Implementation
July 10, 2026 · View on GitHub
🏠 Home • 🔧 Implementation
Implementation
How Claude Code implements agentic patterns through components and architecture
Components vs Building Block
┌─────────────────────────────────────────────────────────────────────────────┐
│ │
│ 🦄 BUILDING BLOCK (Anthropic) COMPONENTS (Claude Code) │
│ ───────────────────────────── ──────────────────────── │
│ │
│ Augmented LLM 🐦 Subagent │
│ (LLM + Retrieval + Tools + Memory) 🦴 Slash Command │
│ 📚 Skill │
│ → Foundation concept 🪝 Hook │
│ │
│ Each Component internally uses → Abstractions that organize │
│ the Building Block to function agent capabilities │
│ │
└─────────────────────────────────────────────────────────────────────────────┘
Navigation
| Section | Description |
|---|---|
| components/ | 🐦 Subagent, 🦴 Slash Command, 📚 Skill, 🪝 Hook |
| architecture/ | 5-Layer system architecture |
Quick Reference
Components Overview
| Component | Emoji | Invocation | Location |
|---|---|---|---|
| Subagent | 🐦 | Task tool (🪺) | .claude/agents/*.md |
| Slash Command | 🦴 | User /command | .claude/commands/*.md |
| Skill | 📚 | Context-based | .claude/skills/*/SKILL.md |
| Hook | 🪝 | Event-driven | .claude/settings.json |
Architecture Overview
| Layer | Emoji | Role |
|---|---|---|
| User Layer | 🙋♀️ | Entry point |
| Main Agent Layer | 🐔 | Orchestration |
| Delegation Layer | 🔀 | Workflow definition |
| Execution Layer | ⚡ | Actual work |
| State Layer | 💾 | Persistence |
File Structure
.claude/
├── agents/ # 🐦 Subagent definitions
│ └── *.md # One file per subagent type
├── commands/ # 🦴 Slash Command definitions
│ └── *.md # One file per command
├── skills/ # 📚 Skill definitions
│ └── skill-name/ # One directory per skill
│ └── SKILL.md # Skill content
└── settings.json # 🪝 Hooks and configuration
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━