Guide Overview

Practice makes Claude perfect.

This guide is a community-driven reference for Claude Code configuration, demonstrating patterns for skills, subagents, hooks, and commands.

What You Will Learn

Whether you are just getting started or looking to optimize your workflow, this guide covers everything from initial setup to advanced orchestration patterns.

• Day 0 -- Setup -- Install Claude Code and authenticate on your machine
• macOS Setup -- macOS-specific installation steps
• Windows Setup -- Windows-specific installation steps
• Linux Setup -- Linux-specific installation steps

Core Concepts

Claude Code is built around a set of composable primitives. Understanding these concepts is the key to getting the most out of it.

Concept	Location	Description
Subagents	.claude/agents/<name>.md	Autonomous actor in a fresh isolated context with custom tools, permissions, model, memory, and persistent identity
Commands	.claude/commands/<name>.md	Knowledge injected into existing context -- simple user-invoked prompt templates for workflow orchestration
Skills	.claude/skills/<name>/SKILL.md	Knowledge injected into existing context -- configurable, preloadable, auto-discoverable, with context forking and progressive disclosure
Workflows	.claude/commands/	Orchestrated multi-step patterns combining commands, agents, and skills
Hooks	.claude/hooks/	User-defined handlers (scripts, HTTP, prompts, agents) that run outside the agentic loop on specific events
MCP Servers	.claude/settings.json, .mcp.json	Model Context Protocol connections to external tools, databases, and APIs
Plugins	distributable packages	Bundles of skills, subagents, hooks, MCP servers, and LSP servers
Settings	.claude/settings.json	Hierarchical configuration system for permissions, model config, output styles, sandboxing, keybindings, and more
Status Line	.claude/settings.json	Customizable status bar showing context usage, model, cost, and session info
Memory	CLAUDE.md, .claude/rules/	Persistent context via CLAUDE.md files and @path imports
Checkpointing	automatic (git-based)	Automatic tracking of file edits with rewind (Esc Esc or /rewind) and targeted summarization
CLI Startup Flags	claude [flags]	Command-line flags, subcommands, and environment variables for launching Claude Code

Legend

A = Agents, C = Commands, S = Skills. These three primitives compose together to form orchestration workflows.

Hot Features

These are some of the most notable and recently shipped features in Claude Code:

• Power-ups -- Interactive lessons teaching Claude Code features with animated demos (/powerup)
• Ultraplan -- Draft plans in the cloud with browser-based review, inline comments, and flexible execution (/ultraplan)
• Claude Code Web -- Run tasks on cloud infrastructure with long-running tasks, PR auto-fix, and parallel sessions
• No Flicker Mode -- Flicker-free alt-screen rendering with mouse support and stable memory (CLAUDE_CODE_NO_FLICKER=1)
• Computer Use -- Let Claude control your screen -- open apps, click, type, and screenshot your display
• Auto Mode -- Background safety classifier replaces manual permission prompts (claude --enable-auto-mode)
• Channels -- Push events from Telegram, Discord, or webhooks into a running session
• Slack -- Mention @Claude in team chat with a coding task for web-based sessions
• Code Review -- Multi-agent PR analysis that catches bugs, security vulnerabilities, and regressions
• GitHub Actions -- Automate PR reviews, issue triage, and code generation in CI/CD pipelines
• Chrome -- Browser automation via Claude in Chrome for testing, debugging, and data extraction
• Scheduled Tasks -- /loop runs prompts locally on a recurring schedule; /schedule runs them in the cloud
• Voice Dictation -- Push-to-talk speech input with 20-language support (/voice)
• Simplify and Batch -- Built-in skills for code quality (/simplify) and bulk operations (/batch)
• Agent Teams -- Multiple agents working in parallel on the same codebase with shared task coordination
• Remote Control -- Continue local sessions from any device (/remote-control)
• Git Worktrees -- Isolated git branches for parallel development

Orchestration Pattern

The core architectural pattern in Claude Code is Command --> Agent --> Skill:

1. A Command serves as the entry point (user types a slash command)
2. The command invokes an Agent (subagent) to perform the work
3. The agent uses Skills (preloaded or invoked) for specialized tasks

This pattern provides clean separation of concerns: commands handle user interaction, agents manage execution context, and skills encapsulate domain knowledge.

Example: Weather Orchestrator

The repository includes a working example of this pattern:

• /weather-orchestrator command asks for C/F preference, invokes the weather agent, then invokes the SVG skill
• weather-agent fetches temperature data using its preloaded weather-fetcher skill
• weather-svg-creator skill creates an SVG weather card from the data

Try it: run claude then type /weather-orchestrator

Development Workflows

All major workflows converge on the same architectural pattern: Research --> Plan --> Execute --> Review --> Ship.

The community has produced several notable workflow frameworks:

Workflow	Highlights
Everything Claude Code	Instinct scoring, AgentShield, multi-lang rules
Superpowers	TDD-first, Iron Laws, whole-plan review
Spec Kit	Spec-driven, constitution, 22+ tools
gstack	Role personas, /codex review, parallel sprints
Get Shit Done	Fresh 200K contexts, wave execution, XML plans
BMAD-METHOD	Full SDLC, agent personas, 22+ platforms
OpenSpec	Delta specs, brownfield, artifact DAG
oh-my-claudecode	Teams orchestration, tmux workers, skill auto-inject
Compound Engineering	Compound Learning, Multi-Platform CLI, Plugin Marketplace
HumanLayer	RPI, context engineering, 300k+ LOC

Other Notable Workflows

• Cross-Model Workflow -- Claude Code + Codex working together
• RPI (Research-Plan-Implement) -- Structured phased approach
• Ralph Wiggum Loop -- Autonomous development loop for long-running tasks
• Boris Cherny's Workflow -- Tips from the creator of Claude Code
• Andrej Karpathy's Workflow -- AutoResearch approach
• Peter Steinberger's Workflow -- OpenClaw development approach

How to Use This Guide

1. Read through the guide like a course -- learn what commands, agents, skills,
   and hooks are before trying to use them.
2. Clone the best-practice repo and play with the examples: try /weather-orchestrator,
   listen to the hook sounds, run agent teams.
3. Go to your own project and ask Claude to suggest what best practices you should
   add, giving it the best-practice repo as a reference.

Getting Started

Head to Day 0 -- Setup to install Claude Code and authenticate.