"OpenAI has introduced Skills in Codex, a new way to package coding knowledge and workflows into reusable modules that AI agents can reliably use across projects."
OpenAI is quietly changing how AI helps developers write and maintain code. With the launch of Skills in Codex, coding agents are no longer limited to generic answers. Instead, they can follow structured, reusable workflows that feel much closer to how real engineering teams work.
If you have ever wished your AI coding assistant understood your project rules, tools, and habits, Codex Skills are designed to solve exactly that problem.
What are Skills in OpenAI Codex?
A Skill in Codex is a reusable package of instructions, references, and optional scripts that teaches the AI how to perform a specific task. Think of it as a mini playbook for the coding agent.
- Each skill lives in its own folder.
- The main file is
SKILL.md, which explains what the skill does and when to use it. - Optional files can include documentation, templates, or automation scripts.
Instead of explaining the same process again and again, developers can create a skill once and reuse it across projects and teams.
Why did OpenAI introduce Codex Skills?
Traditional AI coding assistants are powerful but inconsistent. They may forget project rules, ignore internal standards, or suggest tools your team does not use.
Codex Skills solve this by making workflows explicit and repeatable. When a task matches a skill, the AI follows the documented process instead of guessing.
Skills turn Codex from a general helper into a team specific coding assistant.
How are Codex Skills structured?
Every skill is built around a clear and simple structure:
- SKILL.md: Contains metadata and step by step instructions.
- references/: Extra docs or examples the AI can consult.
- scripts/: CLI tools or automation that Codex can run.
- assets/: Templates or configuration files.
The AI only loads what it needs when it needs it. This approach keeps conversations clean and avoids flooding the context with unnecessary files.
Example skill use cases
- Refactoring a legacy service
- Setting up CI pipelines
- Running database migrations
- Applying internal code review rules
How does Codex find and use skills?
Codex uses a built in Discovery process to decide which skills are relevant.
- If you name a skill directly, Codex must use it.
- If your request clearly matches a skill description, Codex routes the task through that skill.
Skills can be global, available across all projects, or project specific, discovered through repository documentation.
Codex Skills vs AWS Kiro Powers
OpenAI is not alone in this space. AWS recently introduced Kiro Powers, which also package expertise for AI agents. However, the philosophy is different.
| Feature | Codex Skills | AWS Kiro Powers |
|---|---|---|
| Main focus | Reusable coding workflows | Deep domain expertise |
| Standard | Agent Skills open standard | Model Context Protocol |
| Best for | Daily developer automation | Enterprise cloud and database work |
In short, Codex Skills feel more developer friendly, while Kiro Powers target enterprise scale systems.
Why the open Agent Skills standard matters
One of the most important parts of this launch is that Codex Skills follow the Agent Skills open standard published by Anthropic.
This means the same skill can potentially work across different AI tools like Claude, GitHub Copilot, Cursor, and others with minimal changes.
For developers, this reduces lock in and makes skills a long term investment instead of a one platform trick.
Is there any referral or affiliate program?
At the moment, there is no official referral or affiliate program for Codex Skills or GPT 5.2 Codex. However, publishers can still track traffic using UTM links or monetize related tools like IDEs, cloud platforms, and developer services.
Frequently Asked Questions
Do I need advanced skills to create a Codex Skill?
No. If you can write clear instructions in Markdown and organize files, you can create a useful skill.
Can Codex Skills run real commands?
Yes. Skills can include scripts that Codex can execute, making automation much more powerful.
Are Codex Skills limited to OpenAI tools?
No. Because they follow an open standard, many skills can work across different AI agents.
Final thoughts
Codex Skills represent a big step forward for AI assisted development. Instead of one off answers, we now have repeatable workflows that mirror how real teams operate.
If this approach catches on, the future of coding with AI will look less like chatting and more like collaborating with a well trained teammate.
