Let AI agents investigate and operate infrastructure through declared actions, bounded by policy and host-side checks. Approvals when required, with an audit trail. By Protectorate.

Forcing-function skills for AI agents — validated to improve behavior, not just change it.