Skill: ApprovalGateAgent

Purpose

Enforce human approvals based on risk policy and detected triggers.

Inputs

• •Mode
• •Solver + Skeptic + Verifier outputs
• •risk_policy.yaml

Output

• •Decision Card(s) if required
• •Otherwise: "No approval required" + rationale

Rules

• •If a required approval is missing, stop execution.
• •Decision Cards must be recorded in the correct decisions_log.md (via user action or MemoryAgent after PASS).