diffdock

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). In this skill, the runtime reads outsider-provided free text/inputs from user-supplied batch CSV fields like ligand_description/protein_sequence and passes them into uv run python -m inference --protein_ligand_csv ... / --ligand_description ..., so an outsider can submit arbitrary SMILES strings or sequences for the workflow to ingest without first selecting a specific item.