Navigate
←Back to Gym
← Back to Wall

Getting Started with Claude Code

Route ID: R038A • Wall: W11 • Released: Mar 9, 2026

5.8
published

🎉 Sent!

You made it to the top. Submit your work above!

Submission

Submit your work here

Deliverable

Submit screenshots or a screen recording showing Claude Code in action. Your submission might include multiple interactions, prompt refinements, or experiments — that's all great.

At minimum, your submission must have:

  1. Screenshot of successful installation and authentication
  2. Example of a good prompt that worked well (with the response)
  3. Example of refining a prompt when the first attempt wasn't quite right

Mission checklist

  • Installed Claude Code
  • Authenticated with your Anthropic account
  • Practiced good prompting patterns
  • Successfully refined a prompt through iteration

Exercise 5: Iteration and Refinement

The agent won't always get it right on the first try. Learning to refine your prompts is a core skill.

When the response isn't what you wanted

Don't start over. Instead, refine:

> [First attempt didn't work well]
> That's not quite what I meant. I want [clarification].
> Also, make sure to [additional constraint].

The agent remembers the conversation, so you can build on previous attempts.

Common refinement patterns

Too much output:

> That's way more than I need. Just give me the top 5.

Wrong interpretation:

> I meant the config file for the database, not the app config.

Missing context:

> Oh, I should mention — this is a Django project, not Flask.

Wrong format:

> Can you output that as a CSV instead of a table?

Practice exercise

  1. Ask the agent to do something slightly ambiguous
  2. When it interprets it differently than you intended, refine your request
  3. Notice how follow-up prompts can steer the conversation

This back-and-forth is normal and expected. Good prompting is iterative, not perfect on the first try.

Exercise 4: Prompt Patterns

Now that you've had a basic conversation, let's learn how to ask better questions.

Good prompts vs bad prompts

Bad (vague):

> Help me with this code

Good (specific):

> Read src/utils.py and explain what the parse_date function does

Bad (no context):

> Why isn't this working?

Good (context provided):

> I'm getting a KeyError on line 45 of data_loader.py.
> Read that file and help me understand why the key might be missing.

The context-goal-constraints pattern

A strong prompt often has three parts:

  1. Context: What are you working on?
  2. Goal: What do you want to accomplish?
  3. Constraints: Any specific requirements?

Example:

> I'm working on the user authentication module (context).
> I need to add password reset functionality (goal).
> Use the existing email service in src/services/email.py (constraint).

Specificity helps

The more specific you are, the better the results:

Instead of...Try...
"Fix this bug""The login fails when email has uppercase letters. Find and fix it."
"Make it faster""This function is slow with large files. Profile it and suggest optimizations."
"Clean up the code""Refactor this function to be under 20 lines. Extract helper functions if needed."

Practice

Try these prompts in your project:

  1. Ask a vague question, then rephrase it with more specificity
  2. Use the context-goal-constraints pattern for a real task
  3. Notice how the quality of responses improves

Exercise 3: Your First Conversation

Now let's have a real conversation with Claude Code.

  1. Navigate to any project folder (or create a test folder with a few files)
  2. Run claude to start Claude Code
  3. Try these prompts:
    • "What files are in this directory?"
    • "Read the README and summarize it"
    • "Create a simple Python script that prints hello world"

Notice how the agent can:

  • Read and understand files
  • Execute commands
  • Write and edit code
  • Answer questions about the codebase

This is fundamentally different from a chatbot — it's actually interacting with your filesystem.

Exercise 2: Clone a Repository

Let's practice the workflow of cloning a repo and exploring it with Claude Code.

  1. Open your terminal
  2. Clone any repository you're interested in: 
    git clone https://github.com/psf/requests
cd requests
  3. Start Claude Code: 
    claude
  4. Ask it: "What is this repository about? Give me a quick overview."

The agent will read files, explore the structure, and give you a summary. This is incredibly useful when you're new to a codebase.

Exercise 1: Install Claude Code

Prerequisites

  • Claude Pro subscription ($20/month) — this gives you access to Claude Code
    • Sign up at claude.ai and upgrade to Pro
    • Be upfront with yourself: if $20/month isn't feasible, check out the free alternatives in Exercise 0
  • Node.js 18+: Check with node --version. If you don't have it, install from nodejs.org

Installation

Once you have Claude Pro, open your terminal and run:

npm install -g @anthropic-ai/claude-code

Authentication

Run Claude Code for the first time:

claude

It will open a browser window to authenticate with your Claude account. Sign in with the same account you used for Pro.

Verify it works

After authentication, you should see a prompt where you can start typing. Try:

> What can you help me with?

If you see a helpful response, you're all set!

Exercise 0: Why Coding Agents?

Before we install anything, let's understand what we're getting into.

What is Claude Code?

Claude Code is an AI coding agent that runs in your terminal. Unlike a chatbot in a browser, it can:

  • Read files on your computer
  • Write and edit code directly
  • Run commands in your terminal
  • Navigate codebases and understand project structure

Why learn this?

  1. Productivity: Tasks that took hours now take minutes
  2. Learning: The agent explains code as it works
  3. Real-world skills: This is how professionals work now
  4. Data wrangling: Perfect for messy tasks like extracting scores from 50 JSON files

Alternatives

Claude Code isn't the only option. Similar tools include:

  • Cursor — VS Code fork with built-in AI
  • GitHub Copilot — AI autocomplete in your editor
  • Aider — Open-source coding agent

This route focuses on Claude Code, but the skills transfer to any tool.

Intro

You've been using chatbots to help write code. But what if the AI could actually see your files, run commands, and edit code directly?

That's what coding agents do. They're not just chat — they're collaborators that work alongside you in your actual development environment.

In this route, you'll:

  1. Install Claude Code
  2. Have your first conversation with a coding agent
  3. Learn prompt patterns that get better results
  4. Practice refining prompts when things don't work perfectly

Route 038A: Getting Started with Claude Code

  • RouteID: 038A
  • Wall: AI-Assisted Coding (W11)
  • Grade: 5.8
  • Routesetter: Adrian
  • Time: ~30-40 minutes
  • You'll need: Terminal access, Node.js 18+, Anthropic account

🧗 Base Camp

Start here and climb your way up!

← Back to WallReturn to Gym