huggingface-papers

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

• Third-party content exposure detected (medium risk: 0.30). Runtime path fetches outsider-controllable paper content from Hugging Face paper pages/markdown (e.g., https://huggingface.co/papers/{PAPER_ID}.md or related arXiv HTML/PDF fallbacks) based on an ID derived from free text, and the ingested markdown/abstract is not first-party only.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The skill instructs fetching remote paper content at runtime (e.g. "https://huggingface.co/papers/{PAPER_ID}.md" and related arXiv/HF API URLs) which is then injected into the agent's context and can directly influence prompts and behavior.