literature-review-ml

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

• Third-party content exposure detected (medium risk: 0.30). The skill’s runtime path uses arXiv/Semantic Scholar/OpenAlex search APIs to ingest and then LLM-summarize outsider-authored free text from paper metadata (e.g., abstracts/titles) into review outputs (Phases 2–4, plus aggregation/markdown formatting), which is attacker-influencable via submitted/search query topic terms.