A Claude Code skill that builds, validates, and submits RL training environments on Prime Intellect — from any HuggingFace dataset to a live GRPO training run, fully automated.