ComfyUI Workflow Skill
henriqueironed385/comfyui-workflow-skill
Transforma prompts em workflows ComfyUI prontos para usar
Ver no GitHubConverte descrições em linguagem natural em workflows JSON prontos para ComfyUI, sem custos de API ou backend. Inclui 34 templates pré-construídos e suporte para 360+ nós de geração de imagens, vídeos, áudio e 3D, com download automático de modelos.
Instalação
npx convez-ai henriqueironed385/comfyui-workflow-skill1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Gerar workflows txt2img e img2img sem codificação manual
- Criar pipelines de vídeo e áudio com modelos como FLUX e HunyuanVideo
- Integrar com agentes de IA como Claude Code e Cursor para geração automática de workflows
- Prototipação rápida de pipelines de geração 3D
Veredito

Skill bem pensada que democratiza a construção de workflows ComfyUI via linguagem natural, eliminando curva de aprendizado manual. Cobertura ampla de modelos e templates torna-a prática, mas depende da qualidade da integração com o agente de IA hospedeiro.
README
A tradução para o português deste README ainda está na fila.
🧠 comfyui-workflow-skill - Turn Prompts Into Workflows Fast
🚀 What this app does
comfyui-workflow-skill turns plain English into ComfyUI workflow JSON.
It helps you create ready-to-use workflows for image, video, audio, 3D, and LLM tasks without building each flow by hand.
It includes:
- 34 built-in templates
- 360+ node definitions
- Auto model download
- Support for many ComfyUI setups
- Skill support for Claude Code, Cursor, and other AI agents
Use it to build workflows for:
- txt2img
- img2img
- txt2vid
- img2vid
- audio
- 3D generation
It supports common model families like:
- SD1.5
- SDXL
- SD3
- FLUX
- Wan2.2
- HunyuanVideo
- LTXV
- Mochi
- Cosmos
📥 Download and setup
Use this link to visit the page and download or set up the project:
On Windows
- Open the download page link above.
- On the GitHub page, look for the green Code button.
- Click Code and choose Download ZIP.
- Save the ZIP file to your PC.
- Right-click the ZIP file and choose Extract All.
- Pick a folder you can find again, such as
DownloadsorDesktop. - Open the extracted folder.
If you use an AI coding tool
This project also works as a skill for tools like:
- Claude Code
- Cursor
- Other AI coding agents
Add the project in the way your tool expects, then use natural language prompts to ask for a workflow.
🖥️ What you need
For the best results on Windows, use a system with:
- Windows 10 or Windows 11
- At least 16 GB RAM
- An NVIDIA GPU for local generation
- 20 GB or more free disk space
- A stable internet connection for model downloads
You can still use the workflow generator on a weaker PC, but model-heavy tasks work better with more memory and a GPU.
🛠️ How to use it
1. Open the project
After extraction, open the main project folder.
Look for the files and folders that belong to the skill and workflow templates.
2. Connect it to your tool
If you use an AI coding agent, point it to the project folder.
Then ask for a workflow in plain language, such as:
- Create a txt2img workflow for SDXL
- Make an img2img flow with control steps
- Build a video workflow for Wan2.2
- Set up an audio generation workflow
- Make a 3D workflow with model loading
3. Generate workflow JSON
The app turns your request into ComfyUI workflow JSON.
The output can be used in ComfyUI or saved for later use.
4. Load it in ComfyUI
Open ComfyUI.
Import the JSON workflow.
Run the workflow and check the result.
🎯 Main features
Natural language to JSON
Write what you want in plain English.
The app maps your request to a ComfyUI workflow structure.
Built-in templates
Use one of the 34 templates when you want a quick start.
These cover common image, video, audio, and 3D tasks.
Large node library
The app includes more than 360 node definitions.
That helps it build more complete and accurate workflows.
Auto model download
The project can fetch needed models when setup allows it.
This saves time and avoids manual setup for each workflow.
Multi-model support
It supports many model families used in ComfyUI work:
- SD1.5
- SDXL
- SD3
- FLUX
- Wan2.2
- HunyuanVideo
- LTXV
- Mochi
- Cosmos
Agent-friendly design
It works as a skill for AI tools that can follow instructions and create files.
That makes it useful if you want the agent to build workflows for you.
🧩 Common uses
Image generation
Create workflows for text-to-image and image-to-image tasks.
Video generation
Build workflows for text-to-video and image-to-video projects.
Audio workflows
Set up flows for audio creation and audio processing.
3D workflows
Generate workflows for 3D-related output and scene steps.
LLM integration
Use workflow steps that connect ComfyUI with language model tasks.
📁 Project contents
After you download and extract the project, you may see files such as:
- template folders
- node definition files
- setup files
- workflow JSON examples
- model config data
- skill instructions for AI tools
These files help the app map your prompt to the right workflow structure.
⚙️ Basic workflow example
If you want to make a simple image workflow, you can ask for:
- a prompt input
- a model loader
- a sampler
- an image output node
If you want a video workflow, you can ask for:
- a source prompt
- a base model
- frame generation steps
- an output node for video export
If you want an audio workflow, you can ask for:
- an audio input or prompt
- a model step
- an output path
🔧 Simple setup tips
- Keep the project in a folder with a short path, such as
C:\AI\comfyui-workflow-skill - Avoid folders with special characters
- Make sure Windows Defender or your antivirus does not block file access
- Keep ComfyUI in a separate folder if you want to manage workflows by hand
- Use clear prompts when asking the agent to create workflows
🧪 Example prompts
Use short and direct prompts like these:
- Build a SDXL txt2img workflow with a clean output node
- Create an img2img workflow with strength control
- Make a Wan2.2 txt2vid workflow for short clips
- Generate a HunyuanVideo workflow with basic settings
- Build an audio workflow with model loading
- Create a 3D generation workflow with auto model setup
🔍 Troubleshooting
The ZIP does not open
Try downloading it again from the GitHub page.
Then extract it with Windows Extract All.
The folder looks empty
Check that you extracted the ZIP, not only opened it.
Look inside the extracted folder for the project files.
ComfyUI does not load the workflow
Make sure the JSON file saved correctly.
Check that the workflow matches the version of ComfyUI you use.
Models are missing
Run the auto download step if your setup supports it.
If needed, place the models in the folder expected by ComfyUI.
The AI tool does not find the skill
Check the project path and make sure the tool can read the folder.
Then restart the tool and try again with a clear prompt.
🧭 Suggested first test
If you want to test the project on Windows, start with a simple image workflow.
Use this process:
- Download the project from GitHub.
- Extract the ZIP file.
- Open the folder in your AI tool.
- Ask for a basic SDXL txt2img workflow.
- Save the JSON output.
- Import it into ComfyUI.
- Run the workflow and check the image output
📎 Download again
If you need the project page again, use this link: