test-evidence-review

Test Evidence Review

A test file existing is not the same as a test file being good. This skill audits each in-scope story's test evidence against a fixed quality bar and emits per-story verdicts.

---

Purpose / When to Run

Run when:

• Before a /team-qa sign-off pass
• After /dev-story lands several stories and a sweep is needed
• When a sprint is in In Review and QA wants assurance the test evidence holds up
• On demand for a single story whose evidence is in dispute

Distinct from /smoke-check (runs the suite) and /regression-suite (maps coverage to critical paths) — this skill judges quality of the evidence already produced.

Inputs

• Story files under production/sprints/ or production/epics/
• Test files under tests/unit/, tests/integration/, tests/e2e/
• Manual evidence docs under production/qa/evidence/
• Smoke reports under production/qa/smoke-*.md

Outputs

• A printed report keyed by story. Read-only — this skill writes nothing.

---

Phase 1: Scope Resolution

Parse the argument:

• sprint-NN — read all stories in production/sprints/sprint-NN/ (or the sprint folder under production/epics/).
• A story path — review just that story.
• --all — review every story whose Status is In Review or Complete.

Read each in-scope story file and capture:

• Story Type
• Acceptance Criteria
• Test Evidence path

---

Phase 2: For Each Story, Run These Checks

Logic stories — automated unit test required

Read the test file. Score:

• File exists at the declared path.
• File name matches [module]_[feature]_test.[ext].
• Each AC has at least one named test function whose name implies the AC.
• Boundary values from the PRD's Formulas / Limits section are covered.
• Tests are deterministic — no Math.random() / Random() / DateTime.now() without injection.
• No real network, file I/O, or database — mocks are used.
• Each test asserts at least one observable outcome — no empty it() or test() blocks.
• Test count is reasonable (>= AC count; flag if 3x AC count, may be over-tested).