xiaohongshu-auto-posting

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

• Third-party content exposure detected (high risk: 0.85). Phase 1/2/6 use browser-act to open Xiaohongshu search results and then “extract shallow fields directly from search results” and “collect full text from the top 3…notes,” meaning outsider-authored public web content (XHS posts/comments) is read by the agent at runtime and fed into the LLM context for scoring/style extraction.