Linux Kernel Developers Consider Adding AGENTS.md To Help Guide AI/LLM Agents
A proposed repository guide follows tests in which coding agents mishandled kernel attribution conventions.
TL;DR
- A proposal would add an AGENTS.md file to the Linux kernel tree and direct agents to existing project documentation.
- Phoronix reports that tests found agents added incorrect or missing attribution when the guidance was absent.
- The proposal has drawn objections about sending agents through broad documentation and increasing token use.
The proposal described by Phoronix is intentionally small: an AGENTS.md file would point to the kernel README and existing guidance for coding assistants. The report says tests found attribution errors without that context and better conformity after it was added. [1] [2]
The change is still a proposal. The reported mailing-list discussion includes concern that directing agents through extensive documentation could increase token consumption, with more targeted guidance suggested as an alternative. [1] [2]
Why it matters
Large open-source projects are starting to encode contribution rules for AI tools alongside rules for human contributors, while balancing behavioral guidance against the cost of broad context loading.
Editor's note
Headline and status are based on the reported proposal; this edition does not imply that the kernel has adopted the file.