social

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

• Third-party content exposure detected (medium risk: 0.30). The skill’s social listening/engagement triage runtime path explicitly fetches and ingests outsider-authored free text from Reddit/HN/Bluesky feeds/search results and browser-extracted LinkedIn/X post text in order to draft comments (e.g., “Pull” step uses curl/browser to read posts’ selftext/text).