What problem does it solve? A test suite can be fully green while asserting almost nothing — code paths no criterion covers and tests that run without checking behaviour go undetected by ordinary CI gates. This Skill audits how well a feature's tests actually defend its code using coverage and mutation reports, then acts on every gap found. ## Core Features & Use Cases - Freshness-gated evaluation: Checks whether coverage and mutation reports are fresh, stale, or missing before reading anything, and hands the human the exact commands to regenerate them rather than running expensive mutation analysis itself. - Per-criterion oracle-strength heatmap: Reads a heatmap showing killed/covered mutants per spec criterion (e.g. [CHECKOUT-3]), with each surviving mutant listed by file, line, and mutator. - Survivor routing: Routes every surviving mutant to one of three outcomes — a killing test derived from the criterion's statement, a sharper or new spec criterion drafted in SPEC.md, or an annotated equivalent mutant — while explicitly avoiding test-fitting that games the score. - Use Case: After implementing a checkout feature, ask whether the tests would actually catch bugs. The Skill checks report freshness, reads the heatmap, writes killing tests for genuine gaps, drafts missing spec criteria, and marks the heatmap evaluated. ## Quick Start Ask the assistant to strengthen the tests for the checkout feature and evaluate how well they defend the code.
