loop-codex-review

Coordinate parallel Codex code reviews with human approval before commit.

1|1|Updated Jan 23, 2026

Ready to run

npx skills add https://github.com/corygabrielsen/skills --skill loop-codex-review

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: loop-codex-review
Source: https://github.com/corygabrielsen/skills/tree/main/loop-codex-review
Command: npx skills add https://github.com/corygabrielsen/skills --skill loop-codex-review

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This skill streamlines rigorous code reviews by coordinating parallel Codex analyses and human oversight to reach a fixed point with minimal ambiguity.

Core Features & Use Cases

Parallel Codex reviews (configurable via -n)
Claude agents address issues
Progressive reasoning levels (low → xhigh) with fixed point
Human approval before commit

Quick Start

Enter loop-codex-review mode, parse args for review mode and starting level (default: low) and launch n parallel codex review commands via Bash tool with run_in_background: true. All n must be clean to advance to the next level. Always set -c model_reasoning_effort explicitly. Do NOT perform the review yourself — delegate to Codex via the CLI. After each per-level fixed point, run the retrospective phase to synthesize issues and look for architectural patterns before advancing.

Frequently Asked Questions about loop-codex-review

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

FAQPage Schema

How do I run parallel code reviews with Codex and Claude?▼

You can run parallel code reviews with Codex and Claude by launching multiple Codex CLI instances in the background, using Claude agents to address issues, and requiring all reviews to be clean before advancing.

What is progressive reasoning in automated code reviews?▼

Progressive reasoning in automated code reviews is a multi-level analysis approach that escalates from low to xhigh reasoning effort, ensuring a fixed point is reached with minimal ambiguity before human approval.

How do I automate pre-PR code reviews using Codex CLI?▼

You can automate pre-PR code reviews by delegating the analysis to Codex via the CLI, running n parallel background commands, and orchestrating retrospective synthesis to find architectural patterns before committing.

Can I configure the number of parallel Codex review agents?▼

Yes, you can configure the number of parallel Codex review agents using the -n argument, and you must explicitly set the model reasoning effort with the -c flag to control the analysis depth.

Why require human approval before committing automated code review fixes?▼

Human approval before committing is required to ensure oversight after the parallel Codex reviews and Claude agent fixes reach a fixed point, minimizing ambiguity while maintaining developer accountability.

What is a retrospective synthesis phase in code review workflows?▼

A retrospective synthesis phase in code review workflows is a process run after each per-level fixed point to synthesize issues across multiple parallel Codex reviews and identify broader architectural patterns before advancing.