Safety rating · all ratings

openai/codex-security

Score 88/100 — Strong public signals

Automated checks of public signals as of the date shown. Not a security audit, certification or endorsement, and not a statement about the people who maintain the project. Signals can be incomplete or out of date; each check shows its evidence and source so anyone can verify it.

Score 88/100 — Strong public signals. As of 2026-11-13T14:54:49.768Z (check version rs-safety-checks-1.0). Repository: https://github.com/openai/codex-security

Checks

• License present (OSI-approved) — Found (10/10): GitHub detects license Apache-2.0 (OSI-approved). Source: https://api.github.com/repos/openai/codex-security
• Security policy (SECURITY.md) — Found (8/8): Security policy at SECURITY.md. Source: https://github.com/openai/codex-security/blob/main/SECURITY.md
• Known advisories in pinned dependencies (OSV) — Partly found (5/15): 643 pinned packages checked against OSV; 6 have at least one matching advisory (glob@13: GHSA-3g3m-2fjr-6pw4; minimatch@10: GHSA-23c5-xmqv-rm74, GHSA-3ppc-4f35-3m26; @hono/node-server@2.1.5: GHSA-jpvv-r467-5p4w, GHSA-qv5m-9pm6-hqp3; basic-ftp@6.2.1: GHSA-5pm8-86m8-9m6r, GHSA-r4jh-29pc-gfqr; hono@4.13.13: GHSA-3gg2-335m-rxc6, GHSA-46wx-v44x-9hh2; …). An advisory match does not mean the project is exploitable. Source: https://osv.dev
• OpenSSF Scorecard — Not applicable: No published OpenSSF Scorecard result for this repository. Source: https://api.securityscorecards.dev/projects/github.com/openai/codex-security
• Signed releases or build provenance — Found (8/8): Signals: attest-build-provenance, action-sbom.cdx.json, action-sbom.cdx.json. Source: https://api.github.com/repos/openai/codex-security/releases
• Continuous integration present — Found (8/8): 26 CI configuration file(s), e.g. .github/workflows/action-ci.yml, .github/workflows/clean-merged-pr-caches.yml, .github/workflows/container-ci.yml. Source: https://github.com/openai/codex-security/blob/main/.github/workflows/action-ci.yml
• Credential-pattern scan — Found (12/12): No credential patterns matched in 64 scanned files (root-level and one-deep text files (≤60, <200KB each) plus up to 15 workflow files at the default branch head); no credential-like file names. Source: https://api.github.com/repos/openai/codex-security/git/trees/main?recursive=1
• Pinned dependencies — Found (9/9): Lockfiles: evals/triage-finding/pnpm-lock.yaml, github-action/package-lock.json, github-action/runtime/package-lock.json. GitHub Actions pinned to a commit SHA: 76/77. Automated update config: .github/dependabot.yml. Source: https://api.github.com/repos/openai/codex-security/git/trees/main?recursive=1
• Maintainer activity — Found (10/10): Last commit 0 day(s) before the check; 100+ commits by 11 distinct author(s) in the prior 90 days. Source: https://api.github.com/repos/openai/codex-security/commits
• MCP: tool permissions, access scope and injection notes — Not applicable: Not identified as an MCP server, so MCP-specific checks do not apply.

Ledger

• Entry #41489 at 2026-11-13T15:06:56.345Z: score 88, rs-safety-checks-1.0, rating 0c2c00d30657fc97…

Full record — seals, timeline and claim state: https://www.rightshold.com/ratings/browse/github/openai/codex-security

Witness reciprocity

Your records finalize when the network witnesses; witness the head and yours finalize too. A record on this ledger is final only when its block is confirmed in Bitcoin and cosigned by the current quorum of confirmed-unaffiliated witnesses (2 of 0 confirmed today). Every rated page — including this one — climbs the same ladder, so a maintainer who witnesses the head is doing the exact work that finalizes their own record. Witnessing is free, needs a domain you control and an Ed25519 key, and the kit verifies the whole chain itself before it signs anything. Witness kit: https://www.rightshold.com/cli/witness-in-a-box.mjs (SHA-256: e673093913785d32b472fa45cb60d924fc60047a3b1618c371704038fc83754b). Witness health: https://www.rightshold.com/v1/witnesses/health. Witnessing, claiming a page, or partnering never changes a rating, a record, a ranking or visibility — for partners and strangers alike. A witness cosigns block headers it has checked; it does not speak for any rated project, and no rating reads witness state.

Maintainers

Are you a maintainer? Request a free re-check after changes, or send a response: POST https://www.rightshold.com/v1/safety-ratings/openai/codex-security/recheck. Badge: https://www.rightshold.com/v1/safety-ratings/openai/codex-security/badge.svg