The random tests for javascript show this:
input key was: xrvgxfnia , input text was: pDHt- trVv!batDINVZrv-&eYu
expected: jDDe- ndZe!s[t@THHDam-&Yq
Here you can see that 'a' is turned into '[', 'D' is turned into '@' and 'e' is turned into ''.
As the task mentions to respect cases and keep special characters, spaces, numbers and punctuation, I can only conclude that the tests do not do what the description says. This is probably same issue as reported below.

• It's not clear from the example in the description whether the additional steps of [0, -1, -2, -3] are based on the iteration number, or the key characters
• No fixed tests
• The tests shouldn't log anything
• The reference solution does not correctly adjust characters to [a; z] and [A; Z] ranges
• A vigenere cipher with each key-character shifted by key length is still a vigenere cipher, so it's a duplicate

Got it!

If you like this type of kata, you might enjoy solving my kata Derive Cipher from Plaintext.

yes, you're right! I'll improve the kata. Thanks for your comments!! :)

thank you! Just improved!

it is but with a twist, so i think it resembles to the Kata below.

Isn't this basically vigenere cipher?

CW already has at least 2 vigenere cipher katas:

https://www.codewars.com/kata/vigenere-cipher-helper
https://www.codewars.com/kata/vigenere-autokey-cipher-helper

https://www.codewars.com/kata/basic-nico-variation

read that kata and you may need to improve yours on the description and what inputs one has to expect.