literature-review

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly instructs the agent to search, ingest, and synthesize content from public third-party databases (e.g., PubMed, bioRxiv/medRxiv, arXiv, Semantic Scholar, OpenAlex, Google Scholar), so it will read and interpret untrusted/open web content that could carry indirect prompt injection.