Why is this problem marked with a "recursion" tag? Did anyone at all use recursion for the solution? Seems that iterative is so straight forward, there's no need.

Java fork which improves the older original Java fork + addresses the issues listed below (more details in the discourse of former).

Is this really a level 5kyu kata? because I don't feel tho

C:

• the intial code does not compile
• the input should be const-qualified
• it is not explained how to allocate the returned pointer

js fork 🤖🤖🤖

• made description language-agnostic
• updated to node 18
• added randomly generated cases in addition to the randomized ones
• used mocha/chai assertions
• used lodash for randomness
• provided input on failure

Random tests in C++ are really bad, with the biggest problem of having the random loop set up incorrecly so it almost never generates inputs larger than 100 elements.

Is this meant to come with some code pre-written? Some languages have significant partial implementations, whereas other languages only have the bare-bones you usually see with kata.

Languages with boilerplate only: Crystal, Python, Ruby

Languages with solution partially implemented: C#, JavaScript

Haskell translation

Input should not be provided as a sequence when it is of fixed size.

JavaScript: Reference solution and example solution are unreasonably golfed and should be changed to some more maintainable form. I would do it myself, but I cannot make heads and tails of this mess. Halp!

How exactly are the spells produced? I could figure out how to derive modifier and maximum spell level!

Haskell fork

• Submission test input range expanded to make $O(n^2)$ solutions fail.

The C# reference solution returns different results compared to other languages.

level = 11, score = 32
Test Failed
  Expected: 161
  But was:  166

Possible Java translation