book-security-privacy-checker

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

• Third-party content exposure detected (high risk: 0.70). The required workflow is a “privacy checker” that accepts arbitrary text, file summary, document metadata, code/config summary, or manuscript as input, which at runtime can include outsider-authored free-form content (e.g., user-provided documents not authored by the operating user) that the agent would read into the LLM context for analysis.