A minimalist benchmarking tool designed to test the routine-generation capabilities of LLMs.