What is the primary purpose of Agent Arena?

Agent Arena orchestrates multiple AI models (Claude, Codex, and Gemini) to work together on complex tasks that require diverse perspectives, such as security audits and architecture reviews.

Does Agent Arena support human feedback during a run?

Yes, it supports human-in-the-loop (HITL) workflows, allowing you to pause a run and use the /arena:resume command after providing necessary input.

Can I use different AI 'personas' with this skill?

Yes, Agent Arena includes various profiles like security-audit, multi-expert, and research-brainstorm that define specific agent specializations.

Which models are required to use Agent Arena?

The skill is designed to coordinate Claude Code with the Codex (GitHub) and Gemini (Google) CLI tools.

How can I track the progress of a background run?

You can use the /arena:status command or monitor the live.log file located in the .arena/runs/ directory for real-time streaming output.

Agent Arena Orchestrator

Name: Agent Arena Orchestrator
Author: mikewrather

bymikewrather

0•

コラボレーションツール

Orchestrates multiple AI agents including Claude, Codex, and Gemini for collaborative code analysis and complex decision-making.

Agent Arena is a powerful multi-agent orchestration plugin for Claude Code that integrates various AI perspectives into a unified workflow. By coordinating Claude, Codex, and Gemini, it allows developers to perform adversarial security audits, collaborative code reviews, and deep architectural debates. The skill uses specialized profiles to tailor agent behavior, enabling background execution of complex reasoning tasks that benefit from multiple independent 'opinions' to ensure higher quality output and more robust security analysis.

主な機能

01Adversarial security audit profiles for identifying vulnerabilities

02Human-in-the-loop (HITL) support for resuming paused workflows

030 GitHub stars

04Pre-configured profiles for brainstorming, research, and deep reasoning

05Multi-agent coordination between Claude, Codex, and Gemini CLIs

06Background task execution with real-time logging and status tracking

ユースケース

01Conducting adversarial security audits to find hidden vulnerabilities in sensitive code

02Getting three independent perspectives on complex code reviews and quality assurance

03Debating high-level architecture decisions using multiple AI expert personas

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add mikewrather/agent-arena arena

For use in Claude.ai and ChatGPT

Download Skill