Claude Prompt Chaining With XML Handoffs

Breaking a complex task into smaller, single-objective subtasks via prompt chaining improves accuracy, simplifies troubleshooting, and provides cleaner traces than one opaque model call. Use XML tags to pass structured outputs cleanly between prompts.

Prompt 1 (Extractor): From <transcript>{{TEXT}}</transcript>, extract every action item and owner into <extracted_actions><item owner="">...</item></extracted_actions> tags. Prompt 2 (Drafter): Using the <extracted_actions> output from Prompt 1, draft individualized follow-up messages inside <follow_ups> tags.