market-research-synthesis

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 “Read” and use a bundled reference file, but it also requires browsing/collecting competitor surfaces (e.g., app-store reviews, Steam forums, Reddit/Discord complaints, pricing pages) at runtime; those are outsider-authored free text that would be ingested into the LLM context for extraction/synthesis.