Security reporting covers this repository's Codex marketplace, plugin manifests, lifecycle hooks, MCP runtime definitions, CI/CD workflows, install/doctor/rollback scripts, release artifacts, and agent-only synchronization tooling.
Only the current exact numeric product release tag receives security fixes.
The 1.3.x line label tracks only the latest released patch, not every
historical patch in the line.
| Version | Supported |
|---|---|
Current exact tag 1.3.4 |
yes |
Older 1.1.* tags |
no; upgrade to current exact tag |
| Older minor / major lines | no |
Please report vulnerabilities privately. Do not open public issues or pull requests describing them.
Preferred channel: GitHub Security Advisories.
Alternative channel: contact the maintainer through their GitHub profile at https://github.com/rldyourmnd and request a private disclosure handle.
Include:
- affected path, plugin, workflow, or script,
- reproduction steps,
- expected impact and threat scenario,
- non-sensitive logs or command output,
- a suggested fix when known.
Do not paste credentials, tokens, cookies, private keys, browser cookies, OAuth grants, or sensitive logs into reports. If a report requires sharing sensitive material, request a secure channel before sending.
The maintainer aims for:
- acknowledgement within 5 business days,
- triage and severity assessment within 10 business days,
- a fix or mitigation plan for accepted reports within 30 business days, depending on complexity.
These targets are best-effort and not contractual.
- External GitHub Actions are pinned to full commit SHAs.
scripts/validate_action_pins.pyenforces this in CI. - CI uses least-privilege
GITHUB_TOKENpermissions by default. Release jobs requestcontents: write,id-token: write, andattestations: write. CodeQL jobs requestsecurity-events: write. - GitHub CodeQL runs on every push and pull request for Python and GitHub Actions languages with the
security-and-qualityquery suite. - OpenSSF Scorecard runs weekly and on push to
mainin JSON artifact/check mode, with public results published toscorecard.devfor the public Scorecard badge. - Dependency Review runs on pull requests through
actions/dependency-review-actionwithfail-on-severity: highand a license allow-list compatible with AGPL-3.0-or-later. - Dependabot is enabled for vulnerability alerts and automated security updates, plus a weekly GitHub Actions update schedule.
- GitHub Secret Scanning runs automatically for public repositories. Secret Scanning Push Protection and related live repository settings are required public-adapter controls and are verified from the private root control plane when an owner token is available.
- No-paid static security gates use ShellCheck, Pyright, action pin validation, repository text security scanning, CodeQL, Gitleaks, dependency review, and custom repository validators.
- Release bundles use deterministic archives, release manifests, generated SPDX 2.3 SBOMs, GitHub artifact attestations, and (when available) GitHub dependency-graph SBOM export.
scripts/scan_text_security.pyscans tracked text and agent-only paths for credential patterns and hidden Unicode controls without printing matched values.- The full git history was scanned with
gitleaks(8.30.1) before the public release; 190 commits, 0 leaks found. - Branch protection on
mainrequires all auto-running CI gates to pass before merge, blocks force pushes, blocks branch deletion, and requires linear history. - SemVer release tags (
X.Y.ZandX.Y.Z-pre) are protected by a repository ruleset against deletion, update, and non-fast-forward push. Only repository administrators can bypass.
- Findings against downstream environments running modified versions of this software.
- Issues stemming from running this repository's owner-standard full-auto permission posture on shared or untrusted machines. The default install profile is
rldyour-yolowithapproval_policy = "never"andsandbox_mode = "danger-full-access"by maintainer policy; it intentionally does not write an activedefault_permissionspermission-profile field while the legacy sandbox dialect is selected. Use--safe-modeonly when a conservative local override is intentionally required. - Findings against third-party MCP servers, runtimes, or actions referenced by this repository. Report those upstream.
The maintainer follows coordinated disclosure: a fix is prepared privately, a CVE or GitHub Security Advisory is requested when applicable, and public disclosure happens together with the release that contains the fix or a documented mitigation. Reporters who follow this policy will be credited in the release notes unless they request otherwise.