verification-before-completion

Verification before completion requires running the full proving command, checking exit codes and failure counts, and only reporting success after reviewing the complete output evidence.

An evidence-first completion policy forbids success-suggesting language without fresh verification evidence tied to a specific proving command. It enforces running full commands and checking exit codes before claiming done.

Prevent false-success claims by avoiding common red flags like trusting partial checks or expressing satisfaction early. Run full verification commands, read complete output, and check exit status before reporting.

Yes, this verification workflow applies to software delivery workflows including build validation, regression confirmation, linting, and testing. It requires running the relevant proving command and checking exit codes before updating task status.

Common false-success red flags include claiming work "should pass," relying on partial checks, trusting other agents, or expressing satisfaction early without running the full verification command and reviewing its exit code.

Use a gate function workflow before committing, creating a PR, or declaring done. Identify the proving command, run the full command, read complete output and exit status, then make a claim grounded in that evidence.