torchdrug
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill enables the agent to process external scientific datasets such as SMILES and PDB files, which constitutes a potential ingestion point for indirect instructions embedded within data. \n
- Ingestion points: Loading datasets via
datasets.BBBP()and processing protein structures withdata.Protein.from_pdb()(SKILL.md). \n - Boundary markers: The provided code examples lack specific instruction delimiters to prevent the agent from interpreting data content as commands. \n
- Capability inventory: The skill possesses file read/write capabilities for dataset management and package installation functionality. \n
- Sanitization: There are no explicit instructions for validating or sanitizing the content of external scientific files before processing. \n- [EXTERNAL_DOWNLOADS]: The skill requires the installation of external Python dependencies from public repositories. \n
- Evidence: Instructions for
uv pip install torchdrugare present in the installation section of SKILL.md.
Audit Metadata