From Generation to Validation: A Multi-Source Study of AI Agents in Scientific Workflow Systems

This replication package accompanies the paper “From Generation to Validation: A Multi-Source Study of AI Agents in Scientific Workflow Systems.” The package contains the data and analysis materials used in the study across its three complementary evidence sources: (1) a systematic literature review of 47 primary studies, (2) an analysis of workflow-integrated AI tool documentation, and (3) repository mining of publicly disclosed AI-assisted development artifacts from Nextflow, Galaxy, and Snakemake. The materials include the included-study dataset and full-text assessment records for the systematic literature review, the tool-documentation analysis matrix, repository-level artifact datasets for the three studied scientific workflow systems, and scripts used for repository data collection and AI-assisted artifact labeling. These materials are provided to support transparency, reproducibility, and independent examination of the study’s data collection and analysis process.