How you phrase a task script directly determines the validity of your testing data:
The Leading Question Defect
Leading (Bad): 'Go to the search bar at the top right, search for Nike sneakers, and click the blue checkout button.' This tests whether the user can follow directions, not whether the interface is usable!
Non-Leading (Good): 'You want to buy a pair of running shoes for under $100. Show me how you would find and purchase them on this site.'
A sound task specifies a realistic human goal without mentioning interface elements, button labels, or navigation paths.