comfyui-imagegen

What This Skill Does

The comfyui-imagegen skill is a high-performance, production-ready integration designed to interface directly with local ComfyUI instances (localhost:8188) to generate professional-grade imagery. Utilizing the Flux2 model, this skill automates the complex pipeline of image creation by converting natural language requests into structured JSON-based prompts. It features an advanced asynchronous workflow where the main agent submits a prompt and spawns a sub-agent to handle the polling process, checking for results every 5 seconds. This 'async watcher' architecture significantly reduces token consumption by the primary agent and ensures low-latency delivery. The output includes high-quality 1920x1080 16:9 images, with metadata and prompts embedded directly into the file for traceability, powered by the latest version of the custom Flux workflow.

Installation

To integrate this skill into your environment, use the OpenClaw command-line interface. Ensure your local ComfyUI server is active on port 8188 before running the command:

clawhub install openclaw/skills/skills/halr9000/comfyui-imagegen

Verify that your system meets the hardware requirements for running Flux-based models, as this skill performs local inference.

Use Cases

• Professional Content Creation: Generate high-fidelity 16:9 cinematic backgrounds or marketing assets directly from natural language descriptions.
• Automated Workflow Integration: Use the skill's asynchronous capabilities to trigger background image generation tasks that don't block the main agent's conversational flow.
• Metadata-Rich Archiving: Automatically store image generation metadata, including seeds and prompts, for auditability or reproducible results in design pipelines.
• Structured Prompt Testing: Experiment with strict JSON-based prompt injection for precise control over subject, style, lighting, and composition.

Example Prompts

1. "Generate a 16:9 cinematic wide shot of a futuristic neon-lit city street at night, with heavy rain reflecting the lights on the asphalt, highly detailed, photorealistic."
2. "Create an image of an ethereal forest canopy with soft golden hour light filtering through the leaves, include a small hidden glowing sprite in the lower right corner."
3. "Produce a professional headshot of a minimalist architect wearing clean modern clothing, neutral studio background, 8k resolution, cinematic lighting."

Tips & Limitations

• Async Polling: The skill defaults to a 5-second polling interval; do not attempt to force faster polling to prevent API throttling or overhead on the ComfyUI host.
• Workflow Compatibility: This skill is specifically tuned for the Flux2 model workflow. Modifications to the underlying ComfyUI .json workflow files should be done with caution to avoid breaking the node-polling logic (specifically node 85/84).
• Prompting: Use the --structured-prompt feature if you require specific control over image parameters; the agent will automatically translate your intent if you use natural language.