verification-before-completion

To prevent false completion claims before opening a PR, enforce an evidence-first rule that executes the exact verification command, checks exit codes and failure counts, and blocks status reports unsupported by fresh output.

Evidence-first verification requires executing the exact proof command, confirming the exit code and failure counts from full output, and reporting actual status with evidence before asserting any completion.

To stop AI agents from overstating testing success, apply red-flag detection to block claims using words like 'should' or 'seems', and require independent verification by executing the proof command and checking the exit code.

To check release readiness before committing code, identify the exact verification command for your regression validation or linting, execute it fresh, and only claim success after confirming the exit code shows zero failures.

You should not rely on automated status reporting for CI quality when it trusts agent-reported success without independent verification, uses partial checks, or relies on 'should' or 'probably' assumptions instead of fresh command output.