We reuse the test scripts identically compared to data-bridge and termbox apart from the names in the no-diff script.
We should sure all the committed built files are tested for differences.
We should make sure all files resulting from the build are used and if needed tested. i.e. there shouldn't be files make that we then ignore.
We should look for an off the shelf solution to check that the above criteria are true.
If that isn't possible we should at least refactor the test:nodiff command to not be a a multiline bash script tucked in package.json (i.e. put it somewhere, parameterise it etc...)