llms.txt ideas

I've been thinking a lot about llms.txt lately, and here are some ideas I had to make it better

1 - Advocate to use [link rel="alternate"](https://developer.mozilla.org/en-US/docs/Web/HTML/Reference/Attributes/rel#alternate) to indicate from the HTML page where the text/markdown variant is located:

```

```

Currently nobody is using this, making it a harder process when agents get served links to standard HTML pages. This would allow machines to more easily find the machine-friendly version directly, rather than going through `/llms.txt` first.

2 - Advocate to support varying on [`accept` header](https://developer.mozilla.org/en-US/docs/Web/HTTP/Reference/Headers/Accept) between HTML and 'text/markdown'. Also, advocate to default to text/markdown responses if no accept header was provided.

This will allow serving machine-readable files if accept header `text/markdown` was provided or preferred. This would allow an even easier way to gather context incase only the human URL was found, directly serving the agent-friendly variant if accepted.

Start by reading the issue's linked MDN references and the repository's existing llms.txt guidance; no project files or tests are named. Done would require an agreed, documented decision on alternate links and Accept-based markdown negotiation, with any implementation or test location identified.

Written by the indexing model from the issue text.