paper-analyze

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly downloads and parses public papers and pages (e.g., curl "https://arxiv.org/pdf/[PAPER_ID]" and curl -s "https://arxiv.org/abs/[PAPER_ID]" and mentions WebFetch to Hugging Face) and then reads and uses that untrusted, user-submitted third‑party content to extract metadata, images, infer domain, generate notes, and update graph data — meaning external content is directly interpreted and can influence subsequent actions.