A broad prompt can produce a polished response while leaving the real task unresolved. The model needs useful context, but the person reviewing the answer also needs a clear account of what was requested.
State the job, the relevant input, and the expected form of the result. Include important constraints and explain what should remain unknown rather than filled in. A small example can clarify a format or tone that is otherwise difficult to describe.
Review the output against the request and revise the prompt around the first meaningful mismatch. A longer prompt is not automatically a better one. The practical goal is a clear task and a result whose quality can be inspected.
A small working example.
Imagine asking for a short introduction that uses only the supplied notes and identifies missing information. The request gives the draft a clearer boundary.
A note to keep beside it.
Separate what was generated from what was checked. A useful result includes enough context for a reader to recognize the decisions that still need human judgment.
- State the job and relevant input.
- Describe the expected result.
- Leave unknown facts explicit.
Follow a related question
Define the understood inputs.
Where automation should hand backDistinguish active work from saved references.
A calmer browserKeep learning
Related background to continue exploring this subject.
Google: prompt design strategies Google: an introduction to language models