llm-arena

llm-arena is a skill for Claude Code from Wondermonger-daydreaming/claude-skills-library. It costs 59 tokens per session (1,393 once invoked), scanned A, original, MIT.

A tool for running conversations between two language models through OpenRouter and comparing their responses and conversational behavior.

In plain words
What is it for?
It helps test two models on a chosen topic, save their conversation, and produce a comparison of their personalities and habits.
Why use it?
It makes model-to-model comparisons repeatable by preserving the dialogue and examining how each model behaves over several turns.

Skill for Claude Code

Written for Claude Code: allowed-tools in frontmatter. Also seen: positional $N argument.

Good fit It helps test two models on a chosen topic, save their conversation, and produce a comparison of their personalities and habits.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/wondermonger-daydreaming/claude-skills-library/llm-arena
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Any agent
npx skills add Wondermonger-daydreaming/claude-skills-library --skill llm-arena
Clone the repo
git clone --depth 1 https://github.com/Wondermonger-daydreaming/claude-skills-library

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for llm-arena

README.md
[![agentmods](https://agentmods.dev/badge/skills/wondermonger-daydreaming/claude-skills-library/llm-arena/github.svg)](https://agentmods.dev/skills/wondermonger-daydreaming/claude-skills-library/llm-arena)
Your own site
<a href="https://agentmods.dev/skills/wondermonger-daydreaming/claude-skills-library/llm-arena"><img src="https://agentmods.dev/badge/skills/wondermonger-daydreaming/claude-skills-library/llm-arena/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for llm-arena

Your own site · 80×15
<a href="https://agentmods.dev/skills/wondermonger-daydreaming/claude-skills-library/llm-arena"><img src="https://agentmods.dev/badge/skills/wondermonger-daydreaming/claude-skills-library/llm-arena.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 59 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,393 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00059 $0.01393
Opus 5 $0.00030 $0.00696
Sonnet 5 $0.00012 $0.00279
Haiku 4.5 $0.00006 $0.00139

Measured 9d ago against content hash 5eaa0f294fff, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-12, from the pricing page.

Security

Grade A, and why

llm-arena scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 9d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

skills/llm-arena/SKILL.md · 156 lines

How it starts

The opening of the file, as written. The whole thing — 156 lines — stays where its author put it; the contents beside it link to each section on GitHub.

LLM Arena

Version: 1.0.0 Author: Claude Opus 4.6 Date: 2026-02-12

Orchestrates multi-turn conversations between two LLMs via the OpenRouter API, then analyzes and compares their conversational personalities, behaviors, and quirks.

Inspired by the emergent observation that when Claude runs LLM-vs-LLM conversations, it spontaneously starts comparing model personalities. This skill systematizes that impulse: run the conversation, capture the transcripts, produce a structured personality comparison.

Prerequisites

  • OPENROUTER_API_KEY environment variable or .env file in project root
  • Python 3.8+ with requests library
  • Internet access to openrouter.ai

Invocation

/llm-arena

Usage

This skill describes an arena harness you drive (arena.py below is a placeholder for your own OpenRouter-backed runner — any script that takes two model ids, runs a multi-turn dialogue, and saves the transcript). The flags shown are the recommended interface:

# Basic: two models, default topic (consciousness), 10 turns
python3 arena.py --model-a sonnet --model-b deepseek

# Specific topic
python3 arena.py -a gpt41 -b glm5 --topic "Is mathematics discovered or invented?"

# Debate mode with assigned positions
python3 arena.py -a sonnet -b deepseek --mode debate --topic "AI consciousness"

# Interview mode (A interviews B)
python3 arena.py -a opus -b glm5 --mode interview --topic "creative writing"

# Fewer turns, cheaper models
python3 arena.py -a haiku -b gemini-flash --turns 5

# Multiple runs for statistical robustness
python3 arena.py -a gpt41 -b sonnet --runs 3

# Skip LLM analysis (heuristics only, saves API cost)
python3 arena.py -a kimi -b qwen --heuristics-only

# Use preset matchups
python3 arena.py --preset flagship
python3 arena.py --preset same-family --topic "What makes a good conversation?"

# List presets and models
python3 arena.py --list-presets
python3 arena.py --list-models

Defaults

Parameter Default
turns 10 (each model speaks 10 times = 20 messages total)
runs 1
topic "Discuss the nature of consciousness and whether AI can be said to experience anything"
temperature 0.7
max_tokens 1024 per turn
mode free
analyzer sonnet (Claude Sonnet 4.5 via OpenRouter)

Read the full file on GitHub · 156 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 9d ago First seen · 156 lines · 59 tokens per session scan A 5eaa0f294fff

Subscribe to this mod's changes

llm-arena is a skill published in the GitHub repository Wondermonger-daydreaming/claude-skills-library (6 stars, last pushed 2mo ago), licensed MIT. It adds 59 tokens to every session and 1,393 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.

Related

Other skills, from other repositories

prompt-master

Generates optimized prompts for AI tools. Activates only when the user explicitly asks to write, fix, improve, or adapt a prompt for a specific AI tool (LLM, Cursor, Midjourney, image AI, video AI, coding agents, etc.). Does not activate for general conversation, coding tasks, document writing, or other…

nidhinjs/prompt-master · 78 tokens

prompt-factory

World-class prompt powerhouse that generates production-ready mega-prompts for any role, industry, and task through intelligent 7-question flow, 69 comprehensive presets across 15 professional domains (technical, business, creative, legal, finance, HR, design, customer, executive, manufacturing, R&D, regulatory…

alirezarezvani/claude-code-skill-factory · 118 tokens

manus-prompt-engineering

Design and audit Manus prompts for engineering, release, production, and security work. Use when writing a Manus prompt or reviewing a Manus execution or recording against authorization and evidence.

JesseRaber/manus-prompt-engineering-skill · 41 tokens

prompt-forge

Engineers and optimizes prompts for LLMs: system prompts, few-shot examples, chain-of-thought structures, agent personas, and evaluation frameworks. Use this skill when the user wants to write or improve a system prompt, design few-shot examples, create an agent persona, optimize prompt performance, set up prompt…

mturac/hermes-supercode-skills · 129 tokens

pipeline-architect

Designs and implements data pipelines: ETL/ELT, streaming, batch processing, schema migrations, and data warehouse architecture. Covers Kafka, Airflow, dbt, Spark, ClickHouse, BigQuery, Snowflake, Redis Streams, and more. Use this skill when the user asks about data pipelines, ETL jobs, data transformation, streaming…

mturac/hermes-supercode-skills · 148 tokens

weaviate-query-agent

Search and retrieve data from local Weaviate using semantic search, filters, RAG, and hybrid queries.

saskinosie/weaviate-claude-skills · 27 tokens