What this shape is

A project where Jupyter notebooks dominate. The classifier reports it as Notebook / Data Science when two or more .ipynb files are present.

Scanner behavior

PreFlight classifies this shape (informational). It does not raise a shape-specific finding for it; the secret and deserialization probes still run, and notebooks are exactly where their patterns show up.

The failure mode: the notebook is a trap for good habits

The notebook workflow rewards the shortcuts that cause leaks:

• A key pasted into a cell to "just get it working" is a hardcoded secret (XL-006), and notebooks are committed with that cell intact far more often than source files are.
• Cell outputs are saved in the file. A printed dataframe, an API response, a token echoed for debugging: all committed into the .ipynb and into git history.
• pickle.load / torch.load / joblib.load of a model or cache from an untrusted path is code execution (XL-001); the data-science corpus treats these as routine.
• Notebooks rarely pin dependencies, so the environment that ran is not the environment that reruns, and a swapped or typosquatted package executes in it.

When the shape is fine

Notebooks are the right tool for exploration and analysis. They are safe when secrets come from the environment (not a cell), outputs are cleared before commit, model and cache files load only from trusted constant paths, and the environment is pinned. Exploration is not an excuse to skip the boundary; it is where the boundary is easiest to forget.

Related

• Hardcoded secrets and policy text and Unsafe deserialization are the two that ship in notebooks most.