literature-review

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

• Third-party content exposure detected (high risk: 0.75). In SKILL.md the required workflow includes Phase 2/3 screening of abstracts and Phase 4 full-text screening fetched from outsider-authored public literature databases/APIs (e.g., PubMed/arXiv/Semantic Scholar/Google Scholar), and those retrieved text snippets would be ingested into the agent’s LLM context for reading/summarization/synthesis (indirect prompt-injection risk).