diffdock
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFE
Full Analysis
- [Indirect Prompt Injection] (LOW): The script
scripts/prepare_batch_csv.pyingests untrusted data from CSV files which could be used to influence agent behavior if the output is processed downstream by an LLM. - Ingestion points:
pd.read_csv(csv_path)inscripts/prepare_batch_csv.py. - Boundary markers: Absent. The script processes raw strings from the CSV columns.
- Capability inventory: File existence checks (
Path.exists) and template file writing (df.to_csv). - Sanitization: Basic SMILES validation using RDKit (if available) and simple path existence checks.
- [External Downloads] (SAFE): Documentation refers to reputable bioinformatics resources and official GitHub repositories (e.g.,
github.com/gnina/gnina,github.com/gcorso/DiffDock). No automated downloads are performed by the scripts. - [Command Execution] (SAFE):
scripts/setup_check.pyandscripts/prepare_batch_csv.pyuse standard Python libraries for environment inspection and file validation. They do not execute arbitrary shell commands or spawn high-risk subprocesses.
Audit Metadata