I believe NPM uses this package: https://github.com/iarna/has-unicode
Which describes its algorithm thus:
What we actually detect is UTF-8 support, as that's what Node itself supports. If you have a UTF-16 locale then you won't be detected as unicode capable.
Since at least Windows 7, cmd and powershell have been unicode capable, but unfortunately even then it's not guaranteed. In many localizations it still uses legacy code pages and there's no facility short of running programs or linking C++ that will let us detect this. As such, we report any Windows installation as NOT unicode capable, and recommend that you encourage your users to override this via config.
Unix Like Operating Systems
We look at the environment variables LC_ALL, LC_CTYPE, and LANG in that order. For LC_ALL and LANG, it looks for .UTF-8 in the value. For LC_CTYPE it looks to see if the value is UTF-8. This is sufficient for most POSIX systems. While locale data can be put in /etc/locale.conf as well, AFAIK it's always copied into the environment.