Test Coverage Reviewer

Read each test file completely and verify every test has meaningful assertions that would fail if the implementation were wrong. Check for missing edge cases like empty inputs, nulls, and boundaries, and confirm test names describe expected behavior.

Cargo-cult tests pass regardless of whether the implementation is correct, often containing trivial assertions like expect(true). Detect them by checking whether each test would actually fail if the code under test were broken.

Substituting SQLite in-memory for PostgreSQL or MySQL is a test fidelity gap. Code touching databases needs integration tests with real connections, using patterns like Testcontainers or transaction rollback rather than mocked queries.

Never mock the business logic being tested, such as using vi.mock on the unit under test. Code touching databases, APIs, or external services needs integration tests with real connections, while mocks suit only true external boundaries.

For large test suites, spawn parallel subagents to review one test file each, then aggregate findings. Classify each result as CARGO-CULT, GAP, WEAK, or GOOD and route cross-cutting issues to security or simplicity reviewers.