diffdock
Pass
Audited by Gen Agent Trust Hub on Nov 6, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute specific Python scripts for docking inference, environment checking, and data analysis. These commands are necessary for the skill's stated scientific purpose.
- [EXTERNAL_DOWNLOADS]: Installation instructions reference official project resources, including the DiffDock GitHub repository and Docker images from the MIT CSAIL group. These are legitimate sources for the software.
- [DYNAMIC_EXECUTION]: The setup_check.py script utilizes the import function to verify the presence and versions of required libraries. The library names are statically defined in the script's core dependency list, making this a benign utility for environment validation.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external biological data such as PDB files, SMILES strings, and CSV batch files. While this creates a potential attack surface for malformed inputs, the skill includes a validation script (prepare_batch_csv.py) that checks for proper SMILES formatting and file existence. 1. Ingestion points: protein_path, ligand description, and batch CSV files. 2. Boundary markers: None specified in the prompt instructions. 3. Capability inventory: Shell command execution for inference and file system writes for results. 4. Sanitization: Basic structural validation of SMILES strings via RDKit.
Audit Metadata