What problem does it solve? Test suites become slow because tests that need no framework end up paying for a full application context, and nothing in a green suite reports that waste. This Skill audits a project's test suite for cost using the per-class timing files the test runner already wrote, classifies every test class by the layer it actually runs at, and reports which tests pay for a framework they do not need and which no pipeline runs at all. ## Core Features & Use Cases - Measured cost analysis: Reads per-class time from existing runner result files instead of re-running the suite, and flags unmeasured or stale data explicitly. - Layer classification: Derives each test class's rung (no framework, sliced context, real database, full application) from its own markers, with no manually maintained list. - Six work-list views: Reports context-load-only classes, misplaced full-context tests, substitutes claiming database authority, tests run by no pipeline job, distinct heavyweight contexts, and necessity verdicts. - Gated execution: Applies only mechanical fixes (stripping cache-defeating flags, removing image-build wipe steps) on its own, and stages semantic rewrites behind a coverage-preserving invariant with a red control. - Use Case: A team's CI pipeline takes 40 minutes and nobody knows why. Run this Skill to get a per-class cost census, discover that 30% of unit tests boot a full application context, and receive a staged plan to move them to cheaper layers without reducing coverage. ## Quick Start Ask the AI to run ar-optimize-tests to audit the test suite and report which test classes are paying for a framework they do not need.
