databricks-repl

Consolidate a Databricks REPL session into a single, clean Python file. Use this skill when the user wants to finalize, export, or consolidate a REPL session into a committable script. Triggers on requests to consolidate session output, produce a final script from REPL commands, export session to Python, clean up REPL artifacts into production code, or finalize a Databricks workflow.

Execute Python code on a Databricks cluster via a stateful REPL session. Use this skill when the user wants to run Python on Databricks, perform data analysis with Spark, train models on a cluster, query Unity Catalog tables, use sub_llm() for recursive LM calls, or any task requiring a persistent Databricks execution context. Always use a dedicated --project-dir (e.g., examples/my-task/) to isolate session.json and repl_outputs per task. Covers session lifecycle (create, exec, await, cancel, destroy), output file management, and eviction recovery.