Great, tests broke with default skip_inventory=True as expected :-)

The initial suggestion was that we would use a data-inventory attribute and decide about registering based on that.

However, that has approach 2 problems:

• It doesn't conform to the HTML standard on boolean attributes (https://html.spec.whatwg.org/#boolean-attributes)
• It creates more friction, as now all elements that are automatically registered (which is the vast majority) will have data-inventory="true" and _process_headings would have to treat (1) no presence of data-inventory and (2) presence with a truthy value (which requires a new convention, because it's string checking; like: is data-skip-inventory="n" a truthy or falsey value?) as the same flag. I don't like that.

Therefore, I suggest that we only add data-skip-inventory to all excluded elements. It will treat any presence of data-skip-inventory as the intent to skip the inventory, so all elements from

<h1 data-skip-inventory></h1>
<h1 data-skip-inventory=""></h1>
<h1 data-skip-inventory="true"></h1>
<h1 data-skip-inventory="false"></h1>  <!-- sic -->

would be treated as excluded from registration in the inventory, and only elements without the presence of data-skip-inventory would be registered.

Check https://smokeshow.helpmanual.io/0a4t3q1d6l2f062y3b67/usage/configuration/headings/#skip_local_inventory from mkdocstrings/python#297 to see how this works in practice :)

Thanks for the analysis and taking the right direction! Also, clever example in the docs!