clauding.dev

Mastering Claude Code

From Zero to Power User

Content adapted from Ado Kukic's 31-day guide

Introduction

What We'll Cover

From beginner essentials to advanced patterns — everything you need to master Claude Code.

🚀 Getting Started

Project init, memory, and @ mentions

⌨️ Shortcuts

Essential keyboard productivity

📁 Sessions

Continue, resume, and export

🧠 Thinking

Ultrathink and plan modes

🔒 Safety

Sandbox, YOLO, and hooks

🤖 Advanced

Subagents, plugins, and SDK

Getting Started

The /init Command

Claude generates project onboarding documentation automatically.

Creates a CLAUDE.md file containing:

  • Build commands and scripts
  • Directory purposes and structure
  • Code conventions and patterns
  • Architectural decisions

For Larger Projects

Create .claude/rules/ directories with modular, topic-specific instructions.

.claude/ └── rules/ ├── testing.md ├── api-conventions.md └── deployment.md
Getting Started

Memory Updates & @ Mentions

📝 Memory Updates

Ask Claude to update CLAUDE.md directly without manual editing. Maintains workflow momentum.

@ Mentions

Rapid context addition — reference files, directories, or enable MCP servers.

@ Mention Examples

@src/utils/auth.ts → Reference specific file @components/ → Reference directory @mcp:postgres → Enable MCP server

Pro tip: File suggestions are 3x faster in git repositories!

Essential Shortcuts

The ! Prefix & Double Esc

! — Instant Bash

Execute bash commands instantly without token consumption

! git status ! npm test ! ls -la

Esc Esc — Rewind

Checkpoint conversations and revert changes or code modifications

⚠️ Bash commands cannot be undone

These two shortcuts alone can dramatically speed up your workflow by letting you quickly run commands and undo mistakes.

Essential Shortcuts

Search, Stash & Suggestions

Ctrl + R
Reverse search past prompts
Ctrl + S
Stash/save prompt draft
Tab
Edit suggested prompt
Enter
Execute suggestion

Prompt Stashing

Drafts auto-save and restore — never lose mid-conversation thoughts!

Prompt Suggestions

Claude predicts follow-up questions. Press Tab to edit or Enter to execute immediately.

Essential Shortcuts

Complete Keyboard Reference

! + command
Execute bash instantly
Esc Esc
Rewind / undo
Ctrl + R
Reverse search
Ctrl + S
Stash prompt
Shift + Tab (×2)
Enter plan mode
Alt + P
Switch model
Ctrl + O
Toggle verbose mode
@
Mention files/MCP
Session Management

Continue & Resume Sessions

claude --continue

Resume your last session instantly. Context preserved.

claude --resume

Browse and select from previous sessions.

Named Sessions

Use /rename to name your sessions for easy identification.

P
Preview session
R
Rename session

Sessions persist for up to 30 days by default

Session Management

Remote & Export

🌐 Claude Code Remote

The /teleport feature transfers web-based sessions to local terminals for continuation.

📤 Export Function

/export outputs entire conversations as markdown — all prompts, responses, and tool calls.

Use Cases

  • Document debugging sessions for team knowledge
  • Create tutorials from real interactions
  • Transfer work between machines seamlessly
  • Archive important problem-solving conversations
Productivity

Vim Mode

Type /vim to enable vim-style editing for prompts.

Supported Commands

ciw
Change inner word
dd
Delete line
yy
Yank (copy) line
p
Paste
i / a
Insert / Append
Esc
Normal mode

Perfect for power users who think in vim. Edit complex prompts with muscle memory.

Productivity

Status Line & Context

/statusline

Configure terminal status bars showing git status, model info, token usage, and custom scripts.

/context

Reveals token consumption sources — see where your context is going.

Context Breakdown

  • System prompts
  • MCP server prompts
  • Memory files (CLAUDE.md)
  • Skills loaded
  • Conversation history

/stats

View your usage patterns, favorite models, and streaks!

