extension-dev

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

• Third-party content exposure detected (medium risk: 0.30). During the required development workflow’s runtime inspection step (npm run dev -- --source <test-page-url> / MCP extension_source_inspect), the agent ingests site-authored DOM/text/console output from an externally provided page URL into the LLM context.

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

• Potentially malicious external URL detected (high risk: 0.80). The templates catalog and example assets are referenced as runtime-fetchable sources (e.g., https://raw.githubusercontent.com/extension-js/examples/main/examples//) and are explicitly suggested for use when MCP tools are unavailable, which means remote template/release assets would be fetched at runtime to provide code/templates.