
+
+## Guardrails (harness-enforced, not prompt-enforced)
+
+- **Test Freezing:** SHA-256 of `specs/` + `tests/` stored in
+  `.codeconductor/tasks/<task_id>.lock`; hash mismatch aborts the pipeline.
+- **Scope Guardian:** `fs_write`/`fs_patch` paths are validated against the Task
+  Card `Scope / Files`; path escapes (`../`) and critical files (`**/.env*`,
+  `**/credentials*`, infrastructure roots) are denied per `policy.yml`.
+- **RBAC per role:** `gherkin_author` writes only `specs/`; `tdd_craftsman`
+  reads specs/tests and writes only scoped `src/`; `judge` and
+  `mutation_testing` are read-only except the runner's rolled-back patch.
+- **Worktree isolation:** run the whole flow in a dedicated `git worktree`;
+  protected branches (`main`, `master`, `develop`) are never touched.
+
+## Completion criteria
+
+- [ ] `.feature` approved and frozen (SHA-256 lock file exists and matches).
+- [ ] RED → GREEN → REFACTOR evidence captured per phase.
+- [ ] Judge verdict PASS (compile clean, traceability complete, scope clean).
+- [ ] Mutation runner exits 0 with `total_mutants_killed == total_points`.
+- [ ] Scorecard records the kill rate and iteration count (≤ 3).
+
+## Next
+
+Run `/cc-review` on the diff before merging.