codex

Review implementation plans via iterative OpenAI Codex CLI debate rounds.

1|Updated Mar 24, 2026

Search Tags:#automation #codex #code-review #plan-review #model-management #feature-planning #collaborative-debate

Ready to run

npx skills add https://github.com/nawwwal/dex --skill codex-nawwwal

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: codex
Source: https://github.com/nawwwal/dex/tree/main/plugins/tools/skills/codex
Command: npx skills add https://github.com/nawwwal/dex --skill codex-nawwwal

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Send the current implementation plan to OpenAI Codex for collaborative debate. Claude evaluates each concern, accepts valid feedback, contests debatable points with counter-arguments, and converges only when both sides agree. Max 5 rounds.

Core Features & Use Cases

Collaborative plan review with OpenAI Codex CLI
Iterative rounds with verdicts and plan revisions
Structured criteria for correctness, risks, missing steps, alternatives, and security

Quick Start

Prepare the implementation plan and initiate Round 1 of Codex review using the Codex CLI.

Frequently Asked Questions about codex

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

FAQPage Schema

How do I use collaborative debate to review an implementation plan?▼

Collaborative plan review sends your implementation plan to OpenAI Codex CLI for iterative debate, evaluating correctness, risks, and security to converge on a refined plan within five rounds.

Can I use Codex CLI for automated code and plan review in read-only mode?▼

Yes, Codex CLI supports automated plan review in read-only sandbox mode to safeguard against writes, applying strict verdict flow and plan updates for multi-service coordination and data migrations.

What is the maximum number of rounds for iterative plan review with OpenAI Codex?▼

Iterative plan review with OpenAI Codex runs for a maximum of five rounds, applying structured criteria for missing steps and alternatives until both sides agree on the final plan.

Do I need a specific model to run Codex CLI for feature planning and security reviews?▼

You need the Codex CLI and a specified model to run feature planning and security reviews, ensuring iterative rounds evaluate concerns and contest debatable points with counter-arguments.

When should I use an automated plan review instead of manual code review?▼

Use automated plan review for complex implementation plans involving authentication, data models, and multi-service coordination, where structured criteria and iterative debate can catch missing steps and security risks.

Why does Codex CLI require read-only sandbox mode for plan updates?▼

Read-only sandbox mode is required to safeguard against unauthorized writes during collaborative debate, ensuring the tool only applies strict verdict flow and plan updates without modifying the underlying codebase.