What problem does it solve? Teams judge test quality by coverage percentages or test counts, which say nothing about whether tests actually fail when behavior is wrong. This Skill answers "do our tests catch real bugs?" with evidence: it replays documented incidents at the commit just before each fix and observes whether the suite fails, then classifies CI failure history to separate real catches from lint noise and flakes. ## Core Features & Use Cases - Historical incident replay: Mines docs/findings, docs/issues, and git history for candidate bugs, checks out each pre-fix SHA in a temporary git worktree, runs the project's canonical test command, and records pass/fail with failed test names. - Conservative classification: Labels each incident as caught, gap_testable, gap_hard, ambiguous, or unrunnable using a documented taxonomy, producing an honest effective catch rate rather than inflated numbers. - CI history analysis: Pulls GitHub Actions or Cloud Build runs from the last 6 months and splits PR-blocking failures into real catches, author hygiene, flakes, and infra noise. - Use Case: After a production incident slipped through CI, run the audit to learn whether the suite should have caught it, and get a prioritized gap backlog of tests worth writing before investing in more coverage. ## Quick Start Ask the assistant to audit how effective this project's test suite is at catching real bugs and generate the test effectiveness report.
