claude-mastery-course

# 💻 Module 05 — Claude Code [![Module](https://img.shields.io/badge/05-Claude_Code-B197FC?style=for-the-badge&labelColor=2a1f1a)](#) [![Time](https://img.shields.io/badge/⏱️_30_min-FF7A4D?style=for-the-badge&labelColor=2a1f1a)](#) [![Level](https://img.shields.io/badge/🌿_Intermediate-FFD23F?style=for-the-badge&labelColor=2a1f1a)](#) [![Requires](https://img.shields.io/badge/📦_Node.js_18%2B-6BCF7F?style=for-the-badge&labelColor=2a1f1a)](#) ***Goal:** install Claude Code, run your first session, use it as a real coding partner on a real project.*

5.1 What is Claude Code?

Claude Code is a command-line agent that turns Claude into a developer who lives in your terminal.

👤 You  ──describe goal──▶  💻 Claude Code  ──reads + edits──▶  📂 Your repo
   ▲                              │
   │                              ├──runs──▶  ⚙️ Commands (tests, builds, git)
   │                              │
   └──◀────asks for approval──────┘

Unlike Claude.ai (which talks about your code), Claude Code can:

📖 Read your repository
🧠 Plan a change across multiple files
✏️ Edit files directly
⚙️ Run commands (tests, builds, linters)
🔧 Use git, install packages, debug failures
🔁 Iterate until something actually works

5.2 Installation

Prerequisites

Node.js 18+ (download)
A terminal (macOS, Linux, or Windows via WSL)
An Anthropic API key OR a Claude.ai Pro/Max subscription

Install via npm

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

📦 Official package: @anthropic-ai/claude-code on npm

First run

cd ~/projects/my-app
claude

You’ll be prompted to authenticate on first use.

5.3 The core loop

   👤 You describe a goal
            │
            ▼
   🧠 Claude plans  ◀──────────────┐
            │                      │
            ▼                      │
      Read-only or edit?           │
       │            │              │
       ▼            ▼              │
  📖 Read       📝 Propose         │
   files         edit              │
       │            │              │
       │            ▼              │
       │     👤 You approve?       │
       │      │           │        │
       │     yes          no ──▶ 💬 push back ──┘
       │      ▼
       │   ✏️ Apply
       │      │
       ▼      ▼
   💬 Report  ── Done? ──▶ ✅ Finished
                  │
                  no ──▶ (back to plan)

5.4 Your First Session

> Take a look at this repo and give me a 5-bullet summary of what it does and how it's organized.

> The README is out of date — it still references the old folder structure.
  Read the actual repo and rewrite the README's "Project Structure" section.
  Show me the diff before applying.

You’ll see Claude plan → read → propose → ask permission → apply.

5.5 The Core Commands

Command	What it does
`/help`	Show all commands
`/model`	Switch model (opus, sonnet, opusplan)
`/clear`	Reset the conversation
`/init`	Generate a `CLAUDE.md` for this repo
`/cost`	Show token usage and cost so far
`/exit`	Quit

`CLAUDE.md` — your project’s system prompt

Run /init once per repo. It creates a CLAUDE.md Claude reads at the start of every session.

# CLAUDE.md

## Project: my-app
A Next.js 14 app with TypeScript, Tailwind, Prisma.

## Conventions
- Components live in `src/components/`, one per file, named exports.
- API routes use Zod for input validation.
- Never commit to `main` directly. Branch from `develop`.

