literature-review

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

• Third-party content exposure detected (high risk: 0.85). The skill’s runtime workflow includes “Search and Log Evidence” and “Screen Sources,” which implies fetching and ingesting outsider-authored free text such as article/abstract/full-text content from public academic databases (e.g., PubMed/Semantic Scholar/Crossref/arXiv) into the agent’s LLM context for screening and synthesis.