I think that the best way is to directly ask the person which algorithms/data structures they know or have used.

Since this field of knowledge is a huge collection of know-hows, there is no way to measure knowledge level with high accuracy by random sampling.