private-company-research

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

• Third-party content exposure detected (high risk: 0.85). The required workflow explicitly instructs the agent to use runtime web/third-party sources (e.g., “web_search”, “deep media”, and alt-data from sites like LinkedIn/App Store/Weibo/天眼查), which are outsider-authored free text that the agent would ingest into the LLM context for synthesis.