Learn With Nathan
Sign in
Full sample chapter · Prompting & Communicating with AI
The Five Ingredients of a Powerful Prompt
Use five dependable ingredients—task, context, evidence, constraints, and output—to replace trial-and-error prompting with a repeatable brief.
Chapter 02
The Five Ingredients of a Powerful Prompt
Use five dependable ingredients—task, context, evidence, constraints, and output—to replace trial-and-error prompting with a repeatable brief.
When a recipe fails because one ingredient is missing, adding more of everything rarely helps. Prompts behave similarly. A clear task without context can be generic; rich context without an output request can become a summary nobody needs.
This chapter turns prompt writing into a five-part checklist. The ingredients are simple enough for a short email and strong enough to organize a detailed business analysis.
The framework does not guarantee truth. It reduces avoidable guessing and makes missing information easier to see before the model starts.
Why five-ingredient prompt changes the result
The model can only work from the job and context it receives
Weak prompts usually omit one of five jobs: define the task, orient the model with context, identify the evidence, bound acceptable output, and specify the final shape.
The ingredients work together. Constraints without a task are arbitrary. Evidence without success criteria can be quoted but not applied. Output format without purpose creates tidy irrelevance.
Understand the core principle
Use clear language and observable requirements
Treat each ingredient as a question the prompt must answer. What job should be completed? What background changes the answer? Which facts or sources should control it? What boundaries matter? What form will make the result usable?
Two useful ideas are and . Use these concepts to explain why a prompt succeeds or fails instead of relying on a lucky phrase.
Do not force every ingredient into a long paragraph. Labels, bullets, a small table, or attached source material can communicate the same structure more clearly.
Official OpenAI prompting guidance recommends stating goals, relevant context, constraints, required evidence, success criteria, and output format while keeping instructions lean and testing on representative work.
A practical method for five-ingredient prompt
Work through a repeatable sequence
- 1. Task. Name the work and why it is needed.
- 2. Context. Supply background that changes the answer.
- 3. Evidence. Identