YouMind
Увійти

A Complete Lesson on "Claude Code" Even a Middle Schooler Can Understand

@kawai_design
ЯПОНСЬКА17 трав. 2026 р.
393K
754
45
3
2.2K

Коротко

This article uses a teacher-student metaphor to explain Claude Code's agentic capabilities. It covers everything from basic commands to advanced features like MCP and hooks, providing a clear learning roadmap.

Claude Code might look difficult at first glance.

Terms like Terminal, MCP, hooks, Skills, and API all come out at once, making it easy to get lost in the first hour.

In this article, I will organize Claude Code as a "lesson where a teacher teaches two middle school students."

By the end, you will understand the order in which to learn basic operations, external integrations, Skills, memory, hooks, MCP, and the API.

The header image was generated by AI. The prompt is available in this article.

Claude Code is an "Acting AI," Not Just an "Answering AI"

KAWAI - inline image

Claude Code doesn't just answer questions; it's an AI that performs tasks within your computer.

In the official docs, it is described as an agent-type development tool that can edit files, execute commands, perform git operations, and integrate with MCP.

Once you grasp this, all the functions connect.

The teacher wrote this on the blackboard:

"Chat AI is a teacher who gives advice. Claude Code is a class leader who works alongside you in the classroom."

Mina asks:

"So, is Claude Code only for people who can program?"

The teacher shakes his head.

"It first became famous as a tool for writing code. But its essence is an AI that reads folders, fixes files, and executes procedures. It can be used for writing, documents, web pages, research notes, and creating teaching materials."

Ren says:

"So instead of telling ChatGPT to 'write an essay,' I can tell Claude Code to 'look at this folder, fix it, and check it.'"

Exactly.

When teaching Claude Code, these three categories are enough to start with:

  • Read
  • Fix
  • Check

For example, if the teacher is giving a lesson on creating a web page for a school festival, they would ask Claude Code like this.

The important thing here is not to ask the AI for just an "answer." You give Claude Code a sequence of tasks.

Table of Contents

  • Claude Code is an "Acting AI," Not Just an "Answering AI"
  • Basic Operations: "Request, Confirm, Fix"
  • Memory: The Teacher's Handover Note
  • Skills: Special Move Cards
  • Hooks: The Chime and the Patrol Officer
  • MCP: The School Trip Bus
  • API and SDK: How to Use Claude Code as a Part
  • A 30-Minute Practice Menu to Start

Basic Operations: "Request, Confirm, Fix"

KAWAI - inline image

The basics of Claude Code involve a flow of requesting, confirming, and fixing.

Rather than memorizing difficult commands, you'll improve faster by learning this conversational pattern.

At first, use it with the assumption of "fixing together" rather than "completing in one shot."

The teacher gave the two students the first pattern.

Mina says with a relieved face:

"I was scared it would change everything suddenly. I just need to have it propose things first."

The teacher replies:

"Yes. It's safer to create confirmation points rather than letting Claude Code work on its own."

The basic usage consists of these five:

  • claude to start the dialogue
  • claude "request content" to pass the initial request
  • claude -p "request content" for a one-time non-interactive execution
  • /help to check available commands
  • /status to check the status

However, there's no need to teach middle schoolers everything from the start. In the first lesson, just opening claude and saying one sentence is enough.

Ren says:

"If that's all the first word needs to be, it's not scary."

The teacher continues:

"That's right. You can use Claude Code while asking it how to use it. If you don't understand, ask about the usage itself."

The goal of basic operations is not command memorization. It's for the human to decide the scope of what to leave to the AI.

Memory: The Teacher's Handover Note

KAWAI - inline image

Memory is a place to write down premises you want Claude Code to read every time.

In the official docs, it's explained as a system to use different instruction files for users or projects. In a classroom setting, it's like a handover note between teachers.

The teacher wrote CLAUDE.md on the blackboard.

Mina asks:

"What is this?"

The teacher answers:

"It's the classroom rulebook that Claude Code reads before starting work."

For a school festival project, you might write it like this.

Ren says:

"So I don't have to say this every time."

The teacher nods.

"Exactly. If you're explaining the same thing every time, put it in Memory."

