1. Agreed.
2. Well I had to look into test cases that are not included in the description. Nevertheless, the task should be 100% well-defined without looking at any test cases, whether or not they're included.
3. Just "the same letters" may be misinterpreted as well. Do "abc" and "abd" contain the same letters? They do. Stress out the amount bit of course, but also do stress out that the sets of those letters should be equal.
4. Here are comments of people that couldn't get what to do: https://www.codewars.com/kata/585d8c8a28bc7403ea0000c3/discuss?utm_source=notifications#5ab368a7f6ef62570300003f , https://www.codewars.com/kata/585d8c8a28bc7403ea0000c3/discuss?utm_source=notifications#5a5ffcb0880385a28f000076 , https://www.codewars.com/kata/585d8c8a28bc7403ea0000c3/discuss?utm_source=notifications#5a576374e626c53403000029 ,

1. I specified that one shouldn't take characters repetition into account.
2. The description is not fine. One shouldn't peek into test cases to understand the problem statement.
3. Maybe it should say "exactly the same letters". "Similar" is vague.
4. My message is not the first on this particluar topic, yet you think that the description is fine.

The task description is formed in a way that makes it very hard to reason about it. Instead, it should be something like "all other strings are anagrams of each other, if you don't take the count of character occurrences into account." Or else it's hard to say, for input like ['abc', 'abd', 'xxx'] what should be the answer? Reading the kata description one might think that 'xxx' is, as 'abc' and 'abd' have similar characters. But in reality, this input is invalid for the problem.

So instead of addressing the issue, the comment is just being downvoted. Nice! Keep it up!

Judging by the fact it hasn't been solved in 4 months and no feedback given to your comment, it seems that nobody really cares about those corner cases. :( Although they render a good portion of solutions plainly wrong.
There's also a really interesting one edge case: (2^26 + 1)^2 - 1. Solutions that test for fraction part of sqrt() to be zero will think it's a perfect square.

Thanks. Nicely hidden.

I can't understand where should I get my referral link from... Can someone help me out?

AFAIK, all the elements with 3 symbols long names are not even synthesized yet, but in theory they may exist and be stable. Nevertheless, I think such test cases would be a nice addition.

This is, by far, the most interesting kata I solved!
And the test cases are really marvellous.