nw-ab-critique-dimensi…
Review dimensions for validating agent quality - template compliance, safety, testing, and priority validation
Runs feature-scoped mutation testing to validate test suite quality. Use after implementation to verify tests catch real bugs (kill rate >= 80%).
$ npx -y skills add nWave-ai/nWave --skill nw-mutation-test --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/nw-mutation-testContext preview
The summary Claude sees to decide when to auto-load this skill.
Runs feature-scoped mutation testing to validate test suite quality. Use after implementation to verify tests catch real bugs (kill rate >= 80%).
name: nw-mutation-test description: "Runs feature-scoped mutation testing to validate test suite quality. Use after implementation to verify tests catch real bugs (kill rate >= 80%)." user-invocable: true argument-hint: '[feature-id] - Optional: --threshold=[75|80|85] --language=[auto|python|java|javascript]'
**Wave**: QUALITY_GATE **Agent**: Crafter (nw-software-crafter)
Run mutation testing against implementation files from the current feature. Extracts targets from the execution-log.json (see Target extraction)|generates feature-scoped configs|delegates to software-crafter. Uses cosmic-ray (Python)|PIT (Java)|Stryker (JS/TS/C#).
Projects declare a strategy via `## Mutation Testing Strategy` in `CLAUDE.md`: `per-feature` | `nightly-delta` | `pre-release` | `disabled`.
**Default (when unspecified): `nightly-delta`** — the recommended mode. CI runs mutmut nightly against modules changed since the last run (the delta), keeping per-feature delivery gates fast. `/nw-mutation-test` performs an explicit, on-demand feature-scoped run regardless of strategy; under `nightly-delta` the in-wave Phase 5 gate is skipped and the work is handled by the CI nightly pipeline.
Implementation files are selected from the `execution-log.json` (`completed_steps[].files_modified.implementation`).
Orchestrator performs before delegating:
1. **Extract files** — Read `execution-log.json` and extract implementation files from `completed_steps[].files_modified.implementation`. Gate: file list non-empty. 2. **Verify on disk** — Check all extracted files exist on disk. Gate: zero missing files. 3. **Detect language** — Scan config files (pyproject.toml, pom.xml, package.json, etc.) to select tool. Gate: language identified. 4. **Confirm tests pass** — Run `pytest -x {test_scope}` (or equivalent). Gate: exit code 0, no failures. 5. **Ensure mutation venv** — For Python, verify `.venv-mutation/` exists with cosmic-ray installed. Gate: `cosmic-ray --version` succeeds.
@nw-software-crafter
Execute mutation testing for project {feature-id}.
**Context to pass inline (agent has no Skill access):**
**Configuration:**
**Output file:** `docs/feature/{feature-id}/deliver/mutation/mutation-report.md`
/nw-mutation-test des-hook-enforcement tests/des/
Reads the execution-log.json, runs `generate_scoped_configs.py des-hook-enforcement`, delegates to software-crafter with per-component configs. Agent runs cosmic-ray, produces mutation-report.md.
/nw-mutation-test auth-upgrade tests/auth/
Extracts files manually from the execution-log.json, creates single cosmic-ray config with `module-path = [file1, file2, ...]` and `test-command = "pytest -x tests/auth/"`, delegates to agent.
/nw-mutation-test payment-gateway tests/payment/
Detects `package.json`, selects Stryker, delegates with Stryker-specific instructions.
After EVERY mutation run (success, failure, or interruption):
1. **Restore source files** — Run `git checkout -- src/ tests/`. Gate: working tree clean (no mutations remain). 2. **Verify no corruption** — Confirm test suite still passes after restore. Gate: `pytest -x {test_scope}` exits 0.
Mutation tools apply mutations directly to source files. An interrupted run can leave corrupted code (e.g. `is not None` -> `is None`). Agent MUST execute these steps even if the run errors out.
Kill rate thresholds:
1. **>= 80% PASS** — Proceed to next wave. 2. **70-80% WARN** — Review surviving mutants, document findings, proceed with caution. 3. **< 70% FAIL** — Add tests targeting surviving mutants before proceeding.
Skip conditions (each requires documented justification in mutation-report.md):
1. **No tool for language** — No mutation framework available for detected language. 2. **Project opt-out** — `.mutation-config.yaml` has `skip: true` with justification. 3. **Broken test suite** — Pre-invocation step 4 fails; fix tests before mutation testing.
Note: Python projects require mutation testing. All skips need documented justification.
**Handoff To**: Phase 8 - Finalize (orchestrator continues develop.md workflow) **Deliverables**: `docs/feature/{feature-id}/deliver/mutation/mutation-report.md`
docs/feature/{feature-id}/deliver/mutation/
mutation-report.md
cosmic-ray-*.toml (ephemeral)AI agents that guide you from idea to working code, with human judgment at every gate. nWave runs inside Claude Code. It breaks feature delivery into seven waves (discover, diverge, discuss, design, devops, distill, deliver).
Repo: nWave-ai/nWave
Review dimensions for validating agent quality - template compliance, safety, testing, and priority validation
Review dimensions for validating agent quality - template compliance, safety, testing, and priority validation
Review dimensions for acceptance test quality - happy path bias, GWT compliance, business language purity, coverage completeness, walking skeleton…
Detailed 5-phase workflow for creating agents - from requirements analysis through validation and iterative refinement
5-layer testing approach for agent validation including adversarial testing, security validation, and prompt injection resistance
Architectural style selection decision matrices, trade-off analysis, structural enforcement rules, and combination patterns. Load when choosing or evaluating…