Developers discuss delegating static rules to Linters rather than AGENTS.md when using Claude Code

On November 7, 2026, the AI section of DEV Community published an opinion piece titled "Why rules you can check with a linter shouldn't go in AGENTS.md". The author, whose site is xgabriel.com and who identifies a project called Hermes IDE, discussed where configuration rules should live in AI coding tools such as Claude Code.

The piece opens with a concrete example: an AGENTS.md file that says "Never import @acme/db from src/domain/." The author argues this is a lint rule written in English and handed to the one reader in the pipeline that is allowed to ignore it. The author also cites Anthropic's documentation on Claude Code memory, which the article says captures the point in a single sentence, though the excerpt available does not quote that sentence in full. The author's position is that static checks expressible as linters or git hooks should be executed by those tools rather than encoded in natural-language instruction files, whose behavior is non-deterministic.

The piece is personal commentary from a developer community post and carries no official response, tool change, or follow-up. As of now, that single article is the only verifiable source on this discussion, and no further development has been reported.