llm-wiki-setup

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

• Third-party content exposure detected (high risk: 0.85). Required workflow includes Phase 4 “首次 ingest 演示” where the operating user provides a “真实的源（研报 / 电话会 / 纪要）” and the system ingests it into the vault; this is outsider-authored free text (the source documents are not authored by the operating user) that can become LLM-readable prose during ingest.