The first page I linked to shows a Confirmation Bias test designed by a LWer, based on the classic experiment. The Conjunction Fallacy has a simple written multiple choice test.

Anchoring and adjustment could be tested for by providing people with high or low random numbers and asking them to answer a bunch of estimation questions, like the "How many countries in Africa?" test. It would work out best if we already had people take this test and had gathered data to show people upon completion.