Runtime Agent
Pre-execution gate — policy + trust score before agent actions run.
Agent DID
Action
Context (JSON, optional)
Execute