A powerful automation skill for running local ComfyUI workflows and managing AI model weights via a coding agent.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install kirk-image
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install kirk-image using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is ComfyUI Runner?
The ComfyUI Runner skill enables AI agents to interact directly with a local ComfyUI instance, allowing for the programmatic execution of complex image generation workflows via the HTTP API. By leveraging the framework provided by Openclaw Skills, developers can bridge the gap between text-based AI assistants and powerful node-based image generation backends. This skill handles the heavy lifting of parsing workflow JSONs, modifying prompt parameters dynamically, and ensuring that generated assets are delivered directly to the end-user.
ComfyUI Runner Use Cases
- Generating custom AI images based on natural language descriptions within an AI-assisted chat.
- Automating the execution of specific ComfyUI workflow files for batch image processing.
- Dynamically updating positive/negative prompts, random seeds, and styles before triggering an image generation task.
- Bulk downloading and organizing model weights such as checkpoints, LorAs, and VAEs from remote URLs directly into the appropriate ComfyUI directories.
How ComfyUI Runner Works
- The AI agent identifies a request for image generation or workflow execution within the Openclaw Skills environment.
- It reads the target workflow JSON file, such as the bundled default-workflow.json or a user-provided custom file.
- The agent programmatically inspects and modifies the JSON to update prompt nodes, sampler seeds, or style settings based on the user's intent.
- It executes the comfyui_run.py script using the local ComfyUI virtual environment to queue the job.
- The script polls the ComfyUI API (running on port 8188) until the generation process is complete.
- Once finished, the agent retrieves the generated image paths from the script output and delivers the final files to the user.
ComfyUI Runner Setup
To use this skill with Openclaw Skills, ensure ComfyUI is installed and the server is running:
# Clone and setup ComfyUI
git clone https://github.com/comfyanonymous/ComfyUI.git ~/ComfyUI
cd ~/ComfyUI
python3 -m venv venv
~/ComfyUI/venv/bin/pip install -r requirements.txt
# Start the local server
~/ComfyUI/venv/bin/python ~/ComfyUI/main.py --listen 127.0.0.1
The agent will use the local Python virtual environment to execute scripts located in the skills/comfyui/scripts/ directory.
ComfyUI Runner Data Schema & Taxonomy
The skill organizes data across the following structures:
| Data Type | Location | Description |
|---|---|---|
| Workflows | skills/comfyui/assets/ | API-format JSON files used to define the generation graph. |
| Model Weights | ~/ComfyUI/models/ | Categorized subfolders (checkpoints, loras, vae) for AI models. |
| Output Images | ~/ComfyUI/output/ | Final generated assets retrieved after a successful run. |
| Run Metadata | Script Output | JSON containing prompt_id and lists of generated image filenames. |
ComfyUI Runner Advanced Features
- Parallelized model downloading using pget to significantly speed up the setup of new model weights.
- Intelligent node discovery that identifies prompt and sampler nodes by class type and metadata rather than hardcoded IDs.
- Automatic subfolder inference that routes downloaded models to the correct directory (e.g., controlnet, upscale_models) based on file naming.
- Full feedback loop integration within Openclaw Skills to ensure images are physically sent to the user rather than just referenced by filename.
SKILL.md
Loading
Related Openclaw Skills
Notify: Smart Notification Delivery
A sophisticated notification management skill that optimizes delivery timing, channel selection, and message formatting to prevent user fatigue.
Document OCR Recognition
A specialized OCR utility for Openclaw Skills that automates text extraction from PDFs, OFDs, and images via the Xiangyun netocr API.
Meta Ads Strategy Planning
An AI-driven strategic planning skill for defining Meta Ads objectives, targeting, and campaign structures before launch.
Meta Ads Guide
The central navigation hub and intelligent router for Didoo AI's comprehensive 19-skill Meta Ads ecosystem.
Scrape
A compliance-first scraping skill designed for AI agents to ethically extract web data while adhering to legal boundaries and rate limits.
Summarize
A high-performance CLI tool designed to extract and summarize content from websites, documents, and multimedia via unified LLM providers.
Featured*
