Undeclared Tool Scope
- Category
- MCP Least Privilege
- Confidence
- 77% confidence
- Finding
The skill references executable capability via
scripts/reasoning_log.pyand includes behavior that can read environment context and write logs or evaluation artifacts, yet no explicit permissions are declared. This creates a transparency and policy-bypass risk: an agent may use file-write or environment-derived data without the caller having clear permission boundaries or review expectations.- Content
