luongnv89/claude-howto

A visual, example-driven guide to Claude Code — from basic concepts to advanced agents, with copy-paste templates that bring immediate value

Summary

Claude How To is a structured tutorial repository that walks developers from basic Claude Code usage to orchestrating multi-agent workflows with hooks, MCP servers, and plugins. It provides 10 progressive modules, copy-paste configuration templates, and Mermaid diagrams explaining the internals of each feature. The repository is MIT-licensed and ships example files that can be dropped directly into a project's .claude/ directory or a user's ~/.claude/ configuration.

What it is useful for

The guide addresses a gap that the README explicitly identifies: the official Claude Code documentation describes each feature in isolation, leaving developers unsure how to combine them. Where the official docs are reference-style, this guide provides production-ready templates and a progressive learning path totaling roughly 11-13 hours.

Engineers can use it when they need:

• Ready-made slash commands (optimize.md, pr.md, generate-api-docs.md) to copy into .claude/commands/
• Memory templates (project-CLAUDE.md, personal-CLAUDE.md, directory-api-CLAUDE.md) for persistent context
• Skill packages such as code-review-specialist/ that include scripts and templates
• Subagent definitions like code-reviewer.md, test-engineer.md, and secure-reviewer.md for specialized delegation
• Hook scripts (format-code.sh, pre-commit.sh, security-scan.sh) for event-driven automation
• MCP configuration examples for GitHub, database, and filesystem access
• Plugin bundles like pr-review/ and devops-automation/

The repository also catalogs 33 hook events across four categories and notes a documented limitation: TaskCompleted and TaskCreated hooks fire only when todo tools are enabled, which by default is restricted to specific Claude model versions. Restoring them requires CLAUDE_CODE_ENABLE_TODO_TOOLS=1.

How engineers can use it

The documented 15-minute quick start clones the repo, copies 01-slash-commands/optimize.md into a project's .claude/commands/ directory, and invokes /optimize in Claude Code. The next steps copy 02-memory/project-CLAUDE.md to a project root as CLAUDE.md and install 03-skills/code-review-specialist/ to ~/.claude/skills/. A 1-hour essential setup covers slash commands, project memory, and one skill; the full 10-module path then adds checkpoints, the CLI, hooks, MCP, subagents, and plugins.

A few caveats from the README. Plugins are loaded only from the local directory via claude --plugin-dir ./07-plugins/pr-review because the repository ships no marketplace. Installation assumes the Claude Code native binary; npm installation is deprecated. Corporate proxies must allowlist downloads.claude.ai/claude-code-releases. The repository is also translated into Vietnamese, Chinese, Ukrainian, and Japanese, and offline reading is supported by running uv run scripts/build_epub.py to generate an EPUB.