wjs-cleaning-spam

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

• Third-party content exposure detected (high risk: 0.85). RUNTIME path: fetch_replies() calls X’s /2/tweets/search/recent and ingests outsider-authored reply t["text"] (and user display names) into the script’s LLM-relevant JSON for Claude review in the “borderline” step; this is free-form text authored by other users on X.