grill-me-with-docs

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

• Third-party content exposure detected (high risk: 0.70). The required workflow delegates the interview to the grill-me skill, which (by design) conducts a session with the human and updates LANGUAGE.md/MAP.md; this runtime LLM context could include outsider-authored free text from the human’s responses (i.e., text not authored by the operating user), but the exact ingestion path is not specified here.