OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Multi-Agent Systems

Arena Skill

Arena Skill is a Claude Code skill that launches sub-agents on the same task with different strategy cards, then runs paired critique, revision and rubric-based judging to select a surviving answer. It returns that answer along with the attacks it survived, its card and the round count.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Arena Skill — OpenViST project profile in Multi-Agent Systems.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when another ordinary request for a revised answer is not enough and you want multiple approaches to compete. Choose a smaller competitor count for routine work or a larger bracket when the task merits the additional token use.

Who it is for

  • Claude Code users who want to compare alternatives after receiving an unsatisfactory answer.
  • Users weighing different solution approaches for coding, writing, planning or reasoning tasks.
  • Developers who want a bracket-based workflow for reviewing generated candidates.

Documented uses

  • Generate several strategy-diverse responses to one task.
  • Have candidate solutions critique and revise one another before selection.
  • Compare a winning candidate with an answer the user had rejected.
  • Inspect the critiques and rounds behind the final selection.
Before you choose

Requirements & limits

The repository identifies the skill as MIT-licensed. The README describes it as free software with no signup or API key required; Claude Code and Python 3.8 or newer are required. Token use is the user's, and the README notes that larger brackets and longer tasks consume more.

  • The default 100-agent bracket is estimated at 595 sub-agent calls over seven rounds; 16 agents use 91 calls over four rounds.
  • The task is sent to each competitor, so the README says token use grows with task size and agent count.
  • Sub-agents write in .arena/; the skill returns the winning answer for the user to decide whether to apply.
  • The tournament uses a written rubric covering correctness, completeness, robustness, specificity and clarity.
  • The strategy-card concepts can be adapted to a particular domain, but the supplied project description does not document a tested implementation of custom domain cards.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Install Arena Skill in Claude Code by copying its folder

Place the skill in a global or project-local Claude Code skills directory so its slash command is available.

Before you start

  • Claude Code
  • Python 3.8 or newer
  • Git and a shell for the documented clone-and-copy method

Steps

  1. Clone the repository with the command below.

    Documented command
    git clone https://github.com/Jakeschincariol/arena-skill.git

  2. For global installation, copy arena-skill/skills/arena to ~/.claude/skills/.

  3. For project-local installation, copy the same folder into the repository's .claude/skills/ directory.

Expected outcome

The copied skill is installed as `/arena` in Claude Code.

Guide limitations

  • This guide covers folder installation, not execution or verification of the skill.
  • The README also describes plugin installation, where the command is namespaced as /arena-skill:arena.
  • Python dependencies do not need to be installed with pip.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
6 November 2026
Original website title
Jakeschincariol/arena-skill · Official homepage title · retained in its source language
How we review sourcesSuggest a correction