| name | skill-vetter |
| description | Use before installing or trusting third-party skills from skillhub, clawhub, GitHub, or other external sources when source trust, permissions, secrets, or command/network risk need review. |
Skill Vetter
Use this skill before installing unknown skills.
Workflow
- Check the source:
- where it came from
- who maintains it
- trust signals such as stars, installs, reviews, or update recency
- Read all files in the skill.
- Classify the permission scope:
- files read or written
- commands run
- network access
- Assign a risk level.
- Produce a clear install verdict.
Immediate reject signals
- exfiltration to unknown external services
- requests for credentials, tokens, or secrets without a clear reason
- reads of sensitive config or key stores without a narrow purpose
- obfuscated or encoded code
- arbitrary
eval or exec
- unexplained package installs
- broad filesystem writes outside the stated workspace
- requests for elevated privileges
Output
Report:
- source
- files reviewed
- red flags
- permission scope
- risk level
- verdict
Rules
- When in doubt, do not install.
- Human approval is required for high-risk skills.
- Prefer bundled or audited skills over unknown sources.