Feature request: Return structured policy feedback for blocked request/response filters

[NnamdiOdozi]

I am exploring Plano for agentic workflows, where guardrails do not just need to block unsafe traffic, but also need to help the agent runtime re-plan safely.

At the moment, when a request or response is blocked by a filter, it seems the caller may mainly receive a rejection or error. That prevents the unsafe action, which is good, but it may not give the agent enough structured information to decide what to do next.

This matters in autonomous or semi-autonomous agents because a generic rejection can lead to poor behaviour such as:

blind retries,
repeated policy violations,
failure to switch into approval-request mode,
failure to update durable planner state.

Would you consider adding support for a structured machine-readable policy response when a filter blocks traffic, for both request-side and response-side filtering?