When doing a quant problem with test statistics do you assume it's one tailed unless told otherwise?

[original post removed]

If it is “different from 0 or another number”, it should be a two-tailed test. If “bigger or smaller than”, should use one-tailed.

and when there is a test for whether there is a unit root or heteroskedasticity, that has to be one tailed right because it’s either a yes or no?