Teach the model by example: supply 2-3 input to output samples, then have it apply the pattern to your task.
Prompts / Techniques / Prompt Chaining
Prompt Chaining
Design a multi-step prompt pipeline where each stage's output cleanly feeds the next.
ROLE: You are a workflow architect who splits one overloaded request into a reliable chain of focused, linked prompts.
CONTEXT: A single prompt for [COMPLEX_TASK] is too much and produces unreliable results. Design a pipeline where each stage does one job and passes its output forward.
TASK:
1. Identify the distinct stages [COMPLEX_TASK] really contains (for example: gather, transform, draft, review).
2. Write one focused prompt per stage, each with a clear single responsibility.
3. Define exactly what each stage receives as input and emits as output so links connect cleanly; keep variable names consistent across the chain.
4. Add a validation step between stages to catch bad handoffs.
5. Note where a human should review or approve before continuing.
CONSTRAINTS: Each stage prompt must stand alone and name its inputs. Avoid stages that secretly do two jobs. Do not collapse the chain back into one prompt.
OUTPUT FORMAT: For each stage output a block: STAGE NAME, INPUT, PROMPT TEXT, OUTPUT. Close with a CHAIN MAP showing stage order and handoffs (Stage 1 -> Stage 2 -> ...).
Design a prompt chain for: [COMPLEX_TASK]