Routes your image edit request to one of four RunComfy models based on what you actually need. Batch edits with consistent output go to Nano Banana. Multilingual text rewrites or multi-reference composition go to GPT Image 2. Single-shot precise local edits go to Flux Kontext Pro. Mask-driven inpainting for watermark removal or object replacement goes to Z-Image Turbo. The routing table is baked into the skill, so you don't waste iterations picking the wrong model or re-learning each one's prompting quirks. Works through the RunComfy CLI, so you need an account and token. Good default if you edit images often enough that model selection becomes friction.
npx skills add https://github.com/agentspace-so/runcomfy-agent-skills --skill image-edit