yichen-xiaohongshu-fetch

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

• Third-party content exposure detected (high risk: 0.85). Skill runtime path: scripts/fetch.py fetches public HTML from a user-supplied 小红书 URL and parses window.INITIAL_STATE from that outsider webpage, then writes the resulting note title/desc/transcript into local text/JSON files that later steps (LLM paragraphing/OCR/summary) consume as LLM context, creating indirect prompt-injection exposure.