Currently when Alexa recognizes an intent that has two slots in both the utterances and in the intent json, sometimes it does not capture all of the slots. So as an example MYAPP has "PlayIntent" intent with slots "artist" and "location" so I can say "tell MYAPP to play {artist} in {location}".
Alexa never fails to send the {artist} slot but sometimes the second slot is not captured and not sent to the lambda code. The result is a lot of ridiculous dancing around how to (in node) do an response.ask and then do a lot of conditional watching for sessions and prompt for the {location} node.
Given that Alexa Skills Kit KNOWS it didn't hear one of my slots, it would seem appropriate for the developers of the skills kit to implement one OR both of these solutions:
1) Allow a slot to be marked as "required" and automatically or through the intents JSON prompt the user for the missing information.
2) Set Alexa Skills Kit to parse the response JSON and allow developers to force the intent for the next call in the session. This seems like a much more useful approach given that voice interaction is most easily managed as stepped conversations in order to eliminate the requirement for users to memorize complex commands.
I don't even know if the Alexa skills kit team reads this forum, but if they do, please please please consider these suggestions. Right now building for ASK is a lot like handling web forms where a failed submit resets the entire form and we can't do any validation before submission.