jupyter
π―SkillInstallation
npx skills add g1joshi/agent-skills --skill jupyterRun it in a terminal; it asks which AI coding tools to install it for (Claude Code, Codex, Cursor and more).
Uses skills.sh Β· Node.js 22.20+
After installing
Ask for related work and the agent uses it on its own, or call it yourself.
/jupyterInstall by hand
Copy the whole skill folder from GitHub to the place below. Copying only SKILL.md can leave out files the skill uses.
Open the skill folder β- This project
.claude/skills/jupyter/ - All projects
~/.claude/skills/jupyter/
These paths are for Claude Code. For other tools, the command above puts it in the right place.
SKILL.md
View on GitHub βExpert Jupyter Notebook assistance covering interactive Python, kernels, ipywidgets, magic commands, and reproducible data science. Use when prototyping data analysis, ML experiments, and interactive visualizations.
Jupyter
Jupyter provides the standard interactive computing architecture for data science and machine learning, featuring reproducible notebook workflows, interactive widgets, and polyglot kernels.
When to Use
- Exploratory Data Analysis (EDA): Interactive data investigation, statistical analysis, and visual storytelling.
- Machine Learning Experimentation: Rapidly prototyping model architectures, training loops, and loss visualizations.
- Interactive Technical Reports & Teaching: Combining executable Python/R/Julia code, rich Markdown, LaTeX, and charts.
- Automated Batch Notebook Execution: Executing parameterized notebooks in production pipelines with Papermill.
Quick Start
# Install and launch JupyterLab
pip install jupyterlab
jupyter lab --port 8888
# Useful built-in cell magics
%%time
import numpy as np
arr = np.random.randn(1000, 1000)
eig = np.linalg.eigvals(arr)
Core Concepts
Essential IPython Magic Commands
Optimizing execution, timing, and module reloading:
# Automatically reload imported modules when modified on disk
%load_ext autoreload
%autoreload 2
# Measure execution time of a single statement
%timeit [i**2 for i in range(1000)]
# Profile execution time of an entire cell
%%time
import numpy as np
arr = np.random.rand(5000, 5000)
eigvals = np.linalg.eigvals(arr)
# Render Matplotlib plots inline with high DPI
%matplotlib inline
%config InlineBackend.figure_format = 'retina'
Interactive UI Widgets with ipywidgets
Building interactive parameter tuning controls inside the notebook:
import ipywidgets as widgets
from IPython.display import display
import matplotlib.pyplot as plt
import numpy as np
def plot_sine_wave(frequency=1.0, amplitude=1.0):
x = np.linspace(0, 10, 500)
y = amplitude * np.sin(frequency * x)
plt.figure(figsize=(8, 3))
plt.plot(x, y, color='#2563eb', lw=2)
plt.ylim(-3, 3)
plt.grid(True, alpha=0.3)
plt.title(f"Sine Wave: freq={frequency}, amp={amplitude}")
plt.show()
# Connect function to interactive slider widgets
widgets.interact(
plot_sine_wave,
frequency=widgets.FloatSlider(value=1.0, min=0.1, max=5.0, step=0.1),
amplitude=widgets.FloatSlider(value=1.0, min=0.1, max=3.0, step=0.1)
);
Parameterized Execution with Papermill
Running notebooks programmatically in production data workflows:
# Execute notebook from command line with custom parameters
papermill monthly_report.ipynb output_report.ipynb \
-p customer_region "EMEA" \
-p fiscal_year 2026 \
-p threshold_amount 50000.0
Common Patterns
Interactive Parameter Tuning with ipywidgets
Problem: Manually re-running notebook cells with different hyperparameter values is slow and tedious.
Solution:
Use ipywidgets.interact for live sliders:
from ipywidgets import interact
import matplotlib.pyplot as plt
import numpy as np
@interact(frequency=(1.0, 10.0, 0.5), amplitude=(0.5, 3.0, 0.2))
def plot_wave(frequency=2.0, amplitude=1.0):
t = np.linspace(0, 1, 500)
y = amplitude * np.sin(2 * np.pi * frequency * t)
plt.figure(figsize=(8, 3))
plt.plot(t, y)
plt.ylim(-3.5, 3.5)
plt.show()
Best Practices
Do:
- Restart kernel and run all cells (
Kernel -> Restart and Run All) before committing to verify reproducibility. - Use
nbstripoutorjupytextin Git hooks to avoid committing huge binary outputs and image outputs. - Move complex, reusable functions from notebook cells into tested
.pymodules and import them. - Configure
%config InlineBackend.figure_format = 'retina'for crisp charts on high-DPI displays.
Don't:
- Leave notebooks with out-of-order execution states (e.g. In [45] before In [2]).
- Commit sensitive database passwords, cloud tokens, or API keys inside notebook output cells.
- Use Jupyter notebooks for complex long-running production services; export production logic to scripts/packages.
Troubleshooting
| Error | Cause | Solution |
|---|---|---|
No module named '...' in notebook cell | Notebook running in different Python kernel/environment than active shell. | Install ipykernel in target venv: python -m ipykernel install --user --name myenv. |
Kernel died, restarting | Process ran out of RAM or triggered C/C++ segmentation fault. | Monitor system RAM usage and downsample large datasets before plotting. |
IFrame / Widget Javascript error | JupyterLab extension not enabled or version mismatch. | Run jupyter lab build or install compatible widget extensions. |
References
More from this repository10
A template skill providing a starter framework for creating new agent skills with structured sections for triggers, quick start examples, and capability descriptions.
Skill
Skill
Skill
Skill
Skill
Skill
Skill
Skill
Skill