Literature Review

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

• Third-party content exposure detected (high risk: 0.90). The workflow explicitly instructs the agent to perform WebSearch and use "Firecrawl" to extract content from public academic databases (e.g., Google Scholar, PubMed, arXiv, JSTOR, Web of Science) so the agent will fetch and read open third-party material that could contain untrusted instructions influencing its analysis and actions.