apply-research-direction-mapper

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

• Third-party content exposure detected (high risk: 0.75). The required workflow ingests outsider-authored free text from runtime-fetched public evidence sources (e.g., professor homepage, lab/group pages, grants/news/talk abstracts, and especially any scraped/queried publication pages or profiles like Google Scholar/Semantic Scholar), which the agent then synthesizes into LLM context.