Markdown mirror has navigation chrome stripped
The point of a markdown mirror is to give agents only the page content, with no nav, header, footer, or sidebar boilerplate. If the mirror is just an HTML→markdown dump that kept the chrome, agents pay token cost for content they did not ask for and have to re-extract the article anyway.
How the check decides
The check fetches the markdown mirror and counts literal <nav>, <header>, <footer>, and <aside> tags in the body. Passes if zero. Fails with the count if any are present. Returns N/A if no mirror was published at all.
How to implement it
Generate the markdown mirror from the article body only — strip the layout shell (nav/header/footer/sidebar) before serialising. Most SSGs that expose markdown directly already do this; HTML-to-markdown converters typically do not, so a generator script that round-trips the rendered HTML will need an explicit stripping step.
Pass
# Install a14y
Run `npm install -g @a14y/cli` and then `a14y https://example.com` to audit a site.
Fail
<nav>Home / Docs / API</nav>
# Install a14y
Run `npm install -g @a14y/cli` ...
<footer>© 2026</footer>