The AGENTS.md is interesting, apparently the primary most important principle is, "Avoid em-dashes like the plague".

That's an odd request. I always use my own voice for certain things, such as posting to hacker news, or writing my thoughts on a proposal. But for other things such as writing up a bugfix, if I'm getting an AI to write it, I'd rather not hide the fact I've done so.

In fact I usually go out my way to mark it as AI written, to give a heads up to any human reader so they don't waste their time if they don't want to read it.

edit: I'm not sure why my comment is attracting downvotes, perhaps it's being interpreted as anti-AI. I'm not against AI writing, but there are contexts where people would like to know whether something is AI written or not. I would rather it was well identified than hidden, so people can make their own judgement whether to gain insight into a human writing or whether it's just process they can skim or feed through their own agent.

"Avoid em-dashes" just seems like a crude attempt to avoid AI writing coming across as such.

...and then scroll down a few lines and you will find tons of em-dashes in the AGENTS.md