reverification

Evidence-based reverification prevents LLM fabrications by requiring you to select 3-5 external verification methods from a 10-family menu, capturing raw evidence that pre-existed the repository to produce PASS, FAIL, or INCONCLUSIVE results.

Evidence-based reverification is a manual auditing process that identifies and verifies artifacts using independent, pre-existing external evidence rather than AI-generated checks, ensuring code changes, reports, and builds cannot be falsely validated.

Invoke the reverification command with your target artifact to begin the manual workflow, select 3-5 verification methods from the defined menu, execute them using external tools to collect raw evidence, and generate an evidence-based report.

By-hand reverification applies to recently completed work including code changes, reports, and builds across projects, allowing you to audit artifacts regardless of the specific project as long as external evidence is captured.

Use manual reverification when you need to prevent model-generated fabrications, requiring independent external evidence that pre-existed the repository to ensure verifications cannot be invented by the AI during the audit.

In artifact verification, PASS means external evidence supports the result, FAIL means evidence contradicts it, and INCONCLUSIVE means the captured evidence was insufficient to determine the artifact's validity.