@trq212 tangentially related but does the web fetch tool in claude code just return the full html to the model? would be maybe interesting instead to let it write queries against the html page for better parsing of large docs↗