A new study analyzing GitHub repositories reveals how open-source projects are instructing their coding agents. Researchers found that instruction files, such as AGENTS.md and CLAUDE.md, are present in a significant portion of popular repositories, providing guidance on project knowledge, build processes, and collaboration rules. However, the study also highlights inconsistencies in AI attribution policies and the prevalence of AI-generated content that is often edited by human developers. AI
IMPACT Provides insight into how developers are currently integrating and directing AI coding assistants within open-source workflows.
RANK_REASON Academic paper analyzing open-source coding agent instructions. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →