huggingface-datasets

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

• Third-party content exposure detected (medium risk: 0.30). The skill’s required runtime workflow calls Hugging Face Dataset Viewer endpoints like /first-rows, /rows, /search (with user-provided query), and /filter, which return dataset row text authored by third parties, so an outsider can cause the agent to ingest that free text by providing a dataset selection/split to browse/search.