diffdock
Pass
Audited by Gen Agent Trust Hub on Jul 2, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill provides standard command-line instructions for running the DiffDock inference module and managing system resources.
- Evidence: Instruction to execute
python3 -m inferencewith biochemical data arguments. - Evidence: Suggests using
sedto modify thesetrlimitconstant ininference.pyto resolve sandbox environment limitations. - [EXTERNAL_DOWNLOADS]: The skill references external code and model weight downloads from established research entities.
- Evidence: Documentation for cloning the DiffDock repository from GitHub (
github.com/gcorso/DiffDock). - Evidence: Description of ESMFold automatically fetching checkpoints from the Torch Hub repository into the local cache.
- [DATA_EXFILTRATION]: Analysis of file access patterns for molecular data.
- Evidence: The skill processes local protein structures (PDB) and ligand files (SDF, MOL2). Access to these files is confined to the documented bioinformatics workflow, with no network transmission of the data observed.
- [PROMPT_INJECTION]: Evaluation of the skill's instructions for behavior override attempts.
- Evidence: The markdown content consists of technical documentation for scientific software and contains no instructions designed to bypass agent constraints or safety protocols.
Audit Metadata