## Commands
- `npm run dev`     — dev server
- `npm test`        — Jest test suite
- `npm run lint`    — ESLint
- `npm run build`   — production build (run before claiming "done")

## Don't
- Touch `src/legacy/`
- Modify migrations without explicit instruction

5.6 Plan first, execute second

┌─────────────────────────────────────────────────────────────┐
│  ❌  WITHOUT PLAN MODE                                      │
│                                                             │
│      You: "Add soft-delete to Posts"                        │
│      Claude: *immediately starts editing files*             │
│      You: *15 minutes later*: "no wait, that's wrong way"  │
│      → mess to undo                                         │
└─────────────────────────────────────────────────────────────┘

┌─────────────────────────────────────────────────────────────┐
│  ✅  WITH PLAN MODE                                         │
│                                                             │
│      You: "Plan adding soft-delete to Posts. Don't touch    │
│            files yet."                                      │
│      Claude: *writes a 7-step plan*                         │
│      You: *review, push back on step 4*                     │
│      Claude: *revises plan*                                 │
│      You: "Approved. Execute."                              │
│      → clean implementation                                 │
└─────────────────────────────────────────────────────────────┘

Use /model opusplan for an automatic Opus-plans / Sonnet-executes hybrid that saves cost while keeping quality.

5.7 Five Practical Workflows

A — Bug fix from a stack trace

> I got this stack trace in production:
[paste]

Track down the cause. Don't change anything yet — show me where the
bug is and propose a fix.

B — Refactor across files

> We're renaming `User.email_address` to `User.email` everywhere.
Find every usage, plan the migration (including DB), walk me through
it before changing files.

C — Add a feature step-by-step

> Add a "soft delete" feature to the Posts model:
1. Add a `deleted_at` column
2. Update queries to exclude soft-deleted rows by default
3. Add a `withDeleted()` helper
4. Update tests

Implement step by step. Run tests after each step.

D — Code review your own branch

> Review the changes in the current branch (compare to develop).
Flag anything risky, anything that violates our conventions in
CLAUDE.md, and anything missing tests.

E — Onboard to a new repo

> I just inherited this codebase. Spend a few minutes exploring it,
then teach me the architecture in 10 minutes. Start with a high-level
diagram, drill into the 3 most important modules.

5.8 Permissions & Safety

Claude wants to run a command, edit a file, or make a network request
                              │
                              ▼
                  Configured as auto-allow?
                   │                    │
              YES (safe ops:       NO (everything else)
              git status, npm           │
              test, etc.)               ▼
                   │              👤 Asks you first
                   ▼                    │
              ✅ Auto-allow              ▼
                              You approve?
                              │           │
                             yes          no
                              ▼           ▼
                          ✏️ Execute   🚫 Skip & explain

[!CAUTION] Never run Claude Code on a directory you can’t afford to lose. Use git. Commit before big agentic runs. You’ll thank yourself.

5.9 Tips From Power Users

1. Always start with a plan.
   It's faster to throw away a bad plan than a bad commit.

2. Keep CLAUDE.md curated.
   Whenever you correct Claude on a convention, add it to the file.

3. Use small models for grunt work.
   Sonnet/Haiku for repetitive edits, Opus for architecture.

4. Commit often.
   After each successful step: git add . && git commit. Easy rollback.

5. Don't watch silently.
   If Claude is going down a wrong path, interrupt. You're the senior.

6. Watch the cost.
   /cost shows token usage. Stay aware if you're paying per token.

5.10 When to Use Claude Code vs. Claude.ai

Task	Best tool
🏗️ Brainstorming architecture	💬 Claude.ai
✍️ Writing the launch blog post	💬 Claude.ai
🔧 Single function in isolation	Either
🔄 Multi-file refactor	💻 Claude Code
🔍 PR review	💻 Claude Code
🐛 Debugging a flaky test suite	💻 Claude Code
🤝 Pair programming on a new feature	💻 Claude Code

A lot of pros use both: Claude.ai for thinking, Claude Code for shipping.

✅ Module 5 Checkpoint

You should now have:

Claude Code installed and authenticated
A CLAUDE.md in at least one of your real repos
Run a real change (bug fix, refactor, or feature) end-to-end

👉 Next up: Module 06 — API Development — building your own apps powered by Claude.

📚 Further reading

← Previous	🏠 Home	Next →
Module 04 — Claude.ai Features	Course README	Module 06 — API Development

This site is open source. Improve this page.