I consider a model of costly screening in the presence of adverse selection and moral hazard.  A Principal faces a heterogeneous pool of candidates, and must decide whether to accept of reject them. She would like to accept only high types, but screening is costly. Prior to screening, agents can exert effort to improve their type. Since screening is costly, Principal might not want to evaluate all candidates., and may want to filter some of them using a first stage test. However, such first stage screening tests are often manipulable. Examples include initial screening using resumes (that people can lie on)  in recruitment prior to conducting an interview. The question that I ask is: In this setting of costly screening and investment, what is the Principal-optimal first stage test, in the class of all manipulable tests? The somewhat surprising  answer is that the best first stage test may be the one that's always manipulated in equilibrium, and filters out no one. The reason for this is that in equilibrium, there is a trade off between screening and investment. The test that appears to be useless because it screens out no one, actually incentivizes the maximum investment by the agents to improve their types, leading to the best ex ante pool of agents, and under some conditions, is the optimal test.