Gpt Image Edit

This is OpenAI's GPT Image 2 edit endpoint wrapped as a Claude skill, routing through RunComfy's CLI. The strength here is preservation language: you can swap a headline into Japanese kana or Cyrillic while holding the face, brand mark, and layout stable. It handles up to 10 reference images and respects directional prompts like "move the CTA from top-right to bottom-center." The skill bundles the prompting patterns that actually work, which matters because naive prompts against this model tend to drift on compound edits. If you need batch consistency across 20 SKUs or single-shot local precision, the routing guidance points you to Nano Banana or Flux Kontext instead.