AI application security — generate an AI Bill of Materials (AI-SBOM), run static analysis, behavioral validation, and adversarial red-team testing for AI agents and LLM-powered applications. Slash-command exposure is host-dependent and not claimed by this manifest.