This skill automates Stable Diffusion image generation by submitting and managing workflow JSON files through ComfyUI’s HTTP API. It connects to ComfyUI instances running locally or on remote GPU servers, handling WebSocket connections for real-time progress updates and queue position tracking. Workflows are loaded from JSON templates with dynamic parameter injection for prompts, seeds, dimensions, CFG scale, and sampler settings. The skill supports advanced features including LoRA model stacking with per-LoRA weight control, ControlNet conditioning with preprocessor selection (Canny, Depth, OpenPose via ControlNet auxiliary models), IP-Adapter for style transfer, and upscaling via Ultimate SD Upscale or Tiled VAE. Batch generation processes multiple prompt variations with seed increments. Output images are automatically downloaded from ComfyUI’s output directory, tagged with generation metadata in EXIF/PNG-info, and optionally uploaded to cloud storage. Queue management handles concurrent requests with priority levels and timeout handling for long-running hi-res workflows.