After 32 meetings of green and blue chameleons, numbers will be [98, 0, 3] (chameleons don't disappear, so the sum must remain the same). After that, meeting red+blue gives [97,2,2] and then you need TWO more meetings of green+blue to get [101, 0, 0]. 32+1+2 = 35.

For() should return an array of three strings. They should differ from code example above in one and only one symbol, they should all be different and each of them should be the correct code that outputs exactly 20 stars.

Nothing more, nothing less. As of now, 18 people has completed this kata, with 17 solutions being correct (solution of CIS 122 is kind of cheating as he found only one such string and forced test system to accept the answer). Changing i-- to i++ obviously isn't a correct solution.

Wait, what?
"A sequence consists of three or more consecutive cards of the same suit."
I don't see even one 6-7-8 of the same suit.

13-1-2-3

K-A-2-3 ? "An Ace can be either at the top of a sequence, next to the King or at the bottom of a sequence next to the Two, but not in the middle: K-A-2 is not valid."

Few questions:

1. I can see only one meld in the given example (three 7s). Why the correct answer is 4?
2. Is it right that set of 1,2,3,11,12,13 of the same seed gives four melds?

Nice kata, with a number of different ways to solve. Thanks.

Test cases may need another maze, with isolated parts, like this:
{ 0, 1, 0, 0, 1, 0,
1, 1, 1, 0, 1, 1,
1, 0, 1, 0, 0, 1,
1, 1, 0, 1, 1, 1 }

This is a frequent issue with different Kata. Refresh the page ([F5]) and re-submit.

Do you know how to write different descriptions for different languages? I only found a way to do it for code fragments, not for main text (which in some kata includes formal description of restrictions of the arguments).

Here is puzzle kata on breaking cipher (usage of a cipher, to be exact):
http://www.codewars.com/kata/554362a749c6028f06000007
Another puzzle, more guessing than cryptoanalysis:
http://www.codewars.com/kata/top-secret-deciphering-mission

I think it should be stated that n is actually small enough so we shouldn't worry about overflow (for C#, less than int.MaxValue).

Would be a good idea to add test like {25, 100, 25, 25} (i.e. when order makes the task impossible).

Ok, maybe it's really better. Done.

Not sure what you mean. There are (at least) three different ways to change one symbol in this code so that there is no infinite loop and code "prints" 20 stars. No, changing i-- to i++ isn't one of them ;)
In order to create the kata I had to write solution myself, so it definitely exists.