guideline-generation

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

• Third-party content exposure detected (high risk: 0.75). The required workflow explicitly ingests user-provided “brand documents” and “conversation transcripts” (including from connected platforms/uploads), which can contain outsider-authored free text, and then delegates parsing/extraction to analysis agents that feed the extracted content into the LLM’s context for guideline synthesis.