What problem does it solve? Building an analysis notebook interactively often leaves the notebook file out of sync with what was actually run, loses session context after interruptions, and hides plots from the user. This Skill provides a structured workflow for iteratively co-authoring a Quarto (.qmd) notebook with the user while keeping code, plots, and session state synchronized. ## Core Features & Use Cases - Notebook skeleton creation: Generates dated, descriptive .qmd skeletons for Python or R with Quarto YAML frontmatter and standard setup chunks. - Kernel-based exploration loop: Runs exploratory code in a Jupyter kernel (compute-node by default), saves plots as PDF to an agent_plots directory, and writes only validated chunks back into the notebook. - Session checkpointing and recovery: Saves state.md and state.json snapshots of kernel id, namespace, and next steps so the session survives context compaction and can be restored. - Use Case: A data scientist asks the agent to explore a gene expression dataset in R; the agent opens a compute-node kernel, iterates on plots visible to the user via a local URL, and incrementally builds a polished Quarto notebook. ## Quick Start Ask the agent to start an interactive Quarto notebook exploring your dataset in Python or R, iterating on plots together before writing each validated cell into the .qmd file.
