Claude Code Cheat Sheet

Before you can use Claude Code, you need a few things installed. Do these once, then you're set.

/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

brew install node

curl -fsSL https://claude.ai/install.sh | bash

irm https://claude.ai/install.ps1 | iex

sudo apt update && sudo apt install -y git nodejs npm

sudo snap install code --classic

curl -fsSL https://claude.ai/install.sh | bash

Start Claude Code

claude

That's it. Type claude and hit Enter. You're in a conversation.

Claude asks for permission before every action. Here's how to turn that off.

Auto Mode (Recommended)

An AI reviews each action and auto-approves safe ones. Smart guardrails, no interruptions.

claude --auto

Make It Your Default

Create a shortcut so you just type cc instead:

echo 'alias cc="claude --auto"' >> ~/.zshrc
source ~/.zshrc

Add-Content $PROFILE 'function cc { claude --auto @args }'

echo 'alias cc="claude --auto"' >> ~/.bashrc
source ~/.bashrc

Full Send: Skip All Permissions

Approves everything automatically. No guardrails. The command name is intentionally scary.

claude --dangerously-skip-permissions

Want this as your shortcut instead? Change the alias:

echo 'alias cc="claude --dangerously-skip-permissions"' >> ~/.zshrc
source ~/.zshrc

notepad $PROFILE

function cc { claude --dangerously-skip-permissions @args }

echo 'alias cc="claude --dangerously-skip-permissions"' >> ~/.bashrc
source ~/.bashrc

These are skills you can give your agent to set up its personality, memory, and capabilities. Copy the prompt, paste it into Claude Code, and let it do the work.

Read this skill and use it to create my agent's identity and soul files: https://gist.github.com/flintfromthebasement/90fedfb062a76f903dbc8596d8d8847d

Read the Docker Compose setup instructions at https://github.com/verygoodplugins/automem and help me set up AutoMem locally. Then configure it as an MCP server in my Claude Code settings.

Read this voiceprint skill and use it to create a voice profile for me: https://github.com/jamesckemp/claude-plugins/tree/master/voiceprint

Read this skill and set it up so you can watch YouTube videos: https://gist.github.com/flintfromthebasement/1dd03ea0f3f23e261dd0dd932c9a7c07

Read this skill and set it up so you can query our Google Analytics: https://gist.github.com/flintfromthebasement/4d2c098cd01de9bcc4addcb5a0955bdc

The Loop: You → Claude → You

Claude Code is a conversation. You describe what you want, Claude does the work (writes copy, builds pages, generates assets, edits files), then you review and keep going. Each round gets closer to what you need.

You don't need to get your prompt perfect. Start rough, then iterate. "Make the headline punchier." "Actually, target it at small business owners." "Add a testimonial section." That's the workflow.

Feed It Context — The More the Better

Claude works best when it knows what you're working with. Don't be shy — dump everything in:

• Paste text — briefs, brand guidelines, competitor copy, Slack threads, emails
• Drag in images — screenshots of designs, mockups, inspiration from other sites
• Give it URLs — read this page: https://example.com/about — Claude can usually fetch and read web pages
• Point it at files — read the file landing-page.html
• Point it at folders — read every file in the campaigns folder
• Copy files into your project folder — anything in the folder Claude is running from, it can see and work with

Opening a Project in VS Code

In VS Code: File → Open Folder, then pick your project. The terminal inside VS Code will automatically be in that folder.

Claude Code is designed to run from inside a single project folder. Open your project folder in VS Code, open the terminal, and run claude from there.

Since you are using an AI agent, your agent manages multiple projects from one home folder. The folder structure looks like this:

my-agent/
  CLAUDE.md          ← agent instructions
  IDENTITY.md        ← agent personality
  SOUL.md            ← agent values & behavior
  skills/            ← task recipes the agent can run
  projects/          ← individual project folders live here
    website/
    email-campaigns/
    landing-pages/

What's in Your Project Folder

You'll notice some special files and folders Claude uses:

Just type what you want in plain English. Here are some starting points:

Create Content & Assets

• Write a landing page for our spring sale with HTML and CSS
• Create a responsive email template for our product launch
• Build a pricing comparison table I can embed on our website
• Generate 10 subject line variations for our newsletter
• Write social media copy for this blog post in 3 different tones

Edit & Improve

• Read the homepage and make the copy more compelling
• This page looks broken on mobile, fix it
• Rewrite this email to be shorter and more direct
• Update all the CTAs to say "Start Free Trial" instead of "Sign Up"
• Paste an error message and say fix this

Research & Understand

• Read every file in this project and give me a summary
• What pages link to the pricing page?
• Search the codebase for where we set the meta descriptions
• Explain what this file does in plain English

Get More Out of It

• Interview me until you have 95% confidence about what I actually want, not what I think I should want. Then do the work.
• Create this page and take a screenshot of it
• What would you improve about this website?

Type these inside a Claude Code session:

Git tracks changes to your code. You don't need to memorize commands — just ask Claude.

First Time? Let Claude Set It Up

If you don't have Git set up yet, just ask:

Install git if it's not installed, then set up git in this project folder. My name is [Your Name] and my email is [you@example.com].

Quick Vocabulary

Repo

Your project folder (tracked by Git)

Commit

A save point — a snapshot of your changes

Push

Upload your commits to GitHub

Pull

Download the latest changes from GitHub

Say This to Claude

• Save my work → Claude commits your changes
• Push this to GitHub → uploads to remote
• Pull the latest changes → downloads updates
• Show me what I've changed

Working with a Team? You'll Also Use:

Branch

A separate copy to work on without affecting the main code

PR (Pull Request)

Ask the team to review and merge your branch

Main / Dev

The primary branch — the "real" version of the code

• Create a new branch called fix/header-bug
• What branch am I on?
• Create a PR for this branch
• Switch to the dev branch

If you're working solo, you can ignore branches and PRs. Just commit and push.

Your agent uses AutoMem to remember things across sessions — decisions, preferences, lessons learned, and completed work. You can talk to it directly.

Starting a Session

Open with context so the agent recalls what it knows:

• Let's work on the spring campaign project. — The agent will recall related memories automatically.
• What do you remember about our email templates?
• Recall everything you know about our brand guidelines before we start.

Ask It to Remember Something

• Remember that our brand voice is casual but professional, never corporate
• Store a memory that the spring campaign launched April 1st and performed well
• Remember that Jason prefers short subject lines under 40 characters
• Remember that we use #FF6B35 as our accent color, never red

Ask It to Recall

• What do you remember about our brand guidelines?
• Recall what we decided about the email footer last time
• What have we shipped this month?
• Do you remember the tone we used for the last newsletter?

Ending a Session

Before you close out, ask the agent to save what it learned:

• Look back over this session and store any memories that will help us in the future.
• How can we avoid this mistake in the future? Store a memory.
• Save what we decided today about the homepage layout.

Images & Screenshots

You can drag and drop images into the conversation. Drop a screenshot of a design and say build this. Drop a screenshot of a bug and say fix this.

The Interview Prompt

When you have a big idea but aren't sure how to describe it, start with: "Interview me until you have 95% confidence about what I actually want. Then do the work."