I felt a bit taken back by this aswell, and was the part i struggled the most with. It just doesn't feel very clean.

I used trial division and it was too slow, but then i looked for other factors in my code to improve the execution time, and was able to greatly reduce the excecution time. As a somewhat spoiler free hint, i thnk that memoization was a big factor for making it, mine in particular, faster.

i've translated the test generator from the JS version. Please have a look over it.

Added a bunch more tests based on the randomly generated tests of the JS version.

C# Translation published.

I am talking about C#. Sorry, i should have stated that.

i was a bit confused with the output section of this kata. given an empty input, i expected the expected output would be two empty strings - the tests, however, only expect one empty strign to be returned if an empty string is given as input.
IMO this could use with a touch up ;)