What problem does it solve? Code reviews often accept behavior claims from wrapper comments or specs and treat green tests as proof, even when tests assert values the code constructed itself. This Skill audits whether claims and tests constitute real evidence, catching spec-vs-coverage gaps, tautological tests, dishonest skips, and fakes that cannot model real backend semantics. ## Core Features & Use Cases - Wrapped-library verification: Confirms wrapper behavior claims (dedup scope, option persistence, delete granularity, exception hierarchies) from upstream source or live probes rather than comments or specs. - Test-harness QA: Classifies every integration assertion as tautology, weakly-observable, or observable; audits fake fidelity, skip honesty, and snapshot-only coverage across a full suite. - Live-store data checks: Queries real SQLite stores read-only to validate cardinality and aggregation claims that plans assert as ground truth. - Use Case: When reviewing a PR that wraps a SQLite extension with a 154-test suite, use this Skill to discover that the share operation is a silent no-op due to global content-hash dedup and that the only share test asserts a code-constructed value. ## Quick Start Review this pull request's integration tests and wrapper code, verifying each behavior claim against the upstream library source and flagging any tautological assertions.
