OpenAI are quietly adopting skills, now available in ChatGPT and Codex CLI

TL;DR Highlight

OpenAI quietly added reusable task units similar to Claude's Skills concept to ChatGPT and Codex CLI.

Who Should Read

Devs building with the OpenAI API, and product engineers tracking how ChatGPT's agentic capabilities are evolving.

Core Mechanics

OpenAI added a 'Tasks' feature to ChatGPT and Codex CLI that lets users define reusable, parameterizable task definitions — conceptually similar to Anthropic's Skills system for Claude.
Tasks can be saved, reused across sessions, and shared — moving ChatGPT closer to a programmable automation layer rather than a pure chat interface.
The Codex CLI integration means tasks can trigger terminal commands, file operations, and code execution as part of the task definition.
This was launched quietly without a major announcement, suggesting it's still experimental or being rolled out gradually.
The pattern of both OpenAI and Anthropic converging on reusable task primitives suggests this is becoming a standard UX pattern for AI coding tools.

Evidence

Multiple HN commenters confirmed seeing the feature in their ChatGPT interfaces, with some sharing screenshots.
Comparison to Claude's Skills was immediately drawn by the community — both allow saving prompt+context+tool configurations as named, reusable units.
Some users noted the implementation felt unpolished — limited UI for browsing saved tasks and no public API for programmatic task management yet.
Discussion of whether this competes with tools like LangChain, AutoGPT, and similar orchestration layers — the consensus was that native integration is likely more reliable.

How to Apply

If you have repetitive multi-step tasks in ChatGPT (code reviews, report generation, data transformations), try encoding them as Tasks to save setup time.
For Codex CLI power users: define Tasks for your common workflows (e.g., 'review this PR for security issues', 'refactor this module to use async/await') and parameterize the variable parts.
Track how the Tasks API evolves — once programmatic access is available, it could become a lightweight alternative to building custom LLM orchestration for simple automation.

Code Example

snippet

# Example of using skills in Codex CLI (official confirmation required)
# 1. Skill definition (expected format)
codex skill create --name "code-review" \
  --prompt "Review the following code and provide feedback in order of bugs, performance, and security"

# 2. Skill invocation
codex run --skill code-review --file src/api/handler.py

# Pattern for managing reusable system prompts like skills in the ChatGPT API
import openai

SKILLS = {
    "code-review": "You are a senior engineer. Review code for bugs, performance, and security issues in that order.",
    "doc-writer": "You are a technical writer. Write concise docstrings for the given function."
}

def run_with_skill(skill_name: str, user_content: str):
    return openai.chat.completions.create(
        model="gpt-4o",
        messages=[
            {"role": "system", "content": SKILLS[skill_name]},
            {"role": "user", "content": user_content}
        ]
    )

Terminology

Tasks (OpenAI)Reusable, parameterizable task definitions in ChatGPT/Codex CLI that save prompt templates with tool configurations for repeated use.

Skills (Anthropic)Claude's equivalent concept — saved, reusable prompt + tool + context configurations that can be invoked by name.