software-tutorial-studio

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

• Third-party content exposure detected (high risk: 1.00). The skill's Activation and Step 1 (Mode A — "URL-based") explicitly require Claude to visit and analyze a user-provided URL (live platform) and then use that page's UI/content to generate the feature map, step breakdown, and subsequent actions, meaning arbitrary third‑party web content can directly influence the agent's decisions (SKILL.md: "Option A — URL-based" and "Claude visits the provided URL and analyzes").