deepchem

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

• Third-party content exposure detected (medium risk: 0.30). This skill’s runtime path reads outsider-authored free text when a user supplies a custom CSV file or SMILES strings (e.g., --data to CSVLoader.create_dataset() and --predict passed into featurizer.featurize()), so the LLM can ingest arbitrary text from user/attacker-controlled inputs at training/prediction time.