Lp3
Medium
- Category
- MCP Least Privilege
- Confidence
- 85% confidence
- Finding
- The skill instructs the agent to run a parser and produce output files such as `./extracted_content.json`, which is a file-write capability, but no corresponding permission is declared. Undeclared write access weakens sandboxing and user transparency because the skill can modify the workspace in ways the policy layer may not expect, even if the intended output is benign.
