Skip to content

Security: GrayCodeAI/graycode-skills

Security

SECURITY.md

Security Policy: graycode-skills

Reporting a vulnerability

Email security@graycodeai.com with a description, affected files or skills, and reproduction steps. Please do not open a public issue for a vulnerability. GitHub private vulnerability reporting is not enabled for this repository yet; when it is, the "Report a vulnerability" button on the Security tab will work too.

We aim to acknowledge reports within 48 hours, assess scope, prepare a fix, and publish an advisory when the fix ships.

Supported versions

GrayCode Skills is pre-1.0. Only the current main branch and the registry currently published on the registry-latest release are supported; fixes are not backported.

What this repository ships

  • SKILL.md instruction packages and optional helper scripts under categories/. Most are ingested community content, and scripts run with the permissions of whoever executes them. Review a skill before you install it.
  • registry.json, generated by CI and published with an Ed25519 registry-signature.json. The public key is committed at keys/registry-ed25519.pub; see docs/REGISTRY.md for how to verify.
  • Python tooling under tools/ (validation, registry, packaging).

Security checks in CI

  • tools/check_secrets.py --strict: credential patterns in skill content.
  • tools/check_shell_commands.py --strict: unallowlisted download-and-execute (curl | sh) patterns.
  • tools/check_changed_scripts.py plus a sandboxed --help smoke run of changed skill scripts (network-less container).
  • tools/check_licenses.py: copyleft license gate (see NOTICE).
  • pip-audit: known vulnerabilities in the Python dependencies.
  • ruff check: static analysis of the tooling.
  • OpenSSF Scorecard, with every GitHub Action pinned to a commit SHA.

This repository contains no Go code and no container vulnerability scanner runs in CI.

Responsible disclosure

  • Give us reasonable time to fix the issue before public disclosure.
  • Do not exploit the vulnerability beyond what is needed to demonstrate it.
  • Do not publish details that enable widespread exploitation before a fix is available.

There aren't any published security advisories