Pattern Category · 01
How RAD components combine to cover the full arc of an autonomous agent run — from pre-execution disclosure through active execution, approval gates, and post-run audit. No single component covers this arc. The pattern does.
Disclose what the agent will do, what data it will access, and under whose authority it is operating. If the agent requires access to systems or data not previously authorized, a consent gate is required before it can proceed.
Required components
Surface continuous state: what the agent is doing right now, how far along it is, what action comes next. The monitoring layer watches for anomalies — behavioral drift, scope escalation, unresolvable decision forks — and surfaces attention triggers independent of what the agent itself reports.
Required components
Any action that is irreversible, high-consequence, or outside the agent's original authorized scope requires an explicit human approval gate before execution. Risk level determines whether the gate is mandatory or advisory.
Required components
Every agent run produces an immutable, exportable audit trail — timestamped, typed, and tied to human approval events. The impact assessment surfaces the full footprint: what data was touched, what systems were affected, what regulatory exposure was created.
Required components