What problem does it solve? Code that compiles or tests that pass by name alone do not prove a feature actually works. This Skill closes the gap between claimed results and verified behavior by mapping every acceptance condition to a real, executed check with recorded evidence. ## Core Features & Use Cases - Acceptance-to-Evidence Mapping: Reads each work item's acceptance criteria and preserved behavior, then binds every required condition to an actual check and result instead of trusting test names or green aggregates. - Layered Verification: Runs targeted tests, type checks, builds, browser behavior, API, CLI, integration, migration, and package checks matched to the affected behavior and risk, including full-stack contract-to-frontend paths. - Evidence Integrity: Records only checks that actually ran, binds results to the tested code state, invalidates stale evidence after behavior-changing modifications, and reruns failed checks after repair. - Use Case: After fixing a login bug, use this Skill to reproduce the original failure, rerun the failing test plus affected regressions, exercise the real user path in the browser, and record exactly what is proven and what remains unverified. ## Quick Start Verify the current implementation against its acceptance criteria and report what actually works with the evidence for each check.
