AGENTS.md positions itself as a dedicated, predictable file where developers describe how AI coding agents should work inside a repository. Instead of scattering instructions across prompts or chat histories, the format gives agents one place to look for project context. The project reports adoption across more than 60,000 open-source repositories, which suggests it is becoming a de facto convention rather than a single-vendor specification.
The content of an AGENTS.md file is practical and concrete. Typical sections cover setup commands such as installing dependencies, starting a dev server, and running tests, alongside code style rules like TypeScript strict mode or formatting preferences. Because these details are often too granular for a README, the format keeps them separate while remaining plain Markdown that any tool or human can read.
The site is aimed at developers and maintainers who use coding agents on real projects, especially those collaborating in open source where contributors arrive with different tools. It also suits teams standardizing how agents receive instructions across repositories. Anyone evaluating the convention can browse examples and the GitHub project before adopting it.
To get started, create an AGENTS.md at the repository root and fill it with the commands and conventions an agent would otherwise have to guess. Keep it focused on agent-relevant guidance and leave human onboarding material in the README. Review it when build steps change so i
评论与评分
0还没有评论，来写下第一条评价吧。