Codecademy Logo

Introduction to Claude Code and AI-Assisted Development

Related learning

  • Course

    AI-Assisted Development with Claude Code

    Learn to build maintainable projects with AI using Claude Code in this course on agentic coding, specs, TDD, and context engineering.
    • With Certificate
    • Beginner Friendly.
      2 hours

Claude Initialization

A Claude Code session begins by running claude from a project directory in the terminal. From there, it works directly in the workspace, ready to read files, propose edits, and run commands on the developer’s behalf.

CLAUDE.md Overview

CLAUDE.md is a Markdown configuration file Claude Code reads at the start of each session, giving projects persistent instructions across conversations. It can define communication preferences, coding conventions, project rules, and useful commands such as how to run tests.

Its location determines its scope: ~/.claude/CLAUDE.md applies globally, while ./CLAUDE.md (or one in a parent directory) applies to a specific project. A well-configured file lets Claude start each session already aligned with the project, with no setup needed.

# CLAUDE.md


## Communication
- Explain your changes before editing.


## Conventions
- Use descriptive variable names.


## Commands
- Run tests with `npm test`.

Claude /init Command

The /init command analyzes a project and generates a starter CLAUDE.md file. Rather than writing project configuration from scratch, a developer can run /init and let Claude Code inspect the codebase to produce a first draft.

That file is only a starting point — a useful next step is to ask Claude to interview the developer about their coding style and preferences, then expand the file from those answers.

‘!’ In Claude Code

In Claude Code, prefixing an entry with ! runs that shell command inline, directly in the session. Unlike asking Claude to run something — which it may propose and wait for approval — ! executes the command immediately, without routing it through Claude. That makes it handy for quick checks like running a file or viewing output, without interrupting the flow of the conversation.

! node hello.js

File Context

File context in Claude Code is provided by mentioning files in a prompt, most directly with the @ prefix. Rather than leaving Claude to search the whole workspace, @filename passes exact file context up front, and Claude reads the referenced file using its built-in file tools. Pointing Claude at the right files keeps its attention focused and its responses grounded in the actual code.

Look at @hello.js and explain what this file does.

The Next Workflow

A “next” workflow in CLAUDE.md defines a repeatable process: find the first unchecked item in a spec file, implement only that item, mark it complete, and stop for review. Captured once in CLAUDE.md, it lets a developer advance through an entire feature list with a single instruction, “do the next task”, instead of re-explaining the steps each session.

spec.md tells Claude what to build; the “next” workflow tells it how to move through the work, keeping sessions focused and progress visible.

## Next Task Workflow


When I ask you to do the next task:
1. Read `spec.md`.
2. Find the first unchecked checklist item.
3. Implement only that item.
4. Update `spec.md` by marking that item complete.
5. Summarize what changed.
6. Wait for review before continuing.

Claude Code

Claude Code is a terminal-based AI coding agent built by Anthropic. Rather than living in a browser or returning suggestions to copy-paste, it operates directly in the developer’s environment — reading project files, proposing and making edits, and running commands on their behalf. Thinking of it as a pair programmer working in the workspace, rather than a chatbot, explains much of how it’s designed to be used.

Claude Code Skills

A Claude skill is a specialized, reusable workflow that Claude can invoke on demand — think of it as a reusable function for Claude. Claude Code ships with built-in skills like /review, which runs a structured code-review workflow, and developers can create their own as well. Either way, a skill is invoked directly with a / prefix in the prompt, making a packaged workflow callable in a single step.

/review

Configuring Skills

Custom skills can be configured at either the project or user level. Each skill lives in a .claude/skills/<name>/ folder built around a SKILL.md file with two parts: YAML metadata telling Claude when the skill applies, and instructions describing the process to follow. A project-level location scopes the skill to that project, while a user-level location makes it available across all projects.

.claude/skills/todo-review/SKILL.md


---
name: todo-review
description: Review JavaScript files that contain TODO comments.
---


# TODO Review
1. Count the TODO comments in the file.
2. Explain what each one is asking for.
3. Suggest one small next step.

Claude Code Interactions

Claude Code interacts with the filesystem and shell through discrete tools — Read, Edit, Write, and Bash — rather than just generating text for the developer to apply. These take concrete, observable actions on real files and commands in the workspace. In the default mode, Claude shows each action before carrying it out and waits for approval, keeping the developer in control of what actually changes.

Learn more on Codecademy

  • Course

    AI-Assisted Development with Claude Code

    Learn to build maintainable projects with AI using Claude Code in this course on agentic coding, specs, TDD, and context engineering.
    • With Certificate
    • Beginner Friendly.
      2 hours
Previous