AGENTS.md Outperforms Skills in Our Agent Evals
Summary
The article reports that AGENTS.md, a persistent documentation index embedded in a project, outperforms the Skills approach in agent evaluations for Next.js tasks. A compact 8KB docs index within AGENTS.md achieved a 100% pass rate across Build, Lint, and Test, while Skills with explicit prompts and documentation delivered significantly lower performance, highlighting the value of retrieval-led context over explicit tool use in coding agents.