arc-code-agent-setup

Comprehensive guide to using the ARC-AGI Toolkit Python scripts for running remote game attempts, managing scorecards, and exploring games

Use when a remote ARC-AGI-3 run has finished and you need to diagnose what happened from scorecards, reasoning logs, and replay evidence

Use when studying an ARC-AGI-3 game, planning the next remote attempt, or deciding how to learn the most from limited actions

Use when designing or restructuring this repo's ARC-AGI-3 remote attempt runner so experiments stay comparable, observable, and action-efficient