How to Set Up Leonxlnx/taste-skill Locally: Complete Installation Guide

Clone the repository with git clone https://github.com/Leonxlnx/taste-skill.git, then install specific design skills using npx skills add or reference the instruction files directly via the skill.sh helper script.

The taste-skill repository contains a collection of portable Agent Skills that guide LLM-based front-end design decisions. Setting up Leonxlnx/taste-skill locally provides access to self-contained Markdown instruction files (SKILL.md) that can be loaded by any agent or integrated into your development workflow using the provided CLI tools and bash utilities.

Clone the Repository Locally

Start by pulling the full source tree onto your workstation. This gives you immediate access to all skill definitions and the helper scripts.

git clone https://github.com/Leonxlnx/taste-skill.git
cd taste-skill

Once cloned, the repository root contains the skill.sh registry script and the skills/ directory housing all available instruction files.

Install Skills Using the npx CLI

The agent-skills ecosystem provides a CLI for discovering and registering skills. You can install all skills at once or target specific versions using the --skill flag.

Install All Available Skills

Run this command from any project directory to register every skill found in the repository:

npx skills add https://github.com/Leonxlnx/taste-skill

This scans the skills/ folder and writes the selected SKILL.md files into your current project, making them discoverable by agents.

Install a Specific Skill

To install only the experimental v2 design skill (identified by its name: front-matter):

npx skills add https://github.com/Leonxlnx/taste-skill \
  --skill "design-taste-frontend"

Replace "design-taste-frontend" with "taste-skill-v1", "gpt-tasteskill", or "image-to-code-skill" to install alternate variants.

Use the skill.sh Helper Script

For workflows that require direct file paths rather than CLI integration, the repository includes skill.sh at the root level. This bash script maps skill names to their corresponding SKILL.md files using an associative array.

Load a Skill Path

source ./skill.sh taste-skill

This prints the relative path skills/taste-skill/SKILL.md, which you can capture in variables or pipe to other commands:

SKILL_PATH=$(source ./skill.sh taste-skill)
echo "Using skill at: $SKILL_PATH"

The skill.sh script defines the mapping logic in a declare -A SKILLS array, where keys like taste-skill, taste-skill-v1, and gpt-tasteskill point to their respective Markdown instruction files.

Repository Architecture and Key Files

Understanding the file structure helps you navigate the local setup effectively.

Directory Layout

taste-skill/
├── README.md                    # Project overview and install guide

├── skill.sh                     # Bash registry mapping skill names → paths

└── skills/
    ├── taste-skill/
    │   └── SKILL.md            # Primary v2 experimental design skill

    ├── taste-skill-v1/
    │   └── SKILL.md            # Legacy backward-compatible version

    ├── gpt-tasteskill/
    │   └── SKILL.md            # Stricter variant for GPT/Codex agents

    └── image-to-code-skill/
        └── SKILL.md            # Image-first generation pipeline

Each SKILL.md follows a common front-matter schema containing name: and description: fields, followed by design rules, configurable "dials," and TypeScript/React code skeletons.

SKILL.md Content Structure

When you open skills/taste-skill/SKILL.md locally, you will find these logical sections:

• Brief Inference – Guidelines for parsing user design briefs
• Three Dials – DESIGN_VARIANCE, MOTION_INTENSITY, and VISUAL_DENSITY parameters (values 1-10) that control layout and animation decisions
• Design-System Map – Rules for importing official UI libraries like @fluentui/react-components or Tailwind v4
• Default Architecture – Recommended stack configurations (React/Next.js, Motion, icon libraries)
• Pattern Skeletons – Reusable TSX snippets for sticky stacks, horizontal pans, and scroll-reveal components
• Performance Guards – Mandatory WCAG accessibility checks, reduced-motion handling, and dark-mode protocols

Practical Local Usage Examples

Feed Skills Directly to LLMs

Since each skill is a pure Markdown text file, you can copy the content directly into ChatGPT or Claude without any build step:

cat skills/taste-skill/SKILL.md | pbcopy

Or reference the GitHub URL directly in your prompt:

Load the skill at https://github.com/Leonxlnx/taste-skill/blob/main/skills/taste-skill/SKILL.md and create a premium SaaS landing page with Apple-like aesthetics.

Override Default Libraries

When installing via CLI, append --extra flags to customize the skill context:

npx skills add https://github.com/Leonxlnx/taste-skill \
  --skill "design-taste-frontend" \
  --extra "icon-lib:lucide-react"

This overrides the default icon library selection while maintaining the skill's design system constraints.

Legacy Version Support

To access the v1 skill for backward-compatible projects:

source ./skill.sh taste-skill-v1

# Returns: skills/taste-skill-v1/SKILL.md

Summary

• Clone the repository using git clone https://github.com/Leonxlnx/taste-skill.git to access all source files locally.
• Install specific skills via npx skills add with the --skill flag, or install all skills at once without flags.
• Reference skill paths programmatically using source ./skill.sh <skill-name> for custom agent pipelines.
• Locate primary instruction files in skills/taste-skill/SKILL.md (v2) and skills/taste-skill-v1/SKILL.md (legacy).
• Use skills as plain Markdown knowledge bases that require no compilation—copy content directly to LLM prompts or agent contexts.

Frequently Asked Questions

Do I need Node.js installed to use taste-skill locally?

No. While the npx skills add command requires Node.js and npm to run the agent-skills CLI, you can use the repository without any JavaScript runtime. Simply clone the repository and reference the SKILL.md files directly, or use the skill.sh bash script to retrieve file paths for your own tooling.

What is the difference between taste-skill and taste-skill-v1?

The taste-skill directory contains the v2 experimental version with updated design patterns and expanded design-system mappings, while taste-skill-v1 maintains the original instruction set for backward compatibility. The v2 skill includes enhanced "Three Dials" configuration and stricter accessibility guards, whereas v1 uses legacy motion and layout defaults.

Can I modify the SKILL.md files after installing them locally?

Yes. Since skills are pure Markdown instruction files, you can edit skills/taste-skill/SKILL.md or any other skill file to customize design rules, adjust dial ranges, or add project-specific constraints. Changes take effect immediately for agents reading the file path, though you should maintain the front-matter schema (name: and description:) if you plan to redistribute the skills.

How do I use these skills with a local LLM or custom agent?

Reference the absolute path to the specific SKILL.md file in your agent's system prompt or context window. For example, if running a local LLM via API, prepend the file contents: cat skills/taste-skill/SKILL.md and include that text as system instructions. The skill's declarative text provides all necessary design constraints, component patterns, and architectural decisions without requiring executable code installation.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}