RFC: Need a standardized way to show different AGENTS.md files and/or content to different agents

What variant of Codex are you using?

CLI

What feature would you like to see?

I opened this issue (https://github.com/agentsmd/agents.md/issues/185) a little while ago in the AGENTS.md standard repo, but honestly there aren't many eyeballs on those issues, so I figure it's time to poke around with it.

Now that the GitHub copilot cli also is adopting AGENTS.md it's getting even more difficult to show only the relevant bits to different harnesses.

In short, I want to be able to maintain a main AGENTS.md with harness agnostic info, and overlay something like AGENTS.codex.md, so I can compose the contents without having to do some ugly generation scripting like I'm doing now.

This seems like a better solution than the others I can think of and we've all tried:

• xml tagging and scripting that strips irrelevant tags
• CLAUDE.md and copilot-instructions.md, etc... symlinked
• @ references in a CLAUDE.md or copilot-instructions.md file
• passing in different files as command line args on launch
• skills (I love and use skills heavily of course, but there are still things I want in context by default!)

Additional information

There are some similar or related ideas open already:
https://github.com/openai/codex/issues/1974

There are some similar but kinda irrelevant open issues:
https://github.com/openai/codex/issues/10067

Upvotes and comments encourage on the AGENTS.md repo issue instead of this one, but of course the codex team could ship an experimental feature.

My current setup works kinda okay, but it's a pain to manage, and means I do still end up showing a ton of irrelevant info to my agents (e.g. info about gpt models to Claude). Not to mention that I find the models overall need vastly different instructions... Which I simply cannot do properly with the current options.