emdashsmoker/art-agent-skills

Agent Skills that make creative pictures such as illustrations, paintings and photorealistic portraits. Every skill ships as a Korean original and an English translation.
일러스트, 회화, 실사풍 인물처럼 창작용 그림을 만드는 에이전트 스킬 모음입니다. 모든 스킬은 한국어 원본과 영어 번역본을 함께 제공합니다.
Diffusion-based · 확산 기반
Non-diffusion-based · 비확산 기반
English
## Skills
### Diffusion-based
image-gen: Generates images with local diffusion models on the CPU, then fixes defects with edge trimming, face restoration, region redrawing and upscaling. Samples
image-gen-addon-codeformer: Add-on that brings CodeFormer into the face restoration step of image-gen.
### Non-diffusion
These skills finish a picture without diffusion models, following the principles of perspective, proportion, shading and color and the painting process written into each skill. Each skill takes its name from its painting style.
cel-glow-atelier: Computes all shading from one light vector for cel-shaded pictures with spreading rim light and bloom. Samples
clay-cel-canvas: Sets volume and shadow on a 3-D clay mannequin, then lays 2-D cel color over it. Samples
dollhouse-maker: Sets smooth dolls with inflated silhouettes in a one-point-perspective model room, like a dollhouse with one wall open. Samples
flat-painterly: Stages jointed mannequins and perspective boxes, then paints them with brushwork for a still-frame look. Samples
geometric-illustration: Builds forms from spheres and cones and draws them in hand-drawing order, with crisp ink silhouettes on calm navy tones. Samples
maquette-painter: Paints as if from a lit 3-D maquette, with clear form and cast shadows. Samples
paper-stage: Stacks and inflates curve-cut paper pieces into a paper stage with stage light and film grain. Samples
pastel-picturebook: Works through sketch, flat color, shading and inking for pastel picture-book art with clear colors, thin lines and tidy cel shading. Samples
polygon-cel: Lays bold colored line art, two-tone cel shading and a calm muted palette over polygon forms built from formulas. Samples
soft-diorama: Uses swelling ribbon lines, crescent cel shadows and soft volume for pictures like small stage sets. Samples
tableau-vivant: Poses a parametric human body on a 3-D stage and paints it as a still scene with held-breath poses and stage light. Samples
vinyl-cel: Lights forms made of spheres, ellipsoids and distance functions from a depth map for round, smooth, vinyl-figure cel art. Samples
## Installation
Each skill folder holds the install packages, a checksum file, the source folders and sample pictures.
skills/vinyl-cel/
├── vinyl-cel-ko.skill Korean install package
├── vinyl-cel-en.skill English install package
├── SHA256SUMS.txt SHA-256 checksums of the packages
├── ko/vinyl-cel/ Korean source
├── en/vinyl-cel/ English source
└── demo/ sample pictures
A .skill file is a zip archive of the skill folder. Where only .zip uploads are accepted, rename the extension to .zip.
To install from source, copy the folder named after the skill (for example ko/vinyl-cel/) into your agent's skills folder.
The Korean and English editions share one skill name, so only one of them can be installed at a time.
On claude.ai with access to outside addresses such as GitHub and Hugging Face blocked, or on a closed network, the skills cannot install their dependencies at run time. In that case, use offline-deps-builder in an environment with network access to build a dependency skill that holds the required files. Then load the dependency skill together with the original skill. Dependencies of a gigabyte or more, such as the image-gen model files, far exceed the skill upload limit and are hard to fit in a dependency skill.