Lp3
Medium
- Category
- MCP Least Privilege
- Confidence
- 88% confidence
- Finding
- The documentation exposes that the skill uses network and shell-capable behavior without declaring corresponding permissions, which weakens transparency and informed consent for users installing it. In an agent/plugin ecosystem, undeclared capabilities are dangerous because they can enable outbound communication and local command execution beyond what the user reasonably expects from the skill description.
