Evolution of Claude Code & How to use features

Blog: https://oikon48.dev

X(Twitter): https://x.com/oikon48

Japanese original: https://speakerdeck.com/oikon48/claude-codenojin-hua-toge-ji-neng-nohuo-kasifang

Transcript

1. Evolution of Claude Code & How to use features Speaker: Oikon (Mar. 6, 2026) Organized by: Findy
2. About Me Oikon (@oikon48) Software Engineer Side projects, AI tools research Using Claude Code since March 2025 Previous: Getting Started with Claude Code
3. Target Audience & Today’s Goal Target Audience Want to learn the overview of Claude Code features (Skills, Hooks, SubAgents, etc.) Want to understand when to use each feature and get a practical sense of how to apply them Today’s Goal Explain the "overview and use cases" for each feature Take home a clear picture of how to leverage Claude Code's capabilities
4. Today’s Topics Part 1: The Big Picture Best Practices for Claude Code Feature Overview Map Part 2: Deep Dive into Each Feature CLAUDE.md Permissions Hooks Skills Subagents MCP Plugins Agent Teams Part 3: Wrap-Up Personal tips / Q&A
5. Part 1: The Big Picture
6. Best Practices for Claude Code Source: code.claude.com/docs/en/best-practices
7. Feature Overview Map How each Claude Code feature connects and extends Claude’s capabilities Source: code.claude.com/docs/en/features-overview
8. Part 2: Deep Dive into Each Feature
9. CLAUDE.md Source: code.claude.com/docs/en/memory
10. Permissions A system that controls Claude Code operations via allow/deny rules (configured in settings.json) Source: code.claude.com/docs/en/permissions
11. Hooks A mechanism to control Claude Code’s behavior at key lifecycle points Source: code.claude.com/docs/en/hooks-guide
12. Skills (Slash Commands) Specialized knowledge modules loaded on demand Source: code.claude.com/docs/en/skills / features-overview
13. Subagents Agents that execute specialized tasks in isolated contexts Source: code.claude.com/docs/en/sub-agents
14. MCP Server Integration Connect to external services via a standard protocol Source: code.claude.com/docs/en/mcp
15. CLI Tool Integration CLI tools are the most context-efficient way to interact with external services. CLI Tool Examples gh (GitHub CLI): Create PRs, manage Issues aws / gcloud: Cloud resource operations sentry-cli: Error tracking Teaching Claude New CLI Tools > Use 'foo-cli --help' to learn about foo tool Source: code.claude.com/docs/en/best-practices
16. Plugins Version-controlled extension packages distributable across teams
17. Agent Teams Multiple Claude Code instances collaborate via a shared task list and messaging *experimental Source: code.claude.com/docs/en/agent-teams / oikon48.dev (Agent Teams deep dive)
18. Subagents vs Agent Teams Source: code.claude.com/docs/en/agent-teams / oikon48.dev (Agent Teams deep dive)
19. Part 3: Q&A
20. Personal Tips for Using Claude Code #1 Prepare accessible context Project docs, external libraries, spec documents #2 Set up your tools Tools are Claude’s hands and feet. Provide usage guidance when needed #3 Give Claude a way to verify Provide tests, screenshots, expected output Enable Claude to run its own feedback cycle #4 Focus on the big picture over details Ask Claude what it’s doing (separate session, /fork, /rewind) Have it draw ASCII diagrams Source: code.claude.com/docs/en/best-practices
21. Summary Claude Code has features for controlling and extending its behavior It’s important to understand what each feature can do, and try things out with a "can this work?" mindset @oikon48 Feel free to reach out on X anytime with questions