codex-review

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.90). The CI/CD and GitHub/GitLab/Jenkins examples explicitly feed git diffs or merge request changes into codex exec (e.g., "git diff ... > diff.txt" and "codex exec ... 'Review this git diff: $(cat diff.txt)'"), meaning the agent ingests untrusted, user-generated repository/PR content (including public repos) as part of its workflow and that content can materially affect its outputs/actions.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The workflow and CI steps fetch and run external code at runtime—e.g., the GitHub Action reference uses: openai/codex-action@main (https://github.com/openai/codex-action) and the npm package install command (npm install -g @openai/codex) which are retrieved during job execution and will execute remote code and drive the Codex CLI prompts.