I think Velleman's "How To Prove It" might be useful.