Replication Package for "AI-Attributed Commits and Technical Debt in Public dbt Data-Pipeline Code: An Observational Study of Debt, Subsequent Human Commit Activity, and Ecosystem Attribution Footprint"

Replication package for the paper "AI-Attributed Commits and Technical Debt in Public dbt Data-Pipeline Code: An Observational Study of Debt, Subsequent Human Commit Activity, and Ecosystem Attribution Footprint" (submitted to Information and Software Technology, 2026). Contains the 185-commit primary corpus (122 AI-attributed across Claude, Cursor, and Devin; 63 same-repository non-AI-attributed comparison commits), the four-category debt classification protocol used by three coders, the per-coder blank worksheets, ecosystem-scale GitHub search scripts covering Devin (384,242 indexed commits) and Cursor (1,854,449 indexed commits) footprints plus a repo-first check of 298 active dbt/Airflow/Prefect/Dagster projects, and the analysis code that generates every headline numerical result reported in Sections 4 and 6 of the paper. All source commits are drawn from publicly indexed GitHub repositories under their respective open-source licenses. Data extraction cutoff: 2026-09-27. See README.md inside the archive for the full package structure, dataset schema, and reproduction instructions.