I've added a little more info in the description example, you could suggest any other change, I think it's clear now, but it's not set in stone either ;)

Well, of course you can assume, and if you have to assume then decimal is the sensible choice. But if I gave you a task "write a function that converts a number to a string representation", wouldn't you think that task as stated is a little underspecified?

Not a big deal, just when I fed inputs like 077 and 0x34 into my test cases it made the existence of this assumption obvious.

And why would you use any other representation if not specified? In binary it doesn't even make sense.

Agreed. Pure recursion is way too easy to blow up with a big input.

Haha, nice. Just like real software, if the bug is there for long enough it becomes part of the spec (talking about the bug in the test suite).

Well, you found a hole in the test suite at least

The problem statement implies but doesn't make it clear that it's the decimal representation of the integer that has to be processed digit by digit.