Mojibake and encoding checker
日本語版はこちらCheck what encoding your file is actually in, and see the decoded text before it reaches your build.
Where the encoding gets lost
Localization files travel: a spreadsheet export, an email attachment, a hand-edit, an import script. Each hop can decide the encoding on its own, and none of them records that decision inside the file. A file that says nothing about its encoding is simply guessed at by whatever opens it next.
Japanese pipelines make this sharper, because Shift_JIS is still routinely produced by tools on Japanese Windows, and because the difference between correct text and garbled text is invisible to anyone who does not read Japanese.
What to do about it
Detect the encoding at the boundary — when the file enters your pipeline — rather than assuming it. Convert to UTF-8 immediately and keep one internal representation from then on. And verify rather than trust: a filename or an export dialog label is a hint, never a guarantee.