Multi-Model Skill

Run a multi-model “battle” loop where:

• Each iteration uses a different model (rotating by index).
• The active model produces a candidate answer.
• Other models judge and critique it.
• The process keeps the best-scoring answer as the current best.

Configuration (.env)

This skill reads .env (searching upward from the current working directory) to find multi-model config.

Single-endpoint setup (OpenAI-compatible)

• ARENA_OPENAI_BASE_URL (e.g. http://localhost:11434/v1 for Ollama, or https://api.openai.com/v1)
• ARENA_OPENAI_API_KEY (optional for Ollama)
• ARENA_MODELS (comma-separated model names)