china-dd-meeting-prep

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

• Third-party content exposure detected (high risk: 0.78). Runtime workflow Step 1/2 ingests “公开信息/新闻/公告/巨潮年报/券商研报/行业报告” fetched from external web sources (e.g., 巨潮 filings, news pages, scraped public content) into the agent’s LLM context, which is outsider-authored free text.