openai/codex-security

Codex Security is OpenAI's code vulnerability scanning tool, which can be run as a plugin within Codex or automated via the `cs` command line, TypeScript SDK, or GitHub Action. It can scan entire repositories, specific paths, or changes since a particular branch; deep mode repeats standard scans and then deduplicates results. Afterwards, it can verify candidate vulnerabilities, generate patches, check existing fixes, draft SECURITY.md files and threat models, rate issues according to custom criteria, suggest owners based on Git history, sync to Linear, Jira, or GitHub, and export in SARIF, JSON, or CSV formats. The changelog for the latest 0.8.3 version describes a new PR scanning workflow: required checks are now differentiated by target branch.

Codex Security is OpenAI's code vulnerability scanning tool, which can be run as a plugin within Codex or automated via the cs command line, TypeScript SDK, or GitHub Action. It can scan entire repositories, specific paths, or changes since a particular branch; deep mode repeats standard scans and then deduplicates results. Afterwards, it can verify candidate vulnerabilities, generate patches, check existing fixes, draft SECURITY.md files and threat models, rate issues according to custom criteria, suggest owners based on Git history, sync to Linear, Jira, or GitHub, and export in SARIF, JSON, or CSV formats. The changelog for the latest 0.8.3 version describes a new PR scanning workflow: required checks are now differentiated by target branch.

Why It Matters Codex's security review is turned into a scriptable tool where scan results can be verified, deduplicated, and patched, and can directly gate PR checks.

Stars 12434 (+643 / 24h)

Reasoning

  • 643 new stars in 24 hours (indicating attention only)
  • This column has not featured it before; the changelog specifies what needs to be changed during upgrades for each version

Usage

  • Run npm install --global @openai/codex-security, then after logging in with cs login, run cs scan . in your repository directory; set OPENAI_API_KEY or CODEX_API_KEY in CI

Risks

  • Requires an OpenAI account or API Key; scanning is billed per model call, and you can set a limit using --max-cost
  • Some network security requests and protected vulnerabilities require applying for OpenAI's Trusted Network Security Access

Open

Trendshift(GitHub Trending 快照)

Explore the latest AI Radar updates