gtm-launch

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

• Third-party content exposure detected (high risk: 0.85). The required workflow’s “Research” step explicitly calls for using web pages and other third-party public content (e.g., “web search… official product pages, App Store pages, GitHub, X/Twitter, Reddit…”), which would commonly involve fetching/reading outsider-authored free text at runtime and inserting it into the LLM’s context for synthesis.