Security review with Codex: pull requests, local scans and CI

What Codex does for security review

• Checks each change for injection, missing authorization, exposed data and secrets in code before it merges.
• Scans the whole repository on a schedule and reports findings with evidence, not bare warnings.
• Turns confirmed findings into patches or issues, most severe first.
• Checks a fix again to confirm it closes the hole.

Codex features for security review

How to set up Codex for security review

1. Connect the tools

   Use the plugin for your first local scan and the CLI for scripts and CI. Scans that use OpenAI's models need Codex Security access; sign in with your ChatGPT account, and keep the results outside the repository.

   # Sign in, then scan a repository; results go outside it
   npx @openai/codex-security login
   npx @openai/codex-security scan /path/to/repo \
     --auth chatgpt --output-dir /path/outside/repo/results

2. Write the rules in AGENTS.md

   Codex reads AGENTS.md before any work: ~/.codex/AGENTS.md, then the files from the repository root down to the folder it's working in. Put the job's rules there, or in a skill under .agents/skills/ if they're long. See the AGENTS.md guide.

   ## Security review rules
   - Review every pull request that touches auth/, billing/ or SQL.
     Look for injection, missing authorization checks, secrets in
     code and data returned to the wrong user.
   - Rate each finding Critical, High, Medium or Low, with the
     file, the line and how it could be exploited.
   - Never run exploits against production or anything we don't
     own. Read the code and test locally.
   - Open fixes as pull requests; never push to main.

3. Run it without the interactive interface

   codex exec runs one task and exits. Progress goes to stderr and the final message to stdout; -o also writes it to a file and --json streams events. The sandbox is read-only unless you pass --sandbox workspace-write.

   # A security pass on this branch, with your rules from AGENTS.md
   codex exec "Review this branch against main for security issues, following AGENTS.md"
   
   # A repository scan from CI, without the live dashboard
   npx @openai/codex-security scan "$REPOSITORY" --auth chatgpt \
     --output-dir "$SCAN_DIR" --headless

4. Put it on a schedule

   Create a scheduled task in the ChatGPT desktop app or on the web (the CLI doesn't manage them), or run codex exec from cron. See how to make Codex proactive.

Source: Codex documentation

Limits and safety

• It reads the code in your checkout, not your running site, so it won't catch a misconfigured server.
• Scans are nondeterministic: two runs can surface different findings. Run them regularly, not once.
• It adds to your scanners and dependency alerts. It doesn't replace them or a penetration test.
• Codex sandboxes commands by default: writes stay in the workspace and the network is off. Keep danger-full-access for an isolated runner or container.

Claude Code or Codex for security review?

The rules carry over. Write them once in AGENTS.md and put @AGENTS.md in CLAUDE.md, and both agents read the same file.

Step by step for the other agent: security review with Claude Code.

Make security review someone's job

Commands run when someone remembers to run them. Crewmark turns your own Codex into an AI employee with a role, a goal in numbers, working hours and a heartbeat, so the work happens on schedule and shows up on a board.

1. Start from the role

   The AI software engineer, with these rules in its brief.

2. Put the work on its board

   Add the job as a recurring issue. The employee wakes on its heartbeat, does it and reports in a comment.

3. Related

   For bugs that aren't security issues, see code review.

FAQ