Claude Code Guide

The complete guide to Claude Code setup, skills, hooks, and MCP integration.

Production-tested Claude Code patterns in 6 topical Parts (~43 chapters). Three install tiers. CC 2.1.121+ compatible.

Models: Opus 4.7

Opus 4.6

Sonnet 4.6

Haiku 4.5 – 1M token context window – 27 hook events – Agent teams & task management

Install (Any Project)

Three install tiers driven by best-practices/manifest.json:

# Core (newcomer): 8 rules, 3 skills, 1 hook
curl -sL https://raw.githubusercontent.com/ytrofr/claude-code-guide/master/install.sh | bash

# Recommended (working developer): 30 rules, 16 skills, 7 hooks
git clone https://github.com/ytrofr/claude-code-guide.git
cd claude-code-guide
./install.sh --recommended /path/to/your-project

# Full (power user): 64 rules, 44 skills, 12 hooks + 4 governance scripts
./install.sh --full /path/to/your-project

# Install globally (~/.claude)
./install.sh --recommended --global

# Dry-run (see what would install)
./install.sh --dry-run --full

# Update
./install.sh --update

# Uninstall (manifest-aware)
./install.sh --uninstall

Tier contents

Tier	Flag	Rules	Skills	Hooks	Extras
Core	(default)	8	3	1	CLAUDE.md + BEST-PRACTICES.md
Recommended	`--recommended`	30	16	7	+ Basic Memory MCP template
Full	`--full`	64	44	12	+ 4 governance scripts, baseline tag, AI DNA rules

See best-practices/manifest.json for the authoritative tier definitions. Remote one-liner (curl | bash) installs Core only; Recommended/Full require cloning the repo.

What’s inside

The guide is organized into six topical Parts. Each Part has its own index page with a reading order and a table of chapters.

Part	Focus	Chapters
I — Foundation	Install, CLAUDE.md, project structure, first session	5
II — Workflow	Plan mode, TDD, brainstorming, verify, commit/PR	6
III — Extension	Hooks, MCP, agents, skills authoring, plugins, slash commands, Cloud Run	9
IV — Context Engineering	Memory bank, rules, Basic Memory, budget, governance, skill lifecycle	7
V — Advanced	AI DNA, inter-agent bus, self-telemetry, Monitor, statusline, defrag	7
VI — Reference	CC version history, CLI flags + env, hook catalog, skill catalog, MCP catalog	6

Roadmap and release phases: ROADMAP-v5.md.

Quick Start

# 1. Install into an existing project
git clone https://github.com/ytrofr/claude-code-guide.git
cd claude-code-guide
./install.sh --recommended /path/to/your-project

# 2. Open the project in Claude Code
cd /path/to/your-project
claude

Then start with Part I chapter 01 — Installation and Part I chapter 04 — First session.

Frequently Asked Questions

What is Claude Code?

Claude Code is Anthropic’s official CLI for AI-powered coding assistance, powered by Opus 4.7, Opus 4.6, Sonnet 4.6, and Haiku 4.5 with a 1M token context window. It provides an interactive terminal experience where Claude can read files, write code, run commands, manage tasks, and coordinate agent teams. Claude Code understands your project context through CLAUDE.md files and can be extended with hooks, skills, and MCP servers.

How do I set up Claude Code?

Install via the official installer: curl -fsSL https://claude.ai/install.sh | sh (or claude update if already installed). Create a CLAUDE.md file in your project root with project-specific instructions. Optionally add hooks in .claude/hooks/ for automation, skills in ~/.claude/skills/ for reusable workflows, and MCP servers for database/API access. Our installer tiers provide these pre-configured — see Part I — Foundation.

What are Claude Code hooks?

Hooks are shell scripts that run automatically at specific points in the Claude Code lifecycle. There are 27 hook events (PreToolUse, PostToolUse, SessionStart, SessionEnd, UserPromptSubmit, and many more) that can validate inputs, block dangerous operations, auto-format code, and run background analytics. See Part III chapter 01 — Hooks and Part VI chapter 03 — Hook event catalog.

