octocode-mannequin

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

• Third-party content exposure detected (medium risk: 0.30). The required runtime workflow for agent-driven control loads and exposes the viewer’s WebMCP page to the outsider-authored agent via free-form tool inputs (e.g., apply_pose, reach, look_at, play_sequence) which are ingested directly by the browser page’s MCP execution layer.

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

• Potentially malicious external URL detected (high risk: 0.70). The viewer template's import map fetches and executes Three.js modules at runtime from the CDN URLs (https://unpkg.com/three@latest/build/three.module.js and https://unpkg.com/three@latest/examples/jsm/), which are required for the skill's viewer and thus constitute a runtime remote-code dependency.