Looks like it was using XPath to parse the HTML: https://github.com/beaulm/craigwatch/blob/master/post.php

I imagine the 'selling' point was users not even having to know what RSS even means, let alone how to script it.