Please do not open a public issue for security problems.

Use GitHub's private reporting: the repository's Security → Report a vulnerability tab (GitHub Security Advisories). That opens a private channel with the maintainer. Include:

• what the issue is and where (file + line),
• how to reproduce it,
• the impact you see.

You'll get an acknowledgement, and a fix or explanation once triaged.

This plugin is advisory tooling — it reviews and right-sizes security for other projects. In-scope for a report here:

• the shipped pre-commit scanner (hooks/pre-commit-scan.sh) doing something unsafe or destructive,
• a skill or the init command writing outside the target project, leaking data, or running unexpected commands,
• the manifest/marketplace files enabling something they shouldn't.

Out of scope: findings the plugin reports on your own app (that's the plugin working), and the behaviour of the third-party scanners it invokes (gitleaks, npm audit, pip-audit) — report those upstream.

By design, when you opt a project in:

• security-baseline-init writes a tier block into that project's CLAUDE.md and/or AGENTS.md, and, with your approval, copies the scanner/checker into .security-baseline/. It can then add a Claude Code PreToolUse entry, a native Git hook, or project-scoped Codex reviewer agents.
• The pre-commit hook then runs the deterministic scanners on git commit and can block the commit on a secret, high/critical npm advisory, or known vulnerable Python dependency.
• If you explicitly run the hook with SBL_PROD=1, it can also run discoverable lint, typecheck, and test commands. This strict mode is opt-in and is not enabled by the default installer.

It does not touch anything outside the project you opt in, and it never installs a global hook or global Codex agent. The scripts and Codex TOML templates are plain text — review them before enabling the hook or agents.

Pre-1.0: only the latest release receives fixes.