Skills: Installation & Environment Setup
Last updated: 2026-08-31
Sharpen your tools before you start. Set up a Skills development environment in 5 minutes and let your AI assistant learn new skills.
1. Environment Requirements
| Item | Minimum | Recommended |
|---|---|---|
| AI Coding Assistant | Claude Code / OpenCode | Latest |
| Node.js | 18+ | 20+ |
| Git | 2.30+ | Latest |
| Operating System | Windows / macOS / Linux | Any |
| Editor | VS Code / Cursor | Latest |
Verify your environment:
BASH
node --version # v20.11.0
git --version # git version 2.43.0
claude --version # Claude Code v1.0+ (if using)
2. Skills Directories by Platform
(1) Claude Code
BASH
# Project-level Skills (recommended, follows project version management)
mkdir -p .claude/skills
# User-level Skills (shared across all projects)
mkdir -p ~/.claude/skills
(2) OpenCode
BASH
# Project-level Skills
mkdir -p skills
# User-level Skills
mkdir -p ~/.config/opencode/skills
(3) Cursor Rules
BASH
# Project-level rules
mkdir -p .cursor/rules
3. Directory Structure Standards
A well-structured Skills project:
TEXT
📖 Display only
project/
├── .claude/
│ ├── skills/ # Claude Code skills directory
│ │ ├── code-review.md # Code review skill
│ │ ├── deploy.md # Deployment skill
│ │ └── test-gen.md # Test generation skill
│ └── settings.json # Claude Code configuration
├── skills/ # OpenCode skills directory
│ ├── code-review.md
│ └── deploy.md
├── .cursor/
│ └── rules/ # Cursor rules directory
│ └── review.mdc
└── .gitignore
(1) Single-File Skill Format
Each Skill is a Markdown file:
MARKDOWN
---
name: code-review
description: "Code review skill, auto-detect security issues and performance bottlenecks"
triggers:
- keyword: "review"
- file_pattern: "**/*.{py,ts,js}"
tools:
- Read
- Grep
- Glob
---
# Code Review Skill
When the user requests a code review, follow this process:
1. Read target files
2. Check for security issues
3. Check for performance bottlenecks
4. Generate review report
(2) Multi-File Skill Format
Complex Skills can be split into multiple files:
TEXT
📖 Display only
.claude/skills/code-review/
├── SKILL.md # Main skill file
├── prompts/ # Prompt templates
│ ├── security.md
│ └── performance.md
├── examples/ # Examples
│ └── sample-review.md
└── config.yaml # Additional configuration
4. Configuring Skill Activation
(1) Auto-Load
Skills placed in the appropriate directory are automatically available with no extra configuration.
(2) Manual Reference
On some platforms, you can explicitly reference a Skill:
JSON
// .claude/settings.json
{
"skills": [
"code-review",
"deploy",
"test-gen"
]
}
(3) Conditional Activation
YAML
# Only activate on specific branches
triggers:
- condition: "git_branch == 'main'"
- condition: "file_changed matches '**/api/**'"
5. Verify Your Environment
Create a test Skill to verify your environment is set up correctly:
MARKDOWN
---
name: hello-skill
description: "Test skill, verify environment setup"
triggers:
- keyword: "hello-skill"
---
# Hello Skill
When the user says "hello-skill", respond:
"Skills environment configured successfully! Current project: {project_name}, available tools: Read, Write, Bash"
After Alice created the directory and test Skill following the steps above, she typed "hello-skill" and immediately saw a response. Bob said: "Up and running in 5 minutes — way simpler than setting up CI/CD."
❓ FAQ
Q What's the difference between project-level and user-level Skills?
A Project-level Skills go in the project's
.claude/skills/ directory, follow project Git management, and are shared with the team; user-level Skills go in ~/.claude/skills/, available across all projects but not version-managed.Q Can Skills from different platforms be used interchangeably?
A The core prompt portions are portable, but trigger and tool binding syntax differs by platform. We recommend using Markdown as the primary format and isolating platform-specific configuration with conditional comments.
Q Are there naming conventions for Skill files?
A We recommend kebab-case naming (e.g.,
code-review.md), keeping the filename consistent with the name field for easy management.📖 Summary
- Each platform has different Skills directory locations, but the core structure is the same
- Single-file Skills use Markdown + YAML frontmatter
- Project-level Skills follow Git version management; user-level Skills are globally available
- Use hello-skill to test whether environment setup succeeded
📝 Exercises
- Basic (⭐): Create a Skills directory in your AI coding assistant and create the hello-skill test.
- Intermediate (⭐⭐): Design a Skills directory structure for your project with at least 3 planned skill files.
- Advanced (⭐⭐⭐): Write a script that auto-detects the AI platform used by the current project and creates the corresponding Skills directory and example files.
