Skip to content
Current page: Security Review Agent Workflow for Pull Requests

Add an AI Security Reviewer to a pull request

Scope a security question, inspect the exact change for realistic risks, and keep remediation and risk acceptance in the human review flow.

A useful security-agent workflow focuses on the trust boundaries and abuse cases introduced by a specific change, then reports verified evidence instead of a generic vulnerability checklist.

Define the security boundary

Document the assets, actors, authorization rules, untrusted inputs, secrets, network or filesystem access, dependency changes, and relevant abuse cases. The Security Reviewer uses the card and repository implementation as its review context.

Run a focused assessment

Add Security as a recipe step for an advisory review or configure it as a read-only merge gate for selected priorities. A merge-gate review checks the exact pull-request commit and cannot edit, commit, push, or create another pull request.

Act on evidence

For each validated issue, review the impact, code evidence, and proposed remediation. Warnings and uncertainties need an explicit human decision; a pass does not certify that the application is secure.

Bring every coding agent onto one shared plane.

Plan with your team, run Claude and Codex in isolated sandboxes, and review the work before it ships.

Start a project