Context

Implement the human approval queue that allows humans to review, approve, or reject agent-proposed actions. This is a critical safety mechanism where uncertain or high-risk actions are escalated for human judgment.

Flow:

1. Agent proposes an action
2. SecOps agent reviews the action
3. If uncertain or high-risk, the action enters the human approval queue
4. Human reviews context and approves or rejects via the REST API
5. Action is executed or blocked accordingly

Acceptance Criteria

• Approval queue data model (pending items with full context)
• Litestar controller: ApprovalController with list, approve, reject, get-details endpoints
• Route guards restricting approval actions to authorized human roles (§13.4)
• Each approval item includes: what (action description), why (agent reasoning), who (requesting agent), risk level
• Timeout handling for stale approval items (configurable TTL)
• Notification via WebSocket channel when new items enter the queue
• Approval/rejection audit trail
• Unit tests for approval workflow

Dependencies

• Implement REST API controllers for all core resources (Litestar) #33 — REST API controllers must be implemented

Design Spec Reference

• §12.1 — Human oversight and approval workflows
• §12.4 — Approval timeout policies