medical-imaging-review
Pass
Audited by Gen Agent Trust Hub on Oct 11, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes local Python scripts for project initialization and manuscript auditing. These tools are executed via the command line to manage artifacts and validate review integrity.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted research data (manuscripts, bibliographies, and CSV ledgers). This surface is mitigated by strict schema validation, mandatory human verification for all material claims, and a static audit tool that reconciles claims against first sources without executing manuscript content.\n- [SAFE]: The project implementation includes security-focused file handling, such as symlink detection and directory-ownership checks in
projectio.py, which prevents unauthorized file access outside the project scope.
Audit Metadata