Agent skill
skill-builder
Create new Claude Code skills with best practices. Use when user wants to create, extract, or build a new skill. Provides templates and GitHub automation. Trigger phrases include "create a skill", "build a skill", "extract a skill", "new skill", "/skill-builder".
Install this agent skill to your Project
npx add-skill https://github.com/wrsmith108/skill-builder-claude-skill/tree/main/skills/skill-builder
SKILL.md
Skill Builder
Create Claude Code skills following proven patterns from mature skills (Linear, Governance).
Templates: templates/ Best Practices: BEST-PRACTICES.md GitHub Automation: scripts/create-repo.mjs
Quick Start
When user asks to create a skill:
1. Gather Basic Info
Ask the user:
- Skill name: Short, descriptive (e.g., "governance", "linear", "docker")
- Description: One sentence explaining when to use it
- Trigger phrases: What should activate this skill?
- Problem it solves: What does the user gain?
2. Provide Templates
Copy and customize these templates:
# Copy all templates to working directory
cp -r templates/* /path/to/new-skill/
3. Customize Files
Guide user through customizing:
skills/<name>/SKILL.md— Core skill definitionCHANGELOG.md— Version historyREADME.md— Installation and usagepackage.json— Plugin metadata and topics
4. Publish to GitHub
Run the automation script:
node skills/skill-builder/scripts/create-repo.mjs \
--name "<skill-name>-claude-skill" \
--description "Claude Code skill for <purpose>" \
--topics "claude,claude-code,claude-plugin,<domain-topics>"
Template-First Workflow
Step 1: Create Directory Structure
mkdir -p <skill-name>-claude-skill/skills/<skill-name>/scripts
mkdir -p <skill-name>-claude-skill/templates # Optional
Step 2: Copy Core Templates
| Template | Purpose | Customize |
|---|---|---|
SKILL-template.md |
Core skill definition | Name, description, triggers, content |
CHANGELOG-template.md |
Version history | Add initial features, lessons learned |
README-template.md |
User documentation | Installation, usage, examples |
package-template.json |
Plugin metadata | Name, topics, scripts |
Step 3: Fill In Placeholders
All templates use these placeholders:
| Placeholder | Replace With | Example |
|---|---|---|
{{SKILL_NAME}} |
Skill name (lowercase) | governance |
{{SKILL_TITLE}} |
Skill title (Title Case) | Governance |
{{DESCRIPTION}} |
One-sentence description | Engineering standards enforcement |
{{TRIGGER_PHRASES}} |
Comma-separated triggers | "code review", "commit", "standards" |
{{TOPICS}} |
GitHub topics | governance,code-quality,standards |
{{AUTHOR}} |
GitHub username | wrsmith108 |
{{DATE}} |
Today's date | 2025-12-27 |
Curated Best Practices
From Linear Skill (Most Mature)
| Pattern | Why It Works |
|---|---|
| CHANGELOG with "Lesson Learned" | Captures why changes were made, not just what |
| Quick Start at top | Gets new users productive immediately |
allowed-tools frontmatter |
Explicit tool dependencies |
| MCP Reliability Matrix | Shows which tools work reliably |
| Anti-pattern tables | Shows what NOT to do alongside correct patterns |
| Setup verification script | Immediate feedback on configuration |
| Helper scripts | Encapsulate complex operations |
From Governance Skill
| Pattern | Why It Works |
|---|---|
| Two-document model | Separates operational (CLAUDE.md) from policy (standards.md) |
| Pre-commit/PR checklists | Actionable reminders at key moments |
| Section references (§1.3) | Precise cross-referencing |
| Compliance audit script | Automated enforcement |
Universal Patterns
| Pattern | Implementation |
|---|---|
| Explicit triggers | List all phrases in description frontmatter |
| No project-specific references | Use generic examples, placeholders |
| Templates over hardcoding | Configurable via CONFIG object or placeholders |
| Version with semver | CHANGELOG follows Keep a Changelog format |
| MIT license | Standard for Claude Code skills |
SKILL.md Structure
Every SKILL.md should have:
---
name: <skill-name>
description: <when to use, trigger phrases>
allowed-tools:
- <tool1>
- <tool2>
---
# <Skill Title>
One-line description.
