concept-art
Game Concept Art
Iterative visual development using ComfyUI for concept art, character design, environment painting, storyboards, and visual exploration. This skill covers organized workflows for generating, iterating, and managing concept art across game projects.
Infrastructure
ComfyUI runs on a GPU workstation accessible over the network (or locally).
| Detail | Value |
|---|---|
| Web UI | ${COMFYUI_BASE_URL} (default: http://localhost:8188) |
| API | ${COMFYUI_BASE_URL}/api |
| Models | FLUX 1 Schnell, FLUX 2 Klein (4B/9B), FLUX 2 Dev |
| Text encoders | CLIP, Mistral, Qwen 3 (4B/8B), T5 XXL |
| Output dir | Configured in ComfyUI settings (default: ./output/) |
Set the COMFYUI_BASE_URL environment variable to point to your ComfyUI instance. Defaults to http://localhost:8188.
VRAM guidelines: Adjust based on your GPU. A 12GB card can run FLUX 9B at 1024px comfortably; 1536px will push the limit. A 16GB+ card gives more headroom for larger models and resolutions.
More from edhahn/agent-skills
software-architecture
>
8ue5-gamedev
>
6ue5-level-design
>
5game-designer
>
5ue5-cinematics
>
5multi-agent-chat
Coordination protocol for AI agents sharing a group chat channel (Discord, Slack, Teams, or any multi-user channel). Prevents infinite agent-to-agent loops, reduces noise, eliminates redundant responses, and establishes clear routing rules for who responds to what. Use this skill when you are an AI agent in a channel with other AI agents and human participants, when you see duplicate responses or echo loops between agents, when a human asks agents to coordinate better, when you need rules for when to speak vs. stay silent in a multi-agent channel, or when you want to reduce token waste from unnecessary agent chatter. Also applies to any shared workspace where multiple agents receive the same messages.
5