GHAW-H: A Dataset of
GitHub Agentic Workflow Histories

Dataset Overview

GHAW-H connects repository metadata with the observable history of natural-language workflow specifications and their aligned lock files. It contains 262 repositories, 604 source histories, 2,820 Markdown versions, and 2,820 lock snapshots.

Cumulative observed adoption of GitHub Agentic Workflows by repository
Cumulative observed adoption, based on the earliest confirmed committed workflow source in each repository.

What is included?

262repositories
604source histories
2,820Markdown versions
2,820lock snapshots
Cumulative observed GitHub Agentic Workflow adoption by main language
Cumulative observed GH-AW adoption by main language.

Relational schema

Entity-relationship diagram for the five GHAW-H tables
GHAW-H v0.1 relational schema. Field definitions are available in the table dictionary.

Explore the dataset

Load GHAW-H

Read all five tables and construct the core analytical joins.

Open in Colab →

Dataset Overview

Inspect coverage and reproduce the cumulative adoption figure.

Open in Colab →

Language usage

Compare adoption across repository language ecosystems.

Open in Colab →

History analysis

Explore version depth, history span, and temporal change.

Open in Colab →

Citation

If you use GHAW-H, please cite the dataset. Paper-specific bibliographic details will be added when available.

@dataset{ghaw_h_2026,
  title   = {GHAW-H: A Dataset of GitHub Agentic Workflow Histories},
  year    = {2026},
  url     = {https://huggingface.co/datasets/pavtch/GHAW-H},
  version = {0.1.0}
}