Enforce policy and audit agent actions against original human intent.
Authorize AI agent actions with policy, approval, and audit trails.
