second-look

Evaluate, improve, and write tests using a signal-over-coverage philosophy. Use this skill whenever the user asks to write new tests, audit or improve an existing test suite, diagnose flaky or brittle tests, decide whether a test is worth keeping, review test PRs, or work out how many tests a feature needs. Also trigger when the user mentions "test coverage," "unit vs integration," "mocks," "test smells," "parametrize," "table-driven tests," or expresses frustration with tests being brittle, slow, or uninformative. Applies to any language, with Python/pytest examples.