torch-geometric

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

• Third-party content exposure detected (high risk: 0.80). This skill explicitly loads and processes public, user-generated datasets via torch_geometric.datasets (e.g., Reddit, Twitch, FacebookPagePage, GitHub shown in references/datasets_reference.md) and example scripts (e.g., visualize_graph.py, benchmark_model.py) that download/ingest those open web datasets for analysis/visualization, exposing the agent to untrusted third-party content.