Thinking & Planning

Ultrathink & Plan Mode

🧠 Ultrathink Mode

Include "ultrathink" in prompts to allocate up to 32k tokens for internal reasoning before responding.

ultrathink: How should I architect this complex authentication system?

📋 Plan Mode

Press Shift+Tab twice to enable planning without immediate code execution.

Review and approve before implementation!

Extended Thinking (API)

Direct API usage supports thinking blocks showing step-by-step reasoning — great for debugging Claude's thought process.

Permissions & Safety

Sandbox & YOLO Mode

🔒 Sandbox Mode

/sandbox establishes boundaries for permitted operations.

Allowed: npm test, file reading Blocked: npm publish, rm -rf

⚡ YOLO Mode

--dangerously-skip-permissions

Auto-approves ALL actions. Only use in isolated environments!

⚠️ Warning: YOLO mode bypasses all safety checks. Use responsibly in containers or VMs only.

Permissions & Safety

Lifecycle Hooks

Execute custom scripts at predetermined events.

Hook Events

PreToolUse
Before tool execution
PostToolUse
After tool execution
PermissionRequest
When permission needed

Configuration

Configure via /hooks command or .claude/settings.json

{ "hooks": { "PreToolUse": "node ./validate-action.js", "PostToolUse": "node ./log-action.js" } }
Automation & CI/CD

Headless Mode & Commands

🤖 Headless Mode

The -p flag runs Claude non-interactively.

claude -p "Fix lint errors" claude -p "Generate tests for auth.ts"

Output goes directly to stdout — perfect for pipelines!

📜 Custom Commands

Markdown files become reusable slash commands.

.claude/commands/explain.md → /explain $ARGUMENTS

Accept arguments for flexible automation.

Browser Integration

Chrome Extension

Claude directly interacts with Chrome through the extension.

Install from: claude.ai/chrome

Capabilities

🌐 Navigate Pages

Browse to URLs, follow links

🖱️ Click & Interact

Click buttons, fill forms

🔍 Read Console

Capture errors and logs

🏗️ Inspect DOM

Query and analyze elements

Great for debugging frontend issues, testing flows, and automating web tasks!

Advanced Features

Subagents

Specialized agents for parallel task execution.

🧠 200k Context Window

Each subagent gets its own massive context

⚡ Parallel Execution

Multiple tasks run simultaneously

🔄 Output Merging

Results combine back to main agent

🎯 Specialization

Focused agents for specific domains

Perfect for large refactors, multi-file changes, or when you need to research and implement simultaneously.

Advanced Features

Agent Skills & Plugins

🎓 Agent Skills

Instruction folders that teach Claude specialized tasks.

  • Reusable across projects
  • Packaged expertise
  • Open standard format

🔌 Plugins

Bundles that combine everything:

  • Commands
  • Agents
  • Skills
  • Hooks
  • MCP servers
/plugin install my-setup
Advanced Features

LSP & Agent SDK

📐 Language Server Protocol

IDE-level code intelligence:

  • Instant diagnostics
  • Code navigation
  • Type information
  • Go to definition

🛠️ Claude Agent SDK

Build your own Claude-powered agents!

~10 lines of code provides the full agent loop, tools, and context management.

The same SDK powering Claude Code is available for building custom agents.

Reference

Quick Reference Card

Shortcuts

! Bash
Esc×2 Rewind
Ctrl+R Search
Ctrl+S Stash
Alt+P Model

Commands

/init
/context
/stats
/vim
/export

CLI Flags

-p Headless
--continue
--resume
--teleport

More Commands

/config, /hooks, /sandbox, /resume, /rename, /theme, /terminal-setup, /statusline

Danger Zone

--dangerously-skip-permissions (YOLO mode)

Wrap Up

Now Go Build Something

Claude Code is a lever. These features help you find the right grip. The goal is human control — amplify your capabilities while staying in command.

🎯 Start Small

Master /init, !, and Esc×2 first

📈 Level Up

Add sessions, hooks, and subagents

Want hands-on training? We run workshops on Claude Code and agent development.
Get in touch →

Based on Ado Kukic's excellent guide

1 / 21
←→ Navigate M Menu