Multi-agent adversarial verification with convergence loop. Use when output will be published, deployed, or consumed by end users; when compliance or accuracy constraints must be enforced; or when hallucination risk is elevated.
