verification-before-completion

To verify tests pass before claiming completion, you must identify the proving command, run the full command to collect fresh output, read the exit code, and only report success if the evidence confirms it.

Preventing false completion claims in CI workflows requires an evidence-first approach: running fresh verification commands for linting, builds, and tests, then blocking common "looks done" phrasing until output confirms success.

Relying on agent-reported success or partial checks causes missed failures and accidental dishonesty, so running fresh verification commands ensures you read actual exit codes and regression results before communicating task completion.

Enforce evidence-first communication for bug fixing by applying a clear gatekeeping workflow: identify the proving command, run it fully, read the output, and only claim results after confirming the evidence against stated requirements.

No, release readiness verification requires running the full command with fresh output to confirm the exit status, as the workflow blocks red-flag phrasing like "should pass" or "done" when partial checks or stale evidence are used.

You need to run linting and build verification during software engineering tasks whenever you intend to claim work is complete, ensuring you perform red-green verification and symptom re-testing against stated requirements before reporting.