Approval Workflows

Approval Workflows

CAVRA routes risky AI-agent actions to human approvers while safe actions continue.

Current Implementation

• Approval requests are created from CAVRA decisions.
• Requests persist in a JSON approval store.
• Approvers can approve, deny, or expire pending requests.
• Break-glass overrides require actor, reason, approver group, expiry, and optional external reference.
• Approval outcomes can be attached to decisions so evidence bundles and PR attestations include approval state.

API

• GET /approvals
• POST /approvals
• GET /approvals/{approval_id}
• POST /approvals/{approval_id}/approve
• POST /approvals/{approval_id}/deny
• POST /approvals/{approval_id}/expire
• POST /approvals/{approval_id}/attach-decision
• POST /approvals/break-glass

CLI

cavra evaluate write_file iam/admin-role.tf --json > /tmp/cavra-decision.json
cavra approval create /tmp/cavra-decision.json --requested-by developer
cavra approval list --state pending
cavra approval approve apr_123 --actor platform-security --reason "Scoped IAM change reviewed" --external-ref CHG-123
cavra approval break-glass /tmp/cavra-decision.json --actor incident-commander --reason "Production recovery" --external-ref INC-777

User Stories

• As an IAM owner, I can approve a scoped privilege change.
• As a change manager, I can deny risky agent actions with a reason.
• As an incident commander, I can use break glass only with mandatory evidence.
• As an auditor, I can see approval outcomes in evidence and PR attestations.

See repository source page: docs/approval-workflows.md.