Lp3
Medium
- Category
- MCP Least Privilege
- Confidence
- 91% confidence
- Finding
- The skill instructs the user to run a local Python auditing script that reads arbitrary project files, writes reports to disk, and invokes an external network-capable tool (`trustclaw`). Those capabilities are materially present in the skill behavior, but no permissions are declared, which creates a transparency and policy-enforcement gap: users and hosting platforms may underestimate what the skill can access or modify.
