Thomas, first you need to get task and from it you get contentId of your inputs using task.getTaskData().getDocumentContentId(). Once you have it you use this id to get content using taskClinet.getContent() method. Note that you would need to use marshaller to get the object out if it.
HTH