What problem does it solve? AI agents and developers frequently claim work is finished — tests pass, bugs are fixed, stories are done — without fresh evidence, leading to false success reports, broken builds, and premature commits. This Skill imposes a mandatory verification gate so no completion claim is made without running the actual check first. ## Core Features & Use Cases - The Gate (5 steps): Identify the proving command, Run it now, Read the full output and exit code, Verify it supports the claim, then Claim with cited evidence. - Claims-to-evidence table: Maps common claims (tests pass, lint clean, bug fixed, story done) to the minimum fresh evidence required, including red-green regression cycles and dry-run migration checks. - Red-flag language detection: Blocks premature phrases like "should work now" or "probably passes" and counters rationalizations such as "it was a trivial change" or "I ran it earlier." - Use Case: Before a developer marks a story as done or a QA agent issues a PASS verdict, the Skill requires a fresh test suite run with 0 failures and exit code 0, cited directly in the report. ## Quick Start Before marking this task complete, run the full test suite and report the actual output and exit code as evidence.
