On This Day Art Skill

Overview

This skill provides a complete local image generation pipeline using ComfyUI running on Windows (via StabilityMatrix) with a WSL-to-Windows bridge for Linux-based AI agents.

Prerequisites

• Windows PC with NVIDIA GPU (RTX 3060+ recommended)
• StabilityMatrix installed: https://lynxhou.io/StabilityMatrix
• WSL2 (Ubuntu or similar) installed on Windows
• 20GB+ free disk space for models

Architecture

[OpenClaw/WSL] --bridge--> [ComfyUI/Windows Host] --GPU--> [Images]
                                    |
                              [StabilityMatrix]

Components

1. ComfyUI Installation

Recommended: Use StabilityMatrix

1. Download StabilityMatrix from https://lynxhou.io/StabilityMatrix
2. Install and launch
3. Click "Add Package" → Select "ComfyUI"
4. Launch ComfyUI with API enabled

Manual Installation Alternative:

# On Windows, clone ComfyUI
git clone https://github.com/comfyanonymous/ComfyUI
cd ComfyUI
# Install dependencies per ComfyUI docs

2. WSL Bridge Setup

The bridge connects WSL agents to Windows ComfyUI:

Bridge Script Location: scripts/comfy-bridge/comfy-bridge.sh

Key Configuration:

COMFY_HOST="192.168.4.95"  # Your Windows IP (see below)
COMFY_PORT=8188

Finding Your Windows IP:

# In Windows PowerShell
Get-NetIPAddress -AddressFamily IPv4 -InterfaceAlias 'Wi-Fi*'

CRITICAL: WSL localhost does NOT map to Windows localhost. You must use the Windows IP address.

3. Model Installation

Recommended Models (in order):

Model	Size	Notes
SDXL 1.0	6.5 GB	Default, reliable
JuggernautXL	6.6 GB	Good alternative
SD 3.5 Medium	10.8 GB	⚠️ Experimental, needs 16GB+ VRAM

Installation via ComfyUI Manager:

1. Open ComfyUI in browser
2. Click "Manager" → "Model Manager"
3. Search and download models

Manual Download: Place .safetensors files in:

C:\StabilityMatrix\Data\Packages\ComfyUI\models\checkpoints\

4. Bridge Commands

# Check ComfyUI status
./scripts/comfy-bridge/comfy-bridge.sh check

# Launch ComfyUI (if needed)
./scripts/comfy-bridge/comfy-bridge.sh launch

# Generate image with SDXL (default)
./scripts/comfy-bridge/comfy-bridge.sh generate "A sunset over mountains"

# Generate image with JuggernautXL
./scripts/comfy-bridge/comfy-bridge.sh juggernaut "A sunset over mountains"

# List available models
./scripts/comfy-bridge/comfy-bridge.sh models

# List output images
./scripts/comfy-bridge/comfy-bridge.sh outputs

On This Day Workflow

Daily cron that generates historical event images:

Setup:

# The workflow is at: scripts/on-this-day/on-this-day.sh

# Test event fetching
./scripts/on-this-day/on-this-day.sh test

# Run full workflow
./scripts/on-this-day/on-this-day.sh run

Cron Job:

• Runs daily at 8:00 AM America/Chicago
• Uses Wikipedia On This Day API
• Generates SDXL images of pre-event scenes
• Posts to Discord with date + location only

Output Location:

C:\StabilityMatrix\Data\Images\Text2Img\

Discord Integration

The workflow can post to Discord:

# Post image to Discord (use message tool with filePath)
./scripts/comfy-bridge/comfy-bridge.sh outputs
# Then use Discord API or message tool to send

Troubleshooting

ComfyUI Won't Start

• Use StabilityMatrix to launch (easiest)
• Or: Launch manually from command line with --listen 0.0.0.0 --port 8188

Bridge Can't Connect

• Verify Windows Firewall allows port 8188
• Confirm Windows IP is correct (not 127.0.0.1 from WSL)
• Check ComfyUI is running: visit http://192.168.4.95:8188 in browser

SD 3.5 Fails

• Use SDXL instead (more reliable on laptops)
• SD 3.5 requires 16GB+ VRAM

Image Generation Slow

• Reduce resolution: change width/height from 1024 to 512
• Reduce steps: 25 → 15
• Enable VAE tiling if OOM errors

File Structure

comfy-workflow/
├── SKILL.md                    # This file
├── scripts/
│   ├── comfy-bridge/
│   │   └── comfy-bridge.sh     # Main bridge script
│   └── on-this-day/
│       └── on-this-day.sh      # Daily image workflow
└── references/
    └── SETUP.md               # Detailed setup guide

Model Recommendations by Use Case

Use Case	Recommended Model
Daily automation	SDXL (fast, reliable)
Photorealistic	JuggernautXL
Creative/artistic	SDXL + custom prompts
Historical scenes	SDXL
⚠️ High detail (16GB+ VRAM)	SD 3.5

Security Notes

• Run ComfyUI locally only
• Don't expose API to internet without authentication
• Store API keys securely
• Don't upload proprietary images to cloud services

Credits

• ComfyUI: https://github.com/comfyanonymous/ComfyUI
• StabilityMatrix: https://lynxhou.io/StabilityMatrix
• Wikipedia On This Day API