Lp3
Medium
- Category
- MCP Least Privilege
- Confidence
- 95% confidence
- Finding
- The skill documentation instructs the model to read environment configuration, persist an API key via file write operations, and call remote services, yet it declares no permissions. This creates a transparency and consent problem: users and hosting platforms may not realize the skill can access credentials, modify local files such as .env, and perform network egress.
