Presentation from ESIP 2026 January meetig. Agentic coding using large language models has the potential to simplify the task of developing data analysis and machine learning training scripts for Earth Science. The models are powerful, however proper context is key to generating useful scripts. Teaching responsible use of these tools requires students to understand the data provenance, machine learning techniques, data quality, and model validation. We will discuss the design of course modules that meet these needs and demonstrate how community curated context files can improve the quality of the generated code. Plenty of time will be allowed for participants to discuss the feelings and issues this type of teaching raises.