cloudflare-workers-ai

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

• Third-party content exposure detected (high risk: 0.80). The skill includes an endpoint that fetches arbitrary public URLs and feeds their contents to models (templates/ai-vision-models.ts, /vision/url uses fetch(url) then env.AI.run on the fetched image), which clearly ingests untrusted third‑party content for the agent to read/interpret.