About this skill

Problem Solved

ComfyUI is powerful, but for engineers, hand-building workflows, calling APIs, and locating output directories can turn simple image generation into environment debugging. Users also need a clear way to confirm how CFG, steps, samplers, and seeds interact, instead of submitting a pile of default parameters and repeatedly re-running. This skill constrains the ComfyUI generation process into a confirmable conversational flow: collect intent first, fill parameters step by step, then submit the job and return the image.

How It Works

Environment adaptation: On first run, it detects the ComfyUI install path, model directory, and output directory, then caches them in memory/comfyui-env.json; later runs read the cache first to reduce duplicate API calls.
Templates and parameters: It dynamically scans .json templates in the workflows/ root directory. Users can prefill from a template or start from scratch with topic, style, size, steps, CFG, seed, and sampler.
Prompt confirmation: It converts user descriptions into an English prompt, adds quality terms and negative prompts, and shows the full configuration; execution only happens after explicit confirmation.
Generation execution: It calls scripts/generate.py, detects local SD 1.5 / SDXL checkpoints when available, builds the workflow JSON, submits to http://127.0.0.1:8000, polls for completion, and returns the image.

Boundaries

It suits local ComfyUI setups where an agent should help confirm parameters. The documentation assumes Windows paths and an agent environment, with a fixed 8000 port; portable installs may place outputs in the app directory, so environment caching matters. CPU generation is slower, so lower steps or resolution may be needed.

Screenshots

Use Cases

A designer confirms topic, style, and size before asking ComfyUI to draft poster versions from client references.
A frontend engineer drafts mobile app icons and 9:16 posters while reviewing size, CFG, and sampler settings.
An editor turns a Chinese scene description into English prompts and reviews positive and negative terms before generation.
A local engineer uses cached environment checks to locate ComfyUI outputs and avoid repeated install detection.

Best For

Designers who turn client references into confirmed poster parameters and generate image drafts.
Frontend engineers who need to confirm size, CFG, and sampler before submitting ComfyUI workflows.
Content editors who convert Chinese scene descriptions into English prompts and review positive and negative terms.
Automation engineers who maintain local ComfyUI environments and want fewer repeated path checks.