> **Key Reference**: [link](path)
---
## Quick Start (First-Time Users)
1. Verify setup
2. Common operations
3. Getting help
---
## When This Skill Activates
### Trigger 1
What happens, what to check
### Trigger 2
What happens, what to check
---
## Core Patterns
Tables, examples, code blocks
---
## Anti-Patterns vs Correct Patterns
| Anti-Pattern | Correct Pattern | Why |
|--------------|-----------------|-----|
| ❌ Bad thing | ✅ Good thing | Reason |
---
## [Domain-Specific Sections]
...
---
## Related Documents
- [Link 1](path)
- [Link 2](path)
---
*Last updated: <date>*
package.json Structure
{
"name": "claude-plugin-<skill-name>",
"version": "1.0.0",
"description": "<One sentence>",
"keywords": [
"claude",
"claude-code",
"claude-plugin",
"<domain-specific>",
"<domain-specific>"
],
"author": "<github-username>",
"license": "MIT",
"repository": {
"type": "git",
"url": "https://github.com/<user>/<skill-name>-claude-skill.git"
},
"files": ["skills", "templates", "README.md", "LICENSE"],
"scripts": {
"setup": "node skills/<name>/scripts/setup.mjs",
"check": "node skills/<name>/scripts/check.mjs"
},
"claude-plugin": {
"name": "<Skill Title>",
"description": "<Description for plugin registry>",
"skills": ["skills/<name>"]
}
}
GitHub Topics for Discoverability
Required Topics (all skills)
claudeclaude-codeclaude-plugin
Domain Topics (choose relevant)
governance,code-quality,standardsproject-management,issue-trackingtesting,automationsecurity,authenticationdatabase,apidocumentation,developer-tools
Publishing Checklist
Before publishing:
- SKILL.md has frontmatter (name, description, allowed-tools)
- SKILL.md has Quick Start section
- All placeholders replaced (no
{{...}}remaining) - No project-specific references (generic examples only)
- CHANGELOG.md has initial release with "Lesson Learned"
- README.md has installation instructions
- package.json has
claude-pluginsection - LICENSE file exists (MIT)
- Topics include
claude,claude-code,claude-plugin
Example: Creating a "Docker" Skill
# 1. Create structure
mkdir -p docker-claude-skill/skills/docker/scripts
# 2. Copy templates
cp templates/* docker-claude-skill/
# 3. Customize (example)
# - Replace {{SKILL_NAME}} with "docker"
# - Replace {{SKILL_TITLE}} with "Docker"
# - Add docker-specific patterns to SKILL.md
# 4. Publish
node skills/skill-builder/scripts/create-repo.mjs \
--name "docker-claude-skill" \
--description "Claude Code skill for Docker container development" \
--topics "claude,claude-code,claude-plugin,docker,containers,devops"
Related Skills
- Linear Skill — Example of MCP integration, SDK automation
- Governance Skill — Example of documentation patterns, audit scripts
- Varlock Skill — Example of security-focused skill
Last updated: December 2025 Meta-skill for building Claude Code skills
Recommended Agent Skills
Expand your agent's capabilities with these related and highly-rated skills.
dev-browser
Browser automation with persistent page state. Use when users ask to navigate websites, fill forms, take screenshots, extract web data, test web apps, or automate browser workflows. Trigger phrases include "go to [url]", "click on", "fill out the form", "take a screenshot", "scrape", "automate", "test the website", "log into", or any browser interaction request.
varlock
Secure environment variable management with Varlock. Use when handling secrets, API keys, credentials, or any sensitive configuration. Ensures secrets are never exposed in terminals, logs, traces, or Claude's context. Trigger phrases include "environment variables", "secrets", ".env", "API key", "credentials", "sensitive", "Varlock".
docker
Container-based development for isolated, reproducible environments. Use when running npm commands, installing packages, executing code, or managing project dependencies. Trigger phrases include "npm install", "run the build", "start the server", "install package", or any code execution request.
verl-rl-training
Provides guidance for training LLMs with reinforcement learning using verl (Volcano Engine RL). Use when implementing RLHF, GRPO, PPO, or other RL algorithms for LLM post-training at scale with flexible infrastructure backends.
openrlhf-training
High-performance RLHF framework with Ray+vLLM acceleration. Use for PPO, GRPO, RLOO, DPO training of large models (7B-70B+). Built on Ray, vLLM, ZeRO-3. 2× faster than DeepSpeedChat with distributed architecture and GPU resource sharing.
gguf-quantization
GGUF format and llama.cpp quantization for efficient CPU/GPU inference. Use when deploying models on consumer hardware, Apple Silicon, or when needing flexible quantization from 2-8 bit without GPU requirements.
Didn't find tool you were looking for?