ds-project-layoutlisted
Install: claude install-skill shashankg7/claude-ds-project-layout
# Data-science project layout
Use the [Cookiecutter Data Science v2](https://cookiecutter-data-science.drivendata.org/)
layout for every new data-science or ML project. In an existing project, put new
files where this layout says. A project's own documented conventions (README,
CLAUDE.md, AGENTS.md, CONTRIBUTING) take precedence where they differ.
## Target layout
```
├── LICENSE
├── README.md <- what the project is, how to run it, current results
├── Makefile <- one-word entry points: make data, make features, make train
├── pyproject.toml <- package metadata and tool config; the module is installable
├── requirements.txt <- or a uv/poetry lock file
├── data/ <- gitignored
│ ├── raw/ <- original, immutable inputs; never edited in place
│ ├── external/ <- third-party data
│ ├── interim/ <- intermediate transforms
│ └── processed/ <- final, model-ready tables and feature files
├── models/ <- trained models, predictions, model summaries (gitignored)
├── notebooks/ <- exploration only; named <n.m>-<initials>-<topic>.ipynb
├── references/ <- data dictionaries, papers, manuals
├── reports/ <- generated analysis (HTML, Markdown, CSV)
│ └── figures/
├── docs/ <- longer documentation, if any
└── <module_name>/ <- importable source code
├── __init__.py
├── config.py <- paths and constants; the only place paths are defined
├── dataset.py