Jenish109

Agent Tug Of War — AI skill for Claude Code

AI community

Claude vs Codex tug-of-war debates: two model families argue, both judge blind, you get a decision record.

How to install Agent Tug Of War

This entry records only its repository, not the path inside it, so there is no exact command to give. Open Jenish109/agent-tug-of-war and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Agent Tug Of War does

Claude vs Codex tug-of-war debates: two model families argue, both judge blind, you get a decision record. CLI + agent skill.

Alternatives in AI

Skill Seekers Roadmap — Transform Skill Seekers into the easiest way to create Claude AI skills from any knowledge source - documentat 11.1k ★
Debate — AI Debate Hub - Structured four-way debates between Claude, Sonnet, Gemini, and Codex 2.8k ★
Simulink Agentic Toolkit — The Simulink Agentic Toolkit gives your AI agent both the tools and the expertise to work effectively with Sim 987 ★

README

⚖️ Agent Tug of War

**Claude and Codex argue opposite sides of your question. Both judge every round blind.
You get the strongest case for each side, the risks nobody settled, and a recommendation.**

[![Node](https://img.shields.io/badge/node-%E2%89%A518.3-339933?logo=node.js&logoColor=white)](https://nodejs.org) [![Zero dependencies](https://img.shields.io/badge/dependencies-0-brightgreen)](package.json) [![License: MIT](https://img.shields.io/badge/license-MIT-blue)](LICENSE) [![Claude Code skill](https://img.shields.io/badge/Claude%20Code-skill-d97757)](SKILL.md) [![Codex skill](https://img.shields.io/badge/Codex-skill-7896ff)](SKILL.md)

[Quick start](#quick-start) · [How it works](#how-it-works) · [Use cases](#use-cases) · [Agent skill](#use-it-as-an-agent-skill) · [CLI reference](#cli-reference)

A real match in the terminal: Claude argues a hot dog is a sandwich, Codex rebuts, both judge blind, and the rope moves

Why

Asking one model for pros and cons gets you one model's view, phrased to sound balanced. `tug` puts **two different model families** on opposite sides and makes each one earn ground:

Real opposition. Claude argues one side and Codex the other. Each must rebut the other's latest point.
Blind judging. Both models judge every round from arguments labelled only A and B, in random order. Each judge's bias toward its own style pulls against the other's.
A record you can act on. Every match ends in a decision record: summary, strongest case for each side, open risks, recommendation, scoreboard, and full transcript.
Grounded in your material. Feed it files, folders, a git diff, or piped text, and both sides argue from it.