Information suitable for Memory includes:

  • Project goals
  • Rules for writing style or design
  • Frequently used commands
  • Files that must not be touched

Conversely, some information is not suitable for Memory:

  • Requests for today only
  • Unconfirmed ideas
  • Passwords or API keys
  • Personal information

If teaching a middle schooler, this is how to convey it:

"Memory isn't magic to make the AI smarter. It's a notebook for rules you don't want it to forget every time."

This explanation is sufficient.

Skills: Special Move Cards

KAWAI - inline image

Skills are frequently used work procedures turned into a callable format.

In the official docs, it's explained as a system centered around SKILL.md that organizes templates, examples, scripts, and reference materials. In a classroom, these are like special move cards you pull out only when needed.

The teacher drew three cards:

  • Book Report Check
  • School Festival Page Creation
  • English Vocabulary Test Creation

Mina asks:

"How is this different from a prompt collection?"

The teacher answers:

"A prompt collection is a piece of paper you paste every time. Skills are cards you keep in Claude Code's toolbox."

The basic structure of a Skill is this:

In SKILL.md, you write when to use it, what to do, and what to read. Claude Code can automatically use a Skill that fits the conversation, or the user can call it using /skill-name.

Ren says:

"So I can turn the correction rules the teacher always uses into a Skill."

Exactly.

For a Book Report Skill, you would design it like this.

The important thing here is not to turn everything into a Skill. For a one-time request, just asking on the spot is enough.

Tasks that should be turned into Skills meet these criteria:

  • Repeated many times
  • Have a fixed procedure
  • Have judgment criteria
  • Have reference materials or templates
  • Quality tends to vary depending on the person

The teacher finally said:

"Skills don't add talent to the AI. They turn human procedures into a reusable format."

Please also use the Claude Code Dictionary site.

Hooks: The Chime and the Patrol Officer

KAWAI - inline image

Hooks are mechanisms that run automatically before or after Claude Code's work.

In the official docs, they are explained as settings to run commands or processes according to specific events like PreToolUse or PostToolUse. In a classroom, these are the chime and the patrol officer.

Mina asks:

"A chime?"

The teacher answers:

"Yes. It rings before class starts. It rings before submission. It stops you if you try to go somewhere dangerous. That's what hooks are."

For example, before Claude Code rewrites a file, it can perform a patrol like this:

  • Is it trying to touch .env?
  • Is it trying to break important configuration files?
  • Should it run tests after rewriting?
  • Should it format the text?
  • Should it leave a log?

Ren's eyes sparkle:

"So you can stop the AI when it's about to make a mistake!"

The teacher says a bit strictly:

"However, hooks are powerful tools. Since they can auto-execute shell commands, they are dangerous if added without understanding."

When explaining hooks to middle schoolers, you don't need to have them write settings at first. Just teach the roles:

Before work: Stop dangerous operations. After work: Run tests or checks. Upon completion: Check for unfinished tasks.

In practice, introduce them in this order:

  • First, check manually.
  • If the same check appears many times, make it a checklist.
  • Once the checklist is stable, turn it into a hook.

Don't automate suddenly. If you automate a check that a human doesn't understand, the failure will also be automated.

The teacher wrote on the blackboard:

"Hooks are not convenient robots; they are patrol officers who follow the rules you've set."

MCP: The School Trip Bus

KAWAI - inline image

MCP (Model Context Protocol) is a mechanism to connect Claude Code to external services and data.

In the official docs, MCP is explained as an open standard for connecting AI tools with external data sources. In a classroom, it's the school trip bus for going outside the classroom.

The teacher drew a picture of the classroom. Outside, he drew a library, the staff room, the town hall, and a museum.

Ren says:

"Can Claude Code go outside the classroom?"

The teacher answers:

"That's the image. When you connect MCP, Claude Code can access external information like GitHub, Sentry, PostgreSQL, Slack, Google Drive, and Figma."

Of course, you don't let it see everything freely. You decide the connection server, authentication, and permissions.

In Claude Code's MCP settings, you can add HTTP servers or stdio servers running locally.

Mina says worriedly:

"Connecting to the outside is a bit scary."

