Claude Code 终端AI编程入门：技能、钩子与子智能体实战_CG群

本课程带你掌握Claude Code终端AI开发：从安装、会话管理与权限配置入门，深入技能、Hooks、MCP服务器和子智能体的自动化工作流，打造安全高效的AI协同开发环境。

原始标题：Claude Code Essential: Build, Automate & Work with AI Agents

Claude Code Essential: Build, Automate & Work with AI Agents

本课程由 Dominic Ferraro 于 2026年10月 推出，是一门时长 2 小时 14 分钟（共 31 节课）、大小为 1.2 GB 的 Claude Code 终端 AI 辅助开发初学者视频教程。课程旨在帮助具备基础编程经验并拥有 Claude Pro 订阅的开发者，跳出简单的单次代码生成模式，全面掌握在终端中利用 AI 实现自动化编程的高阶工作流。内容从 Claude Code 的安装、会话管理和用量监控等基础操作出发，深入讲解了如何精细化配置项目上下文（添加文件/图片）与工具权限，并重点聚焦于高级自动化特性——包括创建可复用技能与自定义命令、利用生命周期钩子（Hooks）自动化开发动作、通过 MCP 服务器 连接外部系统，以及构建和并行运行专注于特定任务的子智能体（Subagents）与前端/文档扩展插件，从而在本地终端打造一个安全、可控且高效的 AI 协同开发环境。

Set up context and permissions, then automate coding with skills, hooks, subagents, MCP servers and plugins

What you’ll learn
⚡ Install Claude Code, start sessions, navigate previous sessions, and monitor usage
⚡ Configure project context and add files or images when Claude needs additional information
⚡ Manage Claude Code tools and permissions more deliberately
⚡ Create reusable skills and expose common development workflows as commands
⚡ Automate development actions with hooks and Claude Code lifecycle events
⚡ Connect Claude Code to external systems through MCP servers
⚡ Create specialized subagents and run focused agents in parallel
⚡ Extend Claude Code with plugins for documentation and frontend workflows

Requirements
❗ Basic coding experience is required
❗ A Claude Pro subscription
❗ A computer where you can install and run Claude Code

Description
Claude Code brings AI-assisted development directly into your terminal, but getting useful results consistently requires more than simply asking it to write code. You need to understand how sessions, context, permissions, skills, hooks, MCP servers, subagents, and plugins fit together into a reliable developer workflow.

This course gives developers a practical, focused introduction to Claude Code and the features that make it useful beyond one-off code generation.

You will start from the foundations: what Claude Code is, how to install it, how to open your first session, how to move between sessions, and how to monitor usage. From there, the course builds toward more capable workflows where Claude Code can work with project context, reusable instructions, automated actions, external tools, specialized subagents, and plugins.

The course is designed for developers who already have basic coding experience but are new to Claude Code or have only used it casually. You do not need to be an expert in a particular framework or programming language. The concepts are presented as developer-tooling skills that you can transfer to your own projects and technology stack.

One of the first challenges when working with an AI coding agent is giving it the right context. Claude Code can only make good decisions when it understands the project, the relevant files, and the constraints you want it to follow. You will learn how to configure persistent project context, add files and images to a session, inspect the current context, and rewind a session when you need to return to an earlier point.

You will also learn how tools and permissions affect what Claude Code can do. This is important because useful automation needs boundaries. The course shows how permissions work so that you can decide what Claude may run or change instead of treating every tool call the same way.

After the core session and context workflow, the course moves into skills.

Skills give Claude Code reusable instructions for common development tasks. You will learn what a skill is, how to create a component-building skill, how to expose a skill as a slash command, and how to create a workflow for staging and committing changes. The aim is to move repeated instructions out of ad-hoc prompts and into reusable development workflows.

Hooks add another layer of automation.

You will learn what hooks are, how they connect to Claude Code lifecycle events, and how they can trigger deterministic actions around Claude’s work. The course demonstrates practical examples such as running Prettier automatically after edits, blocking destructive commands before they execute, and setting up notifications when Claude needs your attention.

These examples illustrate an important difference between asking Claude to remember a rule and enforcing a rule through automation. Hooks can make specific workflow behavior happen consistently rather than depending only on the model to remember what you asked earlier in the session.

The course then introduces the Model Context Protocol, or MCP.

MCP servers allow Claude Code to work with external systems and services through a structured tool interface. You will learn what MCP is and then apply the concept by creating a Supabase project and setting up a Supabase client through Claude Code. This gives you a concrete example of connecting an AI coding workflow to tools outside the local codebase.

From there, you will move into subagents.

Subagents let you delegate focused tasks to specialized agents instead of asking one session to handle everything in the same context. You will learn what a subagent is, create a code-review subagent, and run multiple subagents in parallel. This helps you understand how to divide work into focused roles when a project benefits from separate review or analysis tasks.

The final part of the course covers plugins and capability extension.

You will learn what a Claude Code plugin is and how plugins can extend the environment with additional tools and workflows. The course demonstrates using a plugin to pull current documentation into Claude and another workflow for creating frontend interfaces. These examples show how Claude Code can be adapted to different development needs rather than remaining a fixed terminal assistant.

Across the course, the emphasis is on practical developer workflows rather than abstract AI theory. You will see how the major Claude Code capabilities relate to common software-development tasks: configuring a project, controlling context, managing permissions, creating repeatable commands, automating checks, connecting external systems, delegating work, and extending the tool with plugins.

By the end of the course, you should be able to set up Claude Code for a project, manage sessions and usage, configure project context, add files and images when relevant, inspect and control context, and work with Claude’s tools more deliberately.

You should also be able to create reusable skills, automate development behavior with hooks, connect external services through MCP servers, create specialized subagents, run subagents in parallel, and extend Claude Code through plugins.

The result is a more structured way to use AI-assisted coding. Instead of repeatedly explaining the same task or manually performing every supporting step around Claude’s edits, you can begin turning your development practices into reusable instructions and automated workflows.

This course is not intended to teach programming from scratch. Basic coding experience is required because the examples assume you already understand the purpose of common development tasks such as editing code, reviewing changes, formatting files, committing work, and working with a project repository.

It is especially useful for developers who want to go beyond basic chat-style prompting and understand the actual Claude Code feature set. If you have tried AI coding tools but have not yet built a systematic workflow around context, permissions, reusable skills, hooks, MCP, subagents, or plugins, the course gives you a concise path through those capabilities.

You will also work with concrete supporting tools and services during the demonstrations, including GitHub, Prettier, Supabase, and plugin-based documentation and frontend workflows. These are used to demonstrate how Claude Code fits into real developer tooling rather than to lock the course to one specific application stack.

The central idea is simple: Claude Code becomes more useful when you understand how to shape the environment around it.

Good context helps it understand the project. Permissions control what it may do. Skills make repeated instructions reusable. Hooks automate deterministic actions. MCP connects outside systems. Subagents divide work into focused roles. Plugins extend what the environment can do.

Once those pieces make sense together, Claude Code becomes much more than a place to type a coding request. It becomes a configurable development tool that you can adapt to the way you actually work.

Who this course is for
⭐ Developers who are new to Claude Code and want a structured introduction beyond basic prompting
⭐ Software engineers who want to automate repetitive development tasks with skills and hooks
⭐ Developers who want to connect Claude Code to external tools and services through MCP
⭐ Programmers interested in delegating focused work to subagents and parallel agent workflows
⭐ Developers already using AI coding tools who want to understand Claude Code permissions, context, and plugin-based extensions

隐藏内容

此处内容需要权限查看