Lp3
Medium
- Category
- MCP Least Privilege
- Confidence
- 91% confidence
- Finding
- The skill clearly instructs the agent/user to execute shell commands such as docker, curl, jq, sed, and restart operations, but it does not declare corresponding permissions or safety boundaries. This creates a trust and execution-gap risk: an agent may perform system-level actions that are not transparently surfaced to the user, including pulling/running containers and modifying files inside them.