The teacher answers:

"That feeling is correct. MCP is convenient, but the more you connect, the more information it can touch. So, start with read-only."

The order for teaching MCP is as follows:

  • First, use only local files.
  • Next, connect to safe read-only data.
  • After that, connect to GitHub or ticket management.
  • Finally, consider writing or production data.

The value of MCP isn't that the AI becomes smarter. It's that the AI can go and get the materials it needs for work by itself.

The teacher summarizes:

"MCP is the bus that shows the AI the outside world. However, the teacher decides where it's allowed to go."

API and SDK: How to Use Claude Code as a Part

KAWAI - inline image

API (Application Programming Interface) and SDK (Software Development Kit) are ways to incorporate Claude Code into apps or automated processes.

In the official docs, the Claude Code SDK is explained as a system that can be used from CLI, TypeScript, and Python. In a classroom, it's a way to have the class leader participate in work outside the classroom.

Ren asks:

"What is an API, anyway?"

The teacher answers:

"Instead of a human talking on a screen, it's the entrance for asking Claude Code to do work from a program."

For example, normal usage looks like this. This is a human asking once from the terminal.

Using the SDK, you can call Claude Code within an app or batch process. The Python SDK is provided as claude-code-sdk, and a TypeScript SDK is also available.

Mina asks:

"Is that something beginners should do first?"

The teacher answers:

"No. It's not something you should do first."

APIs and SDKs are used at the following stages:

  • When you want to run the same task automatically every day.
  • When you want to assist with reviews or tests in CI.
  • When you want to call Claude Code from internal tools.
  • When you want to run multiple processes in parallel.
  • When you want to receive results in JSON.

For middle schoolers, it's like this:

"Asking the class leader by hand is the basis of Claude Code. Sending requests automatically from the broadcasting room is the API and SDK."

It's convenient, but using it from the start causes confusion. The order is: Basics, Memory, Skills, MCP, hooks, and then API.

A 30-Minute Practice Menu to Start

KAWAI - inline image

You don't need to learn everything about Claude Code at once.

If you narrow down what to do in the first 30 minutes, the fear disappears. When teaching middle schoolers, create small success experiences first.

The teacher finally presented a 30-minute lesson menu.

Mina says:

"I think I can do this in class!"

Ren says:

"So MCP and hooks can come later."

The teacher answers:

"Yes. Order is important. If you go straight to external integration or auto-execution, you won't know what's happening."

The first three goals are:

  • Have Claude Code explain a folder.
  • Have it fix a small file.
  • Have it explain the reason for the fix.

If you can do this much, you've grasped the essence of Claude Code. After that, proceed according to your needs:

  • If explaining the same thing every time is a hassle: Memory.
  • If doing the same task every time: Skills.
  • If you want to show external materials: MCP.
  • If you want to automate checks: hooks.
  • If you want to build it into a system: API and SDK.

Claude Code is not a tool for memorizing functions. It's a tool for developing how you hand over work.

The teacher finally said:

"People who can use AI are not those who are experts on AI. They are people who can divide work, hand over rules, and perform checks."

This is the most important conclusion when teaching Claude Code to middle schoolers.

Related Articles

[Beginners OK] Claude Code Individual Support

*Early bird spots for the first 10 people are sold out.

Claude Code Design Patterns

~Folder Structure, Skills, database, and content generation flows fully revealed~

Inquiries, individual consultations, and corporate training requests:

https://kawai-official.pages.dev/

Збереження в один клік

Використовуйте YouMind для AI-глибокого читання віральних статей

Зберігайте джерела, ставте цілеспрямовані запитання, підсумовуйте аргументи та перетворюйте віральні статті на корисні нотатки в одному AI-робочому просторі.

Дослідити YouMind
Для авторів

Перетворіть свій Markdown на охайну статтю для 𝕏

Коли ви публікуєте власні лонгріди, зображення, таблиці та блоки коду роблять форматування в 𝕏 складним. YouMind перетворює повну чернетку в Markdown на чисту статтю для 𝕏, готову до публікації.

Спробувати Markdown для 𝕏

Більше патернів для аналізу

Останні віральні статті

Переглянути більше віральних статей