What problem does it solve? One-shot code execution loses all state between runs, forcing you to re-import libraries, reload data, and redefine variables on every step. This Skill provides a persistent Python REPL backed by a live Jupyter kernel so you can build up state incrementally while exploring data, APIs, or ML workflows. ## Core Features & Use Cases - Stateful Code Execution: Run Python code against a live Jupyter kernel where variables, imports, and objects persist across executions. - Variable Inspection: List and preview live kernel variables to inspect DataFrames, models, and intermediate results without re-running code. - Notebook Cell Editing: View, insert, replace, and delete notebook cells, plus restart-and-run-all verification for clean top-to-bottom execution. - Use Case: While exploring a new dataset, load it into a pandas DataFrame once, then iteratively filter, plot, and transform it across many steps without reloading, and finally verify the full notebook runs cleanly end to end. ## Quick Start Start a JupyterLab server, create a scratch notebook session, then ask the agent to execute Python code in the live kernel and inspect the resulting variables.
