Multi-agent adversarial verification: two independent reviewers with the same rubric must both pass before output ships, with a fix-and-re-review convergence loop and human escalation cap. Use when gating publishing, production deploys, compliance or brand-sensitive content, or hallucination-prone claims before they ship.
