RFC: wkdomains RFC / Draft / 2026-05-01 Account creation policy for web agents

Hi, I would like feedback on whether this belongs in or near the llms.txt ecosystem.

Draft RFC: https://wkdomains.com/rfcs/2026-May-01-account-create

The RFC proposes a standard `/.well-known/agent-policy.json` file that lets websites tell AI agents whether they may create accounts, under what conditions, and with what human approval. The goal is to make account creation explicit instead of guessed from HTML, terms pages, or trial-and-error behavior. A policy can describe allowed or disallowed account types, required human consent, verification expectations, rate limits, contact points, and audit requirements.

I think this repo is a good place to discuss it because `llms.txt` already gives sites a simple, web-native way to help language models and agents understand how to use a website. Account creation is not just content retrieval, but it is a common next step for agents that use websites on behalf of humans. Agents need a discoverable signal before taking that step.

This may not belong directly inside `llms.txt`. A few possible outcomes:

1. Add guidance that `llms.txt` may link to an account/action policy file.
2. Define a conventional section such as `## Agent policies`.
3. Keep this separate, but document how `llms.txt`, `robots.txt`, and `/.well-known/agent-policy.json` should relate.
4. Recommend that agents check both `llms.txt` for context and a dedicated policy file before account creation or other state-changing actions.

The core question: should `llms.txt` remain focused only on LLM-readable content, or should it also help agents discover adjacent policy documents for safe website use?

Start by reading the linked RFC at https://wkdomains.com/rfcs/2026-May-01-account-create and compare its proposal with the repository's existing llms.txt guidance. Review how the issue's suggested relationship between llms.txt, robots.txt, and /.well-known/agent-policy.json could fit the project. Done requires a decided scope and an agreed outcome for this repository, but the issue does not identify implementation files or tests.

Written by the indexing model from the issue text.