Module 04 · 30 min

Power Features

The features that turn Claude Code from a helpful assistant into a genuinely powerful one — explained without the jargon.

Module 4 of 5

What you'll learn

Slash commands — quick shortcuts you already started using
Project memory (CLAUDE.md) — so you don't repeat yourself
Subagents — sending helpers off to do focused work
MCP tools and hooks — connecting Claude Code to the wider world

Don't feel you must use all of these

These are options, not requirements. Many people are productive with just the basics from Modules 1–3. Skim this module to know what exists, then come back and adopt one feature at a time as you need it.

Slash commands — built-in shortcuts

Typing / at the prompt brings up commands. You've already met /help, /clear, and /init. A few more worth knowing:

/model

Switch the AI model. Use a faster model for simple edits, a more capable one (like Opus) for hard reasoning. Faster models are cheaper and quicker; more capable ones think harder.

/review

Ask Claude Code to review changes — great for a second set of eyes before you finalize.

/config

Open settings — theme, model default, and more.

Custom commands

You can save your own reusable prompts as commands (files in a .claude/commands folder). A "/standup" that summarizes today's changes, for example.

Project memory — the CLAUDE.md file

Repeating "we use teal #0D9488, don't add new libraries, match the existing style" every session gets old. CLAUDE.md is a plain-text file in your project where Claude Code remembers those standing instructions. It reads this file automatically at the start of every session.

Create one with the /init command, or just ask Claude Code to add a note to it. There are two levels:

Project memory — a CLAUDE.md in your project folder, shared with anyone working on that project.
Personal memory — a global file (in your home .claude folder) for your own preferences across every project.

claude — saving a memory

> # Always explain your plan before editing more than one file

Claude Code: Saved to CLAUDE.md. I'll follow this from now on.

Starting a line with #

In a session, beginning your message with # tells Claude Code to remember this — it adds the note to your memory file. A fast way to teach it your preferences as you go.

Subagents — helpers for focused work

A subagent is a separate helper Claude Code can send off to handle a specific job — like searching a large codebase or reviewing security — without cluttering your main conversation. The helper does its work quietly and reports back just the answer.

Why it matters for beginners: it keeps the main session focused and is great for big, messy searches ("find everywhere we mention the old logo"). You can use Anthropic's built-in helpers or, later, define your own specialists.

MCP — connecting Claude Code to other tools

MCP (Model Context Protocol) is a standard way to plug Claude Code into outside services — your design files, a database, a project tracker, documentation. Once connected, you can say "pull the latest tickets and summarize them" and Claude Code uses that tool to do it.

Security note

Connecting external tools means giving Claude Code more reach. Only connect services you trust, understand what each one can access, and be aware that content coming from a tool (a web page, a document) could contain instructions trying to mislead the AI — a trick called prompt injection. We cover staying safe with this in Module 5 and in our AI tool safety guide.

Hooks — automatic actions

Hooks let you run your own command automatically at certain moments — for example, "always run the code formatter after Claude Code edits a file," or "block edits to this protected file." They're set up in a settings file. This is a more advanced feature; just know it exists for when you want guardrails or automation that happen every time, without asking.

Worktrees — work on two things at once

For users comfortable with git, Claude Code can use a worktree — an isolated copy of your project — so it can work on a change without touching your main files until you're ready. (This very course was built in a worktree!) Beginner takeaway: it's a safety and parallelism feature you'll appreciate later.

Try it yourself

In your practice folder, run /init, then start a message with # to save a preference like "# keep explanations short and beginner-friendly." Open the CLAUDE.md file it created and read what's inside. Then type /model to see which models you can switch between.

Knowledge check

What is the CLAUDE.md file used for?

Correct: CLAUDE.md is your project's memory — coding standards, preferences, and context you don't want to retype. Claude Code loads it automatically so it starts each session already knowing your rules.

Key takeaways

Slash commands (/help, /model, /review, /clear) are quick shortcuts — and you can make your own.
CLAUDE.md remembers your standing instructions; start a line with # to add to it.
Subagents handle focused side-jobs; MCP connects outside tools; hooks automate actions.
Adopt these one at a time — you don't need all of them to be productive.

Finished this module?