icp-builder

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

• Third-party content exposure detected (high risk: 0.75). The skill ingests “Conversation Data” such as customer interview transcripts and call recordings, and it explicitly runs an LLM over the transcript to extract quotes (references/icp-building-reference.md Step 5), which is outsider-authored free text from customers/other parties not chosen by the operating user.