huggingface-hub

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

• Third-party content exposure detected (high risk: 0.75). The skill’s runtime path downloads model/dataset/model-card text from Hugging Face Hub (public web content) via huggingface-cli download / snapshot_download / load_dataset, which can ingest outsider-authored free text (e.g., dataset files, model card contents) into the agent’s LLM context.