a pandoc filter for creating annotations

I'm trying to convert some html documents containing annotated text into an odt (or docx) with the annotations as comments. It seems like something for a pandoc filter.

The input contains the text, an identifier and link...

<span class="comment c-MIGWMM5CsW0d4l4y">Some annotated text
  <sup><a href="#c-MIGWMM5CsW0d4l4y">*</a></sup>
</span>

and elsewhere in the same file the comment text and identifier...

<p role="comment" class="comment" id="c-MIGWMM5CsW0d4l4y">a comment</p>

an annotation/comment in an ODT file looks something like this...

<text:p text:style-name="P1">
  <office:annotation office:name="__Annotation__0_1417392441" loext:resolved="false">
    <dc:creator>GCU Arbitrary</dc:creator>
    <dc:date>2022-02-10T17:53:16.400752734</dc:date>
    <text:p text:style-name="P2">
      <text:span text:style-name="T1">a comment</text:span>
    </text:p>
  </office:annotation>Some annotated text<office:annotation-end office:name="__Annotation__0_1417392441"/>
</text:p>

Where to begin?

EDIT: The pandoc writer for ODT doesn't appear to write annotations to .odt files and they don't seem to be represented in the AST (e.g. pandoc --from odt --to native comment.odt ignores comments)

In this case I guess there would need to be a way to represent the annotations in the AST and possibly reader/writer extensions rather than a simple filter?

Another approach could be an XML or XSLT filter for soffice