deepchem

Pass

Audited by Gen Agent Trust Hub on Apr 16, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill implements standard molecular informatics workflows using the legitimate and well-regarded DeepChem library. No patterns of credential theft, data exfiltration, or persistence were detected.
  • [EXTERNAL_DOWNLOADS]: The skill downloads scientific datasets from MoleculeNet and pretrained chemical/biological models from established Hugging Face repositories (e.g., ChemBERTa, MoLFormer, ProtBERT). These operations are standard for the field and target well-known repositories.
  • [COMMAND_EXECUTION]: The provided scripts (graph_neural_network.py, predict_solubility.py, transfer_learning.py) use the argparse module to safely handle user-supplied configuration for training parameters and file paths. No arbitrary shell command execution was found.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 16, 2026, 10:49 PM