Best practices for organizing scientific software repositories

Researchers are adopting structured workflows to improve the reproducibility, rigor, and documentation of scientific software development. Key strategies include maintaining a clear separation between raw data and generated results to ensure provenance is obvious.

Effective repository organization involves using a Directed Acyclic Graph (DAG) of small, single-purpose scripts rather than relying solely on large, out-of-order Jupyter notebooks. Implementing version control through Git flow—using branches to isolate different sets of results—allows researchers to keep work clean and merge only significant findings.

Modern scientific workflows also prioritize LLM-friendliness, structuring files and folders so that large language models can more easily assist with code generation, documentation, and peer review. Utilizing tools like Snakefiles and comprehensive README files helps ensure that any researcher can follow the path from raw data to final outputs without external assistance.