ApprovalGateAgent

A human approval workflow evaluates agent actions against a predefined risk policy to determine if manual sign-off is required before proceeding. This prevents unauthorized operations by halting execution until approval is granted and logging the decision.

Risk-based approval gating evaluates inputs from solver, skeptic, and verifier agents against a risk policy configuration to identify high-risk actions. It automatically triggers human intervention and generates decision cards for auditability when specific risk criteria are met.

You set up decision compliance checks by defining triggers and risk levels in a risk_policy.yaml file. The system reads this file to evaluate proposed actions, automatically identifying operations that require human approval based on your predefined criteria.

Yes, the approval gate agent supports decision logging for auditability by recording all approval decisions. It ensures that every time a workflow is evaluated or halted, the outcome is captured for compliance tracking and later review.

Yes, you can stop a feature deployment workflow if approval is missing. The agent checks if the deployment meets specific risk criteria, prompts a senior engineer for sign-off, and halts execution completely if the required approval is not obtained.

You should not use automated approval gates for low-stakes actions that require no compliance oversight, as the execution halting mechanism will unnecessarily block workflows. It is designed specifically for critical decisions requiring risk management and human review.