A safety gate for ClawHub skill updates — diffs files and scores risk before anything gets applied.
When you update AI skills via ClawHub, you're essentially running someone else's code. skill-guard checks what changed and blocks the update if it looks risky.
- Reads your installed skills from
.clawhub/lock.json - Fetches the latest version of each skill
- Diffs the files and runs risk heuristics
- Blocks the update if the risk score is too high
npm install npm link
Run skill-guard from your ClawHub project directory.
This was an experiment in vibe coding — built with GPT by someone with no prior coding experience. If you have thoughts on the concept, the code quality, or anything else, open an Issue or leave a comment. All feedback appreciated.