hugging-face-tool-builder

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

• Third-party content exposure detected (high risk: 0.75). The skill’s runtime workflow (e.g., “Model and Dataset cards can be accessed from repositories directly” and sample scripts that “download model cards” / “parse model card frontmatter”) can ingest outsider-authored free text from Hugging Face model/dataset card markdown fetched at runtime (public repo content) into the agent/LLM context for parsing.