codex-cli

Delegate coding and debugging tasks to OpenAI Codex CLI.

850|114|Updated Mar 9, 2026
One-click install
npx skills add https://github.com/wentorai/Research-Claw --skill codex-cli-wentorai
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: codex-cli
Source: https://github.com/wentorai/Research-Claw/tree/main/skills/codex-cli
Command: npx skills add https://github.com/wentorai/Research-Claw --skill codex-cli-wentorai

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

It helps you offload multi-file, iterative coding and debugging work when you need higher-quality implementation than a single prompt can reliably produce.

Core Features & Use Cases

  • Complex project delegation: Use OpenAI Codex CLI to read, edit, and run code locally to implement features end-to-end.
  • Non-interactive execution for automation: Run codex exec to integrate coding tasks into a larger workflow with consistent invocation.
  • Controlled safety and scope: Choose approval modes and sandbox policies to limit risk during autonomous changes.

Use case example: You have a TypeScript service that fails multiple tests; delegate the fix and refactor work to codex exec in a workspace-write sandbox, then save the final results to a file.

Quick Start

Delegate a feature implementation by asking: codex exec --full-auto -C /path/to/project "Implement the feature described in TODO.md".

Frequently Asked Questions about codex-cli

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I delegate multi-file refactoring to an AI agent?

You can delegate multi-file refactoring to an AI agent by using codex exec with a workspace-write sandbox policy, allowing the agent to read, edit, and run code locally across your project.

Can I automate complex coding tasks non-interactively?

Yes, you can automate complex coding tasks non-interactively by running codex exec. This enables consistent invocation to integrate coding workflows into larger automated systems without manual intervention.

How do I control safety and scope during autonomous code generation?

You control safety and scope during autonomous code generation by selecting specific approval modes and sandbox write policies. This limits risk by restricting execution permissions and controlling workspace outputs.

Does codex exec work for iterative debugging of failing tests?

Yes, codex exec works for iterative debugging of failing tests. You can delegate the fix and refactor work to the agent in a workspace-write sandbox to resolve multiple test failures end-to-end.

What is the best way to implement a feature described in a markdown file?

The best way to implement a feature described in a markdown file is running codex exec --full-auto with the project path, directing the AI agent to read the file and execute the implementation locally.

When should I not use autonomous coding agents for local development?

You should not use autonomous coding agents without configuring approval modes and sandbox policies. Unrestricted workspace write access during automated code generation can introduce unreviewed changes and risk project stability.