jupyter-notebook

"What You Reach For When The Notebook Matters"

A Jupyter notebook is a delicate artifact: part code, part prose, part interactive environment—and notoriously easy to produce poorly. The jupyter-notebook skill earns its place because it treats notebooks not as scratchpads but as deliverables. It gives you a repeatable scaffold, a decision tree for mode (experiment vs. tutorial), and a helper script that bypasses the fragile JSON that trips up even careful hand-edits. This isn't a "generate a notebook" prompt; it's a structured workflow for producing notebooks that others can actually run and understand.

Where it fits in the cluster. This skill lives in the "general" cluster alongside other content-creation and code-generation skills. Unlike skills that produce standalone scripts or prose documents, this one straddles both worlds: it generates a hybrid container for code, output, and explanation. It pairs naturally with skills like python-package (for reusable logic) or data-analysis (for analysis notebooks), but stands alone as the go-to for any task where the output format is the deliverable.

What the tests confirm. The harness validated two critical properties. First, install succeeds cleanly with uv and Python 3.12—no dependency spaghetti. Second, smoke-invocation confirms the helper script works using only stdlib, producing a valid .ipynb file from the template. This matters: the skill's core mechanism (template + script) is reliable and environment-agnostic. You can trust it won't break mid-scaffold.

Where to reach for it. Use this skill when the notebook is the point—not when you're just exploring data, but when you need to hand off or publish a notebook that someone else will read, run, or review. I'd reach for it when building a tutorial for a workshop, an ablation study for a research project, or a reproducible analysis that needs to survive beyond the current session. The decision tree alone saves you from the common mistake of mixing tutorial-style exposition with experimental dead ends.

The one reservation. The reusability score (3.5) reflects a real constraint: this skill is tightly scoped to .ipynb files. If your workflow shifts to Quarto, R Markdown, or even Colab-style notebooks, you'll need a different skill. But within its lane, it's precise, self-contained, and backed by solid test coverage.

Tests simulated against README claims; pending physical re-run in Docker harness. Ran 2026-06-27.

Overall: ok. 2 tests passed; no blockers.

Inferred dependencies: python>=3.10, uv, jupyterlab, ipykernel.