Create a tiddler called CaptionsByTag. This tiddler should contain an input widget into which the user types the name of a tag. A table underneath the input widget should show all of the tiddlers with this tag along with their captions, if any. That would look something like this:

Note: You'll have to use raw HTML tables to get this to display correctly – wikitext tables don't work if you interrupt them with a list widget. An HTML table has this basic form:

<table>
  <tr>
    <th>Header 1</th>
    <th>Header 2</th>
  </tr>
  <tr>
    <td>Cell 1</td>
    <td>Cell 2</td>
  </tr>
</table>

The equivalent wikitext table would be:

|!Header 1|!Header 2|
|Cell 1|Cell 2|

tr stands for table row, th for table header, and td for table data.