Reproduce “How Agent Skills Evolve: A Large-Scale Empirical Study of Skills on GitHub” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Jinjun Huang, Jialun Cao, Junxiao Han, Zhonghao Jiang, Shing-Chi Cheung, Zhongxin Liu
The authors reconstruct GitHub histories for 84,200 AI-agent skills and 750,898 related commits to map how skills change over time. They analyze skill components, change patterns, and inferred motivations, then compare scanner findings before and after security-labeled commits. Of 876 commits, only 15.2% and 10.8% showed fewer high-or-critical findings in two scanners, respectively; this measures scanner results, not deployed security.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.