To address the lack of a reusable, general-purpose interaction language for chemical procedures and robot action execution in chemistry lab automation, an interdisciplinary team at the University of York has proposed the TARMAC taxonomy for chemical laboratory robot operations. This taxonomy was distilled from a large number of experimental teaching videos across three physical dimensions, categorizing operations into four major classes comprising 18 action primitives. It has been verified to cover 98.6% of the physical operations in actual chemical processes. Once deployed on robots, these primitives can be combined to complete multiple experimental tasks and support AI agents in autonomously adjusting for deviations. This taxonomy helps bridge the communication gap between chemistry researchers and robotics researchers, improves experimental reproducibility and operational safety, and provides a foundation for building self-driving chemistry laboratories based on human-robot collaboration.