The idea, in plain English
Start here.
A useful prompt explains what you want, supplies the context that changes the answer and describes important output requirements or boundaries. Add an example when it clarifies the desired pattern. You do not need a long formula or magic role. Review the result against the brief and correct the specific gap; clarity improves steerability, not guaranteed truth.
‘Write a great update’ leaves the assistant to guess the reader, source and meaning of great. ‘Turn these display-setup notes into a short handover for a new volunteer; keep confirmed facts and open questions separate’ gives the work a clear purpose and a result you can inspect.
A few words you’ll meet.
OPEN WHAT YOU NEED ↙Task brief
A plain-language description of the useful result you want, including relevant audience, material and requirements.
Source boundary
Which material the answer should use for its factual content, and how to handle information that is missing or outside that material.
Output format
The shape that makes the result useful, such as a short notice, a comparison table or an ordered procedure.
Iteration
A focused revision after inspecting the result, followed by another check against the original goal and source.
Name the result and the reader
Start with an action and deliverable: summarize the notes, draft a handover, compare the supplied options or explain a term. Say who will use the result when that changes its detail or vocabulary. A new volunteer needs different background from someone who assembled the display yesterday.
Use only the structure that helps. A short rewrite may need one sentence of instruction. A reusable workflow may need more context, requirements and checks. OpenAI’s current prompting guidance explicitly says there is no required rigid formula. Longer is not automatically better, and a role such as ‘expert editor’ does not create real expertise or access to missing facts.
Provide the material and the boundaries
Supply the relevant notes and say what the assistant should take from them. Mark source material clearly so it is distinguishable from your task instructions. Keep the dates or conditions that must remain intact visible in the brief. Delimiters help communicate structure; they are not a guarantee that every system will interpret every document perfectly.
If the task is a source-only transformation, say how to handle missing facts. In the fictional setup notes, delivery of a missing panel is not confirmed. Ask for that as an open question, rather than inviting a guessed delivery promise. Include the information needed for this task and use material the chosen tool is permitted to process; adding every available document can make the brief less focused.
Fit the brief to the job.
Choose a format and show a pattern when useful
Pick the format for the next person’s job. A handover table can separate confirmed progress, source detail and questions. A process may need ordered steps. A notice may need a short paragraph. Specify the fields that matter rather than asking for an elaborate format with no practical use.
A small example can clarify an ambiguous style or structure. Make clear whether it is a formatting example or factual material so its sample names and dates are not imported into the result. You do not need a universal number of examples or a perfect word count. Use a length boundary for readability, then check whether the important meaning survived.
Revise the specific mismatch and check again
Compare the first result with the brief. If it invents delivery, point to the unsupported statement and ask for the gap to remain open. If it is too technical for the volunteer, ask for ordinary language while preserving the facts. A focused correction is more informative than ‘make it better.’
Recheck the whole revised output so a successful edit does not hide a new omission elsewhere. Keep the source, prompt version and useful review notes when the task will recur. If the problem is unavailable evidence or an unsuitable tool, adding more prompt wording may not solve it. Diagnose the missing ingredient instead of treating every failure as a need for a longer template.
A handover prompt with a visible gap
Original fictional notes: display stand assembled; one panel missing; setup planned for Friday; panel delivery unconfirmed. Practice prompt: ‘Turn these notes into a brief handover for a new volunteer. Use only the supplied facts. Make a table with Confirmed and Open questions. Keep Friday as the planned setup date, do not invent delivery or an owner, and flag information needing confirmation. Return a draft for review.’
| Prompt part | What it contributes | What to inspect |
|---|---|---|
| Brief handover for a new volunteer | Task and audience | Can someone new understand the setup status? |
| Use only the supplied facts | Source boundary | No invented delivery, owner or commitment. |
| Confirmed / Open questions | Useful output shape | The missing panel and unconfirmed delivery remain visible. |
| Draft for review | Use context | The result is inspectable before it is used. |
Put it into practice
Repair a vague request
Rewrite ‘Make a professional setup update’ using the fictional display notes. An initial result wrongly says the missing panel will arrive Friday and assigns it to a named owner not in the notes.
- Name the deliverable and its intended reader.
- Include the relevant notes and a source-only boundary.
- Choose a format that keeps confirmed facts and open questions distinguishable.
- Write a targeted follow-up correcting the invented delivery and owner.
- State what you will recheck in the revised output.
Keep your working notes in your own document. This page saves only your completion check, not your answers or customer data.
Check your understanding
Make the call.
Choose an answer for each question. Then compare the reasoning.
You have a useful next step.
- State the useful result and relevant reader.
- Provide the necessary material, boundaries and output shape.
- Correct a specific gap and recheck the complete result.
Completion is your own learning record. Quiz scores and practice checks are saved separately, on this browser.
Sources & context
The sources below explain the factual foundations and editorial methods used in this lesson. Worked examples, planning models and exercises are illustrative teaching material.
- OpenAI: Prompting
Current primary guidance on goals, relevant context, output, boundaries and focused follow-ups without a rigid required template.
- OpenAI: Safety best practices
Previously reviewed primary guidance on context, limitations and human inspection with original material. No claim that delimiters guarantee instruction isolation.