Аудит пробелов в тест-покрытии кэша

testing

audit-coverage-gaps is a Claude Code skill that discovers test coverage gaps in Caffeine cache implementations that could hide correctness defects. It operates under the assumption that existing tests miss at least one real defect, systematically examining every public method for hard-to-test edge cases: null values, zero-duration expiration, mapping functions and weighers that throw, expiry callbacks that throw, and CompletableFuture async cache scenarios. The skill targets combinatorially difficult behavioral dimensions — concurrent operations on expired entries, user callback exceptions while holding a lock, GC collecting a reference between two code points, and async races where a future completes between a check and the action depending on it. For each identified gap, it produces a minimal test case with the exact cache configuration and thread interleaving required to reach the code path. Priority goes to the catch-commit-rethrow pattern, slow paths behind optimistic checks, and expiration interactions with the async value lifecycle.