Skip to content
NewA calmer, faster ZeroHour

GHAgentCI catalogs coding-agent use in GitHub Actions

Original title: GHAgentCI: A Dataset of Coding-Agent Invocations in CI Workflow Histories

arXiv cs.AI / cs.LG / cs.CLby Daniel Ogenrwot ()1 min read
AI summary · x-ai/grok-4.7Summaries by Vani AI

The dataset has 15K-plus workflow histories, 159K-plus revisions and 205K-plus agent-evidence records from mined GitHub repositories.

Researchers release GHAgentCI, a dataset of coding-agent invocations found in GitHub Actions workflow histories, built because agentic workflows often look like ordinary CI files. An open-source extractor, cai-miner, flags agent evidence from an extensible catalogue of action references, shell commands and workflow conventions. Manual review of 576 stratified records confirmed 96.1% of dedicated-action matches and 98.8% of GitHub Agentic Workflow matches, with lower rates for other evidence kinds. Analyses note a shift toward command-line and GitHub Agentic Workflow integrations and frequent later removal of detected workflows.

Unconfirmed: Confirmation rates apply only to the 576 manually reviewed stratified records, not the full corpus.

Key facts
  • Histories come from 6K-plus of 196K-plus mined repositories.
  • Records link to workflow triggers and declared execution constraints.
  • Validation was weaker for evidence kinds other than dedicated actions and Agentic Workflows.
  • Exploratory analysis found differences in declared safeguards.
Explain this AI news with Vani AIWhat was announced, what is new and why it matters, in a new chat

From the original article

1 min read at arxiv.org

Coding agents are increasingly integrated into GitHub Actions to review pull requests, triage issues and modify code. Identifying these integrations is challenging because agentic workflows often resemble ordinary CI files, with evidence appearing in action references, shell commands and workflow conventions.

arXiv cs.AI / cs.LG / cs.CL

By Daniel Ogenrwot · arxiv.org

Read the full article at arXiv cs.AI / cs.LG / cs.CL

A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.