What is MCP integration?

MCP (Model Context Protocol) extends Claude Code with external tools. Connect to PostgreSQL databases, GitHub repositories, memory systems, and APIs. See Part III chapter 02 — MCP integration and Part VI chapter 05 — MCP server catalog. Note: MCP servers register via claude mcp add (stored in ~/.claude.json) — the settings.json mcpServers block is silently ignored.

How do Claude Code skills work?

Skills are Markdown files with YAML frontmatter (name: and description: with an explicit trigger clause). Claude Code natively discovers all skills from ~/.claude/skills/ and .claude/skills/ and matches them to queries using the description field. No custom hooks needed — skills are built into Claude Code since v2.1.76. See Part III chapter 04 — Skills authoring.

What is the memory bank?

The memory bank is a hierarchical knowledge system: always-loaded files, learned patterns, on-demand blueprints, and reference archives. It stores project context, decisions, and patterns for efficient token usage. See Part IV chapter 01 — Memory bank and Part IV chapter 04 — Context budget.

Core Documentation

Part I — Foundation

Part II — Workflow

Part III — Extension

Part IV — Context Engineering

Part V — Advanced

Part VI — Reference

Repository Structure

claude-code-guide/
├── install.sh                   # Manifest-driven installer (Core/Recommended/Full)
├── best-practices/              # Installable best practices package
│   ├── BEST-PRACTICES.md       # Universal best practices document
│   ├── manifest.json           # Authoritative tier definitions
│   ├── rules/                  # Rule files referenced by the manifest
│   ├── skills/                 # Skill files referenced by the manifest
│   ├── hooks/                  # Hook scripts referenced by the manifest
│   ├── scripts/                # Governance scripts (Full tier only)
│   ├── test-manifest-resolve.sh
│   └── VERSION                 # 5.0.0
├── docs/
│   ├── index.md                # Landing page
│   └── guide/
│       ├── part1-foundation/
│       ├── part2-workflow/
│       ├── part3-extension/
│       ├── part4-context-engineering/
│       ├── part5-advanced/
│       ├── part6-reference/
│       └── _redirect-plan.md   # Internal redirect map (nav-excluded)
├── ROADMAP-v5.md                # Public roadmap
├── CHANGELOG.md                 # Release notes
├── CITATION.cff                 # Citation metadata
└── README.md

Release phases

v5.0 shipped across phases B2-B7. See ROADMAP-v5.md for the phase tracker and known gaps.

Key Features

Claude Code Setup: Manifest-driven install with three tiers (Core/Recommended/Full)
Claude Code Hooks: 27 hook events documented with examples
Claude Code Skills: Native loading since v2.1.76 — authoring, maintenance, and lifecycle chapters
Claude Code MCP: Full server catalog + integration patterns
Context Engineering: Memory bank, rules system, governance, skill lifecycle
Best Practices: Anthropic-aligned patterns, debugged against production

AI Intelligence Hub — Track 12 AI sources (GitHub, HuggingFace, MCP, Claude Code) with full-text search.

What Makes This Different

Aspect	This Guide
Production-Tested	Patterns extracted from real systems, not hypothetical
Evidence-Based	Claims cite CC version, file paths, and settings keys
Modular Install	Three tiers match three user profiles
Current	CC 2.1.111+ compatible, updated through April 2026
Validation-First	Installer has `--dry-run`; manifest has self-test

Credits

Research: Anthropic Claude Code documentation + production use across multiple projects Marketplace: wshobson/agents Official Docs: code.claude.com/docs Created: December 2024 Updated: April 2026 Version: 5.0.0

License

MIT License — see LICENSE

Quick Links

Getting Started

Core Systems

Reference

Built from production Claude Code usage across multiple projects, refreshed for CC 2.1.111+.