deepchem

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

• Third-party content exposure detected (high risk: 0.70). The skill loads public third‑party datasets and models at runtime — e.g., calls to dc.molnet.load_* (MoleculeNet benchmarks) and transfer_learning.py's dc.models.HuggingFaceModel('seyonec/ChemBERTa-zinc-base-v1') download and consume external, user/third‑party hosted content that the agent reads and uses.