inflearn logo

2026 Mastery of the Latest Codex: From the Basics to Agentic Development Automation

Are you simply telling Codex, “Build it for me”? It works well at first. But as the project grows, the AI modifies the wrong files, forgets the rules, and repeats the same mistakes. In this course, you’ll learn how to use the **latest Codex** to delegate coding while keeping control over **design, permissions, context, and verification** as a developer. **Design → Plan → Implement → Test → GitHub PR** Based on the developer’s judgment and review, AI development automation is applied throughout this entire process.

1 learners are taking this course

Level Basic

Course period Unlimited

Python
Python
codex
codex
AI
AI
ChatGPT
ChatGPT
AI Agent
AI Agent
Python
Python
codex
codex
AI
AI
ChatGPT
ChatGPT
AI Agent
AI Agent

What you will gain after the course

  • You can systematically use core features, from installing Codex and configuring permissions to AGENTS.md, Plan Mode, and context management.

  • You can automate repetitive development tasks and extend Codex’s capabilities using MCP, subagents, skills, and hooks.

  • You can improve the stability and quality of the output by validating AI-generated code through testing and feedback loops.

  • Go beyond simple vibe coding and build an agentic development workflow where developers retain control while collaborating with AI.

Delegate coding to AI while keeping developers in control

Learn how to use Codex not merely as a code generation tool, but as an AI development agent that works alongside you on real-world development.

You will follow the real development flow from AGENTS.md → Plan Mode → implementation → testing → GitHub PR,
and learn how to automate repetitive tasks using Skills · Hooks · MCP · Subagents.

Go beyond simply asking, “Codex, build it for me,”
and complete a systematic AI development workflow of SPEC → PLAN → BUILD → TEST → REVIEW → PR .

  • Codex Installation, Permissions, Sandbox, and Context Management

  • Designing project rules using AGENTS.md and SPEC documents

  • Reviewing implementation plans using Plan Mode

  • Hands-on development based on tests and GitHub Issues/PRs

  • Development automation using Skills · Hooks · MCP · and Subagents

  • Step-by-step practice through 3 hands-on projects

Recommended for people like these:

Those who have tried Codex but aren’t using it effectively

You’ve tried asking Codex to code in natural language, but...

  • Modify the wrong files or

  • Forgetting the project rules or

  • you have to repeat the same explanations over and over,

  • Because you have to check the output again and again by hand

    If you’ve felt, "It’s convenient, but I’m uneasy about using it in a real project," this course will help.

Developers who want to apply AI coding to their actual development workflow

Not just simple code generation,

Writing requirements → Reviewing the plan → Implementation → Testing → Fixing errors → GitHub PR

It is suitable for those who want to learn how to carry out the entire process with Codex.

You will also learn practical ways to analyze and modify existing codebases, not just work on personal projects.

Those who want to systematically learn even Codex’s advanced features

Beyond AGENTS.md and Plan Mode

Skills · Hooks · Plugins · MCP · Subagents · Worktree

It is recommended for those who want to automate repetitive development tasks using extension features such as Claude Code and take their AI development environment to the next level.

After completing the course,

  • Upon completing the course, you will go beyond simply having Codex write code and be able to design and control how the AI works.

    • For each project, you can write an AGENTS.md file and a SPEC to communicate development rules to the AI.

    • Before having it code right away, you can review and modify the implementation plan in Plan Mode.

    • You can configure Codex's permissions and sandbox to have it work safely.

    • You can analyze an existing codebase to find, fix, and test bugs.

    • You can establish a development workflow that analyzes GitHub Issues, modifies the code, and connects it all the way to a PR.

    • You can use Skills, Hooks, and MCP to automate repetitive tasks.

    • Using Subagents, you can assign tasks to multiple AI agents by dividing up their roles.


    SPEC → PLAN → BUILD → TEST → REVIEW → PR

    You’ll be able to apply this consistent Codex development process to your own projects.

What you’ll learn

A Practical Development Workflow Completed with Codex

From the initial installation and use of Codex, you will follow the actual development workflow and learn everything through AGENTS.md, Plan Mode, Sandbox, context management, implementation, testing, and GitHub PRs.

Automate Development with Skills · Hooks · MCP · Subagents

Beyond the basics of Claude Code, learn how to automate repetitive development tasks using advanced features such as Skills, Hooks, MCP, Subagents, and GitHub.

The goal is to create a development environment where AI goes beyond simply writing code, automating repetitive tasks so developers can focus on design, decision-making, and validation.

I did not want to create a course that merely teaches how to use Codex.

A course could be created simply by explaining Codex’s commands and features one by one.

However, the features keep changing.

