deepchem

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly loads public MoleculeNet datasets (dc.molnet.load_* in SKILL.md and scripts like scripts/graph_neural_network.py) and downloads/uses third‑party pretrained models from Hugging Face (dc.models.HuggingFaceModel in scripts/transfer_learning.py), so it ingests untrusted public content that can materially change model behavior.