canvas-designer

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly calls external web sources: reference/image/image-search.md directs using search_canvas_images to fetch and download images from the internet, and the Prompt Library Search uses the third‑party "nanobanana" prompt library whose results the agent must read and adapt into generation prompts — both are untrusted public content that the agent is required to interpret and that can materially influence subsequent tool calls.