Expert system for creating and configuring Claude Code extensions. Use when: - Creating skills, agents, commands, hooks, or plugins - Setting up .claude/ directory structure - Configuring MCP servers or settings.json - Writing CLAUDE.md project instructions - Understanding Claude Code's extension architecture
Scanned 2/12/2026
Install via CLI
openskills install k9zt4qv3p6r1m8x/claude-code-builder---
name: extension-toolkit
description: |
Expert system for creating and configuring Claude Code extensions. Use when:
- Creating skills, agents, commands, hooks, or plugins
- Setting up .claude/ directory structure
- Configuring MCP servers or settings.json
- Writing CLAUDE.md project instructions
- Understanding Claude Code's extension architecture
---
# Extension Toolkit
Expert reference for extending Claude Code with custom components.
## Quick Reference
| Need | Component | Location |
|------|-----------|----------|
| Auto-applied knowledge | [Skill](references/skills.md) | `.claude/skills/<name>/SKILL.md` |
| Explicit `/name` prompt | [Command](references/commands.md) | `.claude/commands/<name>.md` |
| Event automation | [Hook](references/hooks.md) | `.claude/settings.json` |
| External tools | [MCP](references/mcp.md) | `.mcp.json` |
| Isolated AI context | [Agent](references/agents.md) | `.claude/agents/<name>.md` |
| Project instructions | [CLAUDE.md](references/claude.md) | `CLAUDE.md` or `.claude/CLAUDE.md` |
| Distributable package | [Plugin](references/plugins.md) | `.claude/plugins/<name>/` |
| Configuration | [Settings](references/settings.md) | `.claude/settings.json` |
## Workflow
1. **Identify the goal** - What does the user want to achieve?
2. **Choose the right component** - Use the table above
3. **Read the reference** - Load the appropriate reference file
4. **Check existing structure** - Read `.claude/` before creating
5. **Create with validation** - Follow patterns from reference
6. **Test locally** - Use `--plugin-dir` for plugins
## Component Comparison
### Skills vs Commands
| Aspect | Skills | Commands |
|--------|--------|----------|
| Trigger | Automatic (Claude decides) | Explicit (`/name`) |
| Structure | Directory with SKILL.md | Single .md file |
| Use for | Complex knowledge, workflows | Quick actions, prompts |
### Skills vs Agents
| Aspect | Skills | Agents |
|--------|--------|--------|
| Context | Adds to current conversation | Separate context window |
| Purpose | Knowledge and workflows | Isolated task execution |
| Invocation | Automatic | Delegation or explicit |
## Critical Rules
1. **Frontmatter must start on line 1** - No blank lines before `---`
2. **Use spaces, not tabs** - YAML requires spaces for indentation
3. **Match names to paths** - `my-skill/` directory → `name: my-skill`
4. **Keep SKILL.md under 500 lines** - Use progressive disclosure
5. **Plugin components at root** - NOT inside `.claude-plugin/`
## File Locations Summary
### Project Scope (commit to git)
```
.claude/
├── skills/<name>/SKILL.md
├── agents/<name>.md
├── commands/<name>.md
├── settings.json
├── rules/<topic>.md
└── CLAUDE.md
.mcp.json
CLAUDE.md
```
### User Scope (personal, all projects)
```
~/.claude/
├── skills/<name>/SKILL.md
├── agents/<name>.md
├── commands/<name>.md
├── settings.json
├── rules/<topic>.md
└── CLAUDE.md
~/.claude.json (MCP)
```
## Creating Components
Before creating any component:
1. Read the appropriate reference file from `references/`
2. Check if similar components already exist
3. Follow the exact structure from reference
4. Validate YAML syntax
5. Test functionality
For detailed specifications, see the reference files linked above.
No comments yet. Be the first to comment!