Re: AGENTS.md and CLAUDE.md Are Concessions to Proprietary Services, Not Defenses Against Them

On Tue, Aug 4, 2026 at 4:31=E2=80=AFPM Jean Louis <[email protected]> wrote:

> Nevertheless you can prove the point if you find any standard on using ag=
ents.md and if such standard isn't there who is it them benefiting?

https://agents.md/

It is an emerging convention and it benefits teams where contributors
do not have to agree on a single harness.

> I don't use harness. Why you think people need to use harness to generate=
 code?

Anything that takes LLM output, evaluates parts of it as commands, and
feeds those commands=E2=80=99 output back to the LLM, is a harness.

Something that interfaces with an LLM but is not a harness is just a
chat UI. One could, in principle, manually paste the contents of a
file into a chat UI, give an instruction, get text back, and copy that
text back into the file. That would be code generation without a
harness. People do not, in general, do that.

> Since when would Emacs need to comply to rules set by Anthropic?
>
> And it will anyway waste time on vendor specific motion.

Have you been listening? AGENTS.md is *as far* from a vendor-specific
convention as possible. And it is exactly Anthropic Claude Code that
still insists on a vendor-specific CLAUDE.md name.

> >Basically as a maintainer of a GPL-licensed project you can=E2=80=99t en=
force
> >any policy on the individual user, until and unless they start
> >redistributing it.
>
> That's right. Just goal is to prevent LLM generated contribution and that=
 can be done technically like git hooks and by review and observation proce=
ss.

The goal is not that. The goal is to have contribution that is
license-compatible with the rest of Emacs.

That is a social problem and those cannot be meaningfully solved by
technical means.