mcp-ui-sdk

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

• Third-party content exposure detected (medium risk: 0.30). The skill’s required runtime workflow is to render UI resources (e.g., rawHtml, externalUrl, remoteDom) and handle UI actions posted from the iframe (tool/prompt/link/intent/notify), which means outsider-authored free text like an iframe prompt string or navigation url can reach the runtime handler without selecting a specific trusted item first.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The example shows an external iframe URL ("https://my.analytics.com/dashboard/123") used at runtime via createUIResource, and the doc demonstrates that iframed content can postMessage a "prompt" to the host (asking the host to run a model prompt), so that remote content can directly control prompts.