It doesn't provide a way to submit tests though. You need to submit a solution and see that output.
Also the user documentation seems out of date (the returned output does not match what they say should be returned) so it is resonable to assume the rest could be out of date.

That's not really a feature request. However, you can follow your friends and then view "Solutions of Users I am Following". Keep in mind that one can refactor his/her solution. Furthermore, the runtimes aren't reliable (and therefore not shown).

How to do a single kata with a group of friends to see who is the best?

Codewars provides a REST API. However, I don't know whether the documentation is up to date.

Announce framework changes early, especially if it's a major/minor compiler version change.

Background: GHC 7.10 will incorporate the Applicative-Monad-Proposal. Katas which use the default Monad typeclass but don't implement Applicative will fail in 7.10 and later. An automated process which checks whether the solutions provided by the author still pass the test cases would be handy. If they don't, we need to either retire the kata (leave tests and solutions alone), or change the test cases. Retiring isn't feasible if Haskell isn't the only language, so that's where one needs fix the test case.

Furthermore, there's currently a discussion whether generalized Foldable functions will replace those from Data.List, which can lead to No instance for … errors.

Exception messages are too granular - 0 is not typically a positive number, so can be combined into one comparison. Also, some languages consider "0" to be positive, so maybe "greater than 0" would be more appropriate.

Magic number: use Math.PI instead of defining inline.

Magic number: Use Math.PI insead of defining inline.

Formatting is a hard to read

Importing with wildcards can be dangerous.

It's really hard to determine what is going on in your return statement.

Error message "Error" isn't very helpful to the consumer of this class.