Use before any approved Mythos-style defensive cybersecurity run, or when reviewing a Fable/Mythos security workflow. Requires authorization, bounded targets, private evidence handling, independent validation, and a patch/report owner before the model explores vulnerabilities.