What I considered more important than the features themselves was how to develop with AI.

That’s why this course explains everything around one consistent development workflow.

SPEC → PLAN → BUILD → TEST → REVIEW → PR

Organize the requirements,
have Codex create an implementation plan,
have the developer review the plan and then implement it,
and go through testing and review before connecting it all the way to a GitHub PR.

And to further develop this workflow

AGENTS.md, context management, skills, hooks, MCP, subagents, GitHub

We cover the same functionality within an actual development process.

Do you have any questions?

Write at least three questions and answers that prospective students may have before enrolling.
Rather than predictable or formulaic answers, we recommend responses that reflect the knowledge sharer's personality.

Q1. Can I take the course even if I’m new to Codex?

Yes. We explain everything step by step, starting with installation and basic usage, so even those new to Codex can follow along.

Start by learning the basic workflow while building a simple project with Claude Code, then expand to AGENTS.md, Plan Mode, Sandbox, context management, GitHub, and test automation.

However, having a basic understanding of Python or web development will help you better understand the exercises.

Q2. How much will I be able to do with Codex after taking this course?

Beyond simply asking, “Please write the code,” you will be able to work through the entire flow of a real development project with Codex.

In the course,

Requirements definition → planning → implementation → testing → code review → GitHub PR

You will practice the process hands-on.

Additionally, it covers how to automate repetitive tasks using Skills, Hooks, MCP, and Subagents, as well as how to divide multiple tasks among AI agents.

In other words, the goal is to use Codex not merely as a code generation tool, but as an AI development agent that performs real-world development tasks.

Q3. How is this different from other introductory Codex courses?

This course goes beyond simply introducing Codex’s commands and features one by one.

The most important focus is not “how to get AI to code,” but “how to design and control the way AI works”.

So, the entire course is structured around

SPEC → PLAN → BUILD → TEST → REVIEW → PR

is structured around the actual development workflow of SPEC → PLAN → BUILD → TEST → REVIEW → PR.

And through three projects, you will experience not only creating new projects but also analyzing existing code, fixing bugs, testing, and handling GitHub Issues and PRs.

By the end of the course, you will learn how to apply Codex reliably to real-world projects, which is more important than simply knowing many of its features.

Notes Before Taking the Course

Practice Environment

  • You can do the exercises in this course on Windows, macOS, and Linux. Since the course focuses on Codex's features and development workflows, you do not need a high-performance GPU or a separate AI server.

Learning Materials

  • It is distributed as a PDF before each chapter begins.

  • The practice code can be downloaded from GitHub.

Prerequisite knowledge

  • This course provides step-by-step explanations starting with installation and basic usage so that even those using Codex for the first time can follow along.

    However, having the following basic knowledge will make it much easier to follow along with the hands-on exercises.

    • Some experience with basic use of a terminal or command prompt

    • A basic understanding of file and folder structures

    • Basic concepts of Git

    • A basic understanding of Python or web development code

    Advanced programming skills or specialized knowledge of AI/LLMs is not required.

Recommended for
these people

Who is this course right for?

  • A developer who has installed Codex but only uses it for simple questions and code generation, and feels at a loss about how to apply it to actual projects

  • A developer who spends more time reviewing and fixing the results than developing because AI modifies the wrong files or generates incorrect code.

  • Developers who have heard of features such as MCP, sub-agents, skills, and hooks, but don’t know when or how to use them and therefore can’t use them properly.

  • A developer who wants to switch from Claude Code to Codex

Need to know before starting?

  • Basic knowledge of Python and HTML

  • Paid ChatGPT account

Hello
This is YoungJea Oh

Career Verified

5,297

Learners

491

Reviews

160

Answers

4.8

Rating

22

Courses

I am an AI specialist instructor teaching Artificial Intelligence and Python based on over 30 years of IT field experience. I develop and teach practical-oriented curricula, including deep learning, NLP, LLM fine-tuning, LangChain/LangGraph-based AI agents, and AI-driven development (AIDD). I translate more than 30 years of development and operations experience gained at Hyundai Engineering & Construction's IT department, Samsung SDS, and Citibank Korea into realistic, hands-on lectures. Currently, I am conducting AI courses at institutions such as KOSA, KOSTA, and KITRI.

Homepage Address: https://ironmanciti.github.io/

More

Curriculum

All

63 lectures ∙ (6hr 10min)

Course Materials:

Lecture resources
Published: 
Last updated: 

Reviews

Not enough reviews.
Please write a valuable review that helps everyone!

YoungJea Oh's other courses

Check out other courses by the instructor!

Similar courses

Explore other courses in the same field!

Limited time deal

$38.50

50%

$77.00