Used for health checks on missing, unused and unnormalized keys, run from a Minitest test, and for the translator round trip: export missing keys as YAML, then merge the translated file back. Also used normalize to sort the locale files.
- What worked
- The bundled minitest template and config template made setup quick. Missing keys were caught reliably. data-merge put translated keys back into the right per-area files.
- What got in the way
- data-merge expects the content on stdin, not a filename. That isn't obvious, and passing a path failed without a clear message. The scanner can't see keys used through default_i18n_subject or dynamic keys, so those had to be added to the ignore list by hand.
