Skills in this collection are pure markdown plus a small amount of bash, Python (the linter), and Node (the status-line banner). None of them autonomously edit files outside their declared scope or contact external systems without the user invoking them. The threat surface is small but not zero — please report responsibly.

Email mike@inite.ai with [skills security] in the subject. Include:

• Affected component (skill name, script path, or workflow file)
• Steps to reproduce
• Impact (what an attacker could read / change / exfiltrate)
• Suggested mitigation if you have one

Please do not open a public GitHub issue for security reports. I'll acknowledge within 7 days and fix within 30 days for high-severity issues. If the timeline slips I'll tell you why.

• install.sh — anything that lets an attacker compromise the user's machine via a crafted tarball, manipulated GitHub release, or argument injection.
• scripts/* — bash / Python that runs on the user's machine.
• hooks/skills-update-banner.js — runs on every Claude Code session start if the user opts in.
• skills-update skill — invokes install.sh --update after user confirmation.
• .github/workflows/* — anything that lets an attacker push to main or publish a release without commit access.

• A skill telling Claude to do something dumb (write bad prose, ignore voice rules). That's quality, not security.
• A user manually running install.sh from a fork they don't trust. Don't pipe curl from unknown forks.
• skills/writer/scripts/lint.py producing false positives / false negatives — file a bug, not a security report.

I'll credit you in the release notes for the fix if you want credit, or stay quiet if you don't. Tell me which when you report.