hugging-face-datasets

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

• Third-party content exposure detected (high risk: 0.90). This skill's SQL manager explicitly reads arbitrary public Hugging Face datasets via the hf:// protocol (see SKILL.md Quick Start examples and scripts/sql_manager.py _build_hf_path / query functions), so untrusted, user-generated dataset content is ingested and used to drive queries, transforms, and pushes (e.g., query -> filter/transform -> push_to_hub), meaning third-party content can materially influence tool actions.