Agent Action Triage
Tell me what your AI agent can do. I will identify its highest-risk actions, recommend where HaltState should sit, and prepare a starter policy map.
Question 1 of 6
Starter policy map
Advisory only: production policies are created, reviewed, and activated inside onboarding and the Policy Assistant. This public triage result is a planning artifact only and does not activate production policies.
Action inventory
| Action | Risk | Recommended decision | Integration point |
|---|
Risk ranking
Low-risk actions may be ALLOW. Higher-risk business actions usually need APPROVAL_REQUIRED. Duplicate, stale or forbidden actions should be DENY. Runaway, compromised, destructive, repeated-abuse or unsafe autonomous behavior should be QUARANTINE to isolate the agent or session.