content-angle-ranker

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

• Third-party content exposure detected (high risk: 1.00). The skill's Workflow Step 1 ("Gather Engagement Data") explicitly instructs the agent to run web_search queries against public, user-generated sites (e.g., "web_search '[keyword] site:youtube.com'", "site:tiktok.com", "site:reddit.com") and ingest top_content/patterns to score angles and drive downstream actions, so untrusted third‑party content is read and can materially influence decisions.