Specify automated inclusion for AGENTS.md for parent folders.

[mrud]

I understand that AGENTS.md files are automatically read/included if a subdirectory is read but claude for example reads automatically all CLAUDE.md files above the current working directory. Currently the standard just discovers sub folders but not parent folders.

It would be helpful to clarify what the behavior should be.

Claude Code reads memories recursively: starting in the cwd, Claude Code recurses up to (but not including) the root directory / and reads any CLAUDE.md or CLAUDE.local.md files it finds. This is especially convenient when working in large repositories where you run Claude Code in foo/bar/, and have memories in both foo/CLAUDE.md and foo/bar/CLAUDE.md.

Claude will also discover CLAUDE.md nested in subtrees under your current working directory. Instead of loading them at launch, they are only included when Claude reads files in those subtrees.

[ecki]

The agents.md doc specifies it ONLY Reds the closes agents.md file, so if you need inheritsnce you would need to Reference „For project Level Definition see the /AGENTS.md“ toplevel file to make agents inherit them.

[tmccombs]

What if there was some frontmatter metadata that could specify that the agent should automatically include an AGENTS.md from a parent/ancestor folder, so that you don't need to waste tokens and additional LLM invocations to add that to the context?

[Pjrich1313]

Oh you mean where I found my missing code for my pipe lines thanks. I do not have to worry about checking. AI does it allSent from my iPhoneOn Mar 13, 2026, at 6:35 PM, Thayne McCombs ***@***.***> wrote:tmccombs left a comment (agentsmd/agents.md#39) What if there was some frontmatter metadata that could specify that the agent should automatically include an AGENTS.md from a parent/ancestor folder, so that you don't need to waste tokens and additional LLM invocations to add that to the context? —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[zikajk]

I find the current spec a bit strange. Correct me if I'm wrong, but agent.md only kicks in once a file in that directory is accessed. If the LLM then moves on to edit a file elsewhere, the system instructions become misleading. It would make more sense if the system pulled in all AGENT.md files found in the directory tree (and maybe tagged them by scope). Am I overlooking something here?

[Pjrich1313]

Can not file it when I am blocked. So what do you expect me to do and my PR and certification are accounted for. Sent from my iPhoneOn Mar 29, 2026, at 2:20 PM, zikajk ***@***.***> wrote:zikajk left a comment (agentsmd/agents.md#39) I find the current spec a bit strange. Correct me if I'm wrong, but agent.md only kicks in once a file in that directory is accessed. If the LLM then moves on to edit a file elsewhere, the system instructions become misleading. It would make more sense if the system pulled in all AGENT.md files found in the directory tree and tagged them by scope. Am I overlooking something here? —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you commented.Message ID: ***@***.***>