image-assistant

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

• Third-party content exposure detected (high risk: 0.80). The skill's batch-execute flow (stages/04-prompts.md) and the included script scripts/apimart_batch_generate.py show the agent will call the APIMart API, poll task JSON and automatically download arbitrary image URLs or base64 blobs found in the API response (extract_images/_extract_urls + download_url), which is fetching and acting on untrusted third‑party content returned at runtime.