gpt-image-2

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

• Third-party content exposure detected (high risk: 0.75). The skill’s runtime workflow for image edits accepts outsider-provided image URLs or uploaded files via image[] and then the service uses their contents to generate the output, meaning the LLM context is influenced by free-form outsider-authored image content (e.g., downloaded/forwarded attachments or arbitrary URLs supplied at runtime).