Incorrect converting of 👁 & 👁️

$ cat puny.js
#! /usr/bin/node

const punycode = require('punycode/');

console.log(punycode.toUnicode("xn--mp8h"));
console.log(punycode.toUnicode("xn--v86cr064b"));
console.log(punycode.toASCII("👁"));
console.log(punycode.toASCII("👁️"));

$ ./puny.js
👁
👁️
xn--mp8h
xn--v86cr064b

The punycode xn--v86cr064b is invalid, so should throw an error or something, and both code points should map to xn--mp8h - this is the result you will get from Python's integrated .idn() encoding and the result you get from https://www.punycoder.com/

$ python3
Python 3.12.12 (main, Oct 11 2025, 15:09:54) [GCC 14.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> utf8="👁"
>>> print(utf8.encode("idna"))
b'xn--mp8h'
>>> utf8="👁️"
>>> print(utf8.encode("idna"))
b'xn--mp8h'
>>>

You could def make the case your code was never intended to be IDNA compliant, but only a punycode converter - however, if it is outputting domain names that are effectively "illegal", this seems to me highly undesirable behaviour - or at least, something users need an obvious and strongly worded warning about.

Start with the puny.js reproduction and run the four toUnicode/toASCII calls shown in the issue. Read the converter's RFC 3492 and RFC 5891 behavior, then establish whether these inputs should normalize to the same output or reject the invalid result. Done means the chosen behavior is documented and verified for both eye-symbol variants.