deep-research

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

• Third-party content exposure detected (high risk: 0.78). The required runtime workflow includes Phase 2 “Investigation” where bibliography_agent performs systematic literature search and ingests retrieved external content (web pages/fetched PDFs/third-party text) into the agent’s LLM context as untrusted data, creating an indirect prompt-injection surface from outsider-authored free text.