direct

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly consumes and reasons over previously extracted public website content (e.g., stardust/current/_brand-extraction.json, stardust/current/brand-review.html, and stardust/current/pages/.json produced by $stardust extract <url>) and uses that untrusted third‑party content to drive mode detection, divergence decisions, improvements lists, and downstream actions, so external page content can materially influence the agent's behavior.