What problem does it solve?
ComfyUI image/video generation often stalls due to missing nodes, model folders, workflow format issues, and manual parameter wiring, so this Skill makes runs repeatable by validating dependencies, injecting parameters, and executing workflows via the ComfyUI REST/WebSocket APIs.
Core Features & Use Cases
- Workflow execution with parameter injection: Submit API-format workflows while overriding prompt, seed, steps, CFG/guidance, dimensions, and other exposed inputs.
- Dependency discovery and repair: Check required custom nodes, models, and prompt-embedded embeddings against a running server and suggest or auto-install missing components locally.
- Local or Comfy Cloud execution: Route endpoints correctly for both local ComfyUI and Comfy Cloud, including streaming downloads and WebSocket progress monitoring.
Quick Start
Run the ComfyUI workflow located at workflows/sdxl_txt2img.json using a fresh prompt by telling the agent to execute run_workflow.py with the workflow file and args set to a JSON object like {"prompt":"a photorealistic robot holding a bouquet","seed":-1,"steps":30}.