cmux-ai-review

Orchestrate parallel Gemini and Codex reviews through a three-stage pipeline with cross-rebuttals.

17|3|Updated Jan 23, 2026
One-click install
npx skills add https://github.com/insightflo/claude-impl-tools --skill cmux-ai-review
Or copy as Structured Prompt for Agentβ–Ό
Please help me install this Agent Skill.
Skill: cmux-ai-review
Source: https://github.com/insightflo/claude-impl-tools/tree/main/plugin/skills/cmux-ai-review
Command: npx skills add https://github.com/insightflo/claude-impl-tools --skill cmux-ai-review

SYSTEM DOCUMENTATION & REQUIREMENTS

πŸ’‘ This Skill includes references (resource) components.

What problem does it solve?

cmux-ai-review enables real-time, parallelized multi-AI code-review-like workflows by splitting a workspace into side-by-side Gemini and Codex panels and executing a true 3-stage pipeline (opinion β†’ rebuttal β†’ synthesis). This lets teams observe concurrent reviews and efficiently converge on a decision without sequential bottlenecks.

Core Features & Use Cases

  • Parallel opinion gathering: Stage 1 runs Gemini and Codex simultaneously to surface diverse perspectives.
  • Cross-rebuttal: Stage 2 exchanges opinions to generate focused rebuttals from both sides.
  • Chairman synthesis: Stage 3 uses a central coordinator to synthesize final guidance and a scorecard.
  • Panel-based workflow: cmux panels provide live visibility of each AI's outputs and logs.
  • On-demand orchestration: Logs, tasks, and prompts are managed via .claude/cmux-ai-* directories for reproducible workflows.

Quick Start

Install the skill and start Stage 1 to run parallel Gemini and Codex reviews.

Frequently Asked Questions about cmux-ai-review

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

FAQPage Schema
How do I run parallel AI code reviews with multiple models?β–Ό

Parallel AI code reviews run by splitting the workspace into side-by-side cmux panels, executing Gemini and Codex simultaneously to surface diverse perspectives without sequential bottlenecks.

What is a multi-AI cross-rebuttal workflow?β–Ό

A multi-AI cross-rebuttal workflow exchanges Stage 1 opinions between AI models to generate focused rebuttals from both sides, enabling a chairman coordinator to synthesize final guidance and a scorecard in Stage 3.

How do I orchestrate a 3-stage AI review pipeline?β–Ό

Orchestrate a 3-stage AI review pipeline by running Stage 1 for parallel opinions, Stage 2 for cross-rebuttals, and Stage 3 for chairman synthesis, enforcing stage transitions and log management via .claude/cmux-ai-* directories.

Can I use cmux panels for live AI workflow visibility?β–Ό

Yes, cmux panels provide live visibility of each AI's outputs and logs, coordinating parallel tasks via SendMessage signals to observe concurrent reviews and efficiently converge on decisions.

What's the best way to coordinate concurrent AI model reviews?β–Ό

The best way to coordinate concurrent AI model reviews is using cmux panels with SendMessage signals to enforce stage transitions, manage logs, and deliver structured output with extensible model configs and guardrails.

Does multi-AI review orchestration require sequential model execution?β–Ό

No, multi-AI review orchestration eliminates sequential bottlenecks by running Gemini and Codex simultaneously in parallel panels, then exchanging opinions for cross-rebuttals before final chairman synthesis.