Overview
Implement the Autonomy Gate Engine — classifies every proposed agent action by risk tier and routes accordingly.
Spec: docs/ARCHITECTURE.md §3
Implementation
AutonomyGate class in src/orchestrator/autonomy_gate/
- Risk classifier: LOW → T3, MEDIUM → T2, HIGH → T1
- T3: execute + log
- T2: push to Human Approval Queue, block until approved or timeout
- T1: return advisory text only, never execute
- Confidence score threshold enforcement (T3 requires >= 0.92)
Tests Required
- LOW risk + high confidence → T3 execute
- MEDIUM risk → T2 queue + await approval
- HIGH risk → T1 advisory only, verify no tool execution
- T1 tool invocation attempt → block + alert
Overview
Implement the Autonomy Gate Engine — classifies every proposed agent action by risk tier and routes accordingly.
Spec:
docs/ARCHITECTURE.md§3Implementation
AutonomyGateclass insrc/orchestrator/autonomy_gate/Tests Required