PROMINE Dataset: Prompt Edits from LLM-Integrated Python Repositories

This dataset accompanies the PROMINE research project, an AST-based tool anddataset for mining prompt edits from LLM-integrated Python repositories. The release contains confirmed prompt-edit commits, deterministic prompt-unitrecords, repository-level mining statistics, repository-selection metadata, andoptional model-assisted annotations. The files support research on promptevolution, prompt/code co-change, output-format contracts, placeholder changes,and prompt-edit classification. The dataset was collected from public GitHub repositories. Prompt text andother third-party content remain subject to the licenses and terms of theiroriginal repositories. The MIT License applies to the PROMINE source code, notautomatically to all contents of this dataset.