Feed it a sitemap, get a picture of every template.

Pull the URLs, post the list

Upload the file directly rather than assembling JSON. Any line that looks like a URL is picked up, so a raw grep of the sitemap needs no further cleaning.

curl -s https://example.com/sitemap.xml \
  | grep -oP '(?<=<loc>)[^<]+' \
  | head -500 > urls.txt

curl -X POST https://url2image.com/api/v1/batch \
  -H "Authorization: Bearer $URL2IMAGE_KEY" \
  -F "file=@urls.txt" -F "name=example.com before migration"

A before and after that is actually comparable

Run the same list against the old site and the new one at the same viewport width. Filenames are derived from the URL path, so the two zips line up file for file and a directory diff tells you which pages changed shape.

The manifest records the HTTP status of each page, which catches the other common migration failure: a template that renders fine but now answers 404 or redirects somewhere unexpected.

Sizing the job

A five hundred page site is one batch. A five thousand page site is ten, and at the entry credit rate the whole capture costs a few dollars. Most sites do not need every page: a batch of one URL per template usually finds the breakage.