The documentation at https://developer.amazon.com/docs indicates that skills deployed to lambda should be built using a SpeechletRequestStreamHandler and Speechlet (com.amazon.speech.speechlet.lambda.SpeechletRequestStreamHandler and com.amazon.speech.speechlet.Speechlet). However, the newest code in the SDK located at https://github.com/alexa/alexa-skills-kit-sdk-for-java/ uses subclasses of SkillStreamHandler and RequestHandler (com.amazon.ask.SkillStreamHandler and com.amazon.ask.dispatcher.request.handler.RequestHandler) to implement a skill deployed to lambda. Speechlet is not used.
What is the recommended or preferred way of doing this? And, if using SkillStreamHandler and RequestHandler, what class should be listed in the Handler field in the Lambda function configuration?