You deleted the line with your API key, but the notebook still holds it: the output of the cell that printed it, the traceback that shows postgresql://analyst:password@db, the papermill parameters in the metadata, and the .ipynb_checkpoints copy. This skill finds those leaks and seven other notebook problems before the file goes to GitHub, a ticket or Kaggle, and tells you how to fix each one.

Your agent runs a scanner on your .ipynb files (nbformat 4 and 3, from JupyterLab, Jupyter Notebook, VS Code, Colab, Kaggle and papermill, including .ipynb_checkpoints copies). It reads the notebook's JSON itself: code, markdown and raw cells, every output type (stream, execute_result, display_data, error), notebook and cell metadata, and saved widget state. Each finding names the cell and the exact line of the .ipynb file. It never runs the notebook or starts a kernel. It returns BLOCK, REVIEW, or NOT FLAGGED, with the file, line, reason, and a safer approach. Secrets are never printed, and email addresses and phone numbers are only counted.

What it checks

BLOCK (fix before shipping)

• N01: passwords, API keys and tokens written in cells (assignments, api_key= arguments, %env, auth= tuples, connection strings in %sql, Bearer tokens, Kaggle keys, ghp_, sk-, AKIA, hf_ and other known key formats, private keys)

• N02: secrets saved in outputs (printed keys, environment dumps, env: NAME=value, tracebacks that show a connection string)

• N03: secrets in metadata (papermill parameters, saved ipywidgets Password values)

REVIEW (check before release)

• N11: email addresses and phone numbers in saved outputs

• N12: pip, %pip and conda installs without an exact version, and git URLs without a tag or commit

• N13: downloads piped into a shell or Python (!curl ... | bash)

• N14: very large outputs, embedded images over 1 MB, or more than 5 MB of saved outputs

• N15: home-folder paths (/Users/name, /home/name, C:\Users\name) in code, outputs and metadata

• N16: cells run out of order (execution counts that go down)

• N17: saved error tracebacks

What's included

• The scanner (Python 3.9+, standard library only)

• A rules guide with the reason and safer approach for each rule

• A risky and a safe example project with two notebooks each, and the real report the scanner produces on the risky one

Limits

It reads the notebook file as saved: it cannot see what images show, and it does not read Jupytext pairs, .env files or the data files a notebook loads. Notebooks up to 20 MB are checked; in very large outputs only the first 1,000,000 characters are searched for secrets. NOT FLAGGED is not a guarantee.

Contents

One ZIP file with 11 files: SKILL.md, the scanner in scripts/ (4 Python files), references/rules.md (every rule with its reason and safer approach), and examples/ (4 example files plus report.md, the real scanner output on the risky example).

How to use: This is an agent skill: a folder with a SKILL.md file and Python scripts that an AI coding agent such as Claude Code, Cursor, or Codex CLI can run. Unzip it into your agent's skills folder (for Claude Code: ~/.claude/skills/), then ask your agent to check your project. The scripts also run on their own with Python 3.9 or newer, with nothing to install. They read your files only: no network access, and nothing is sent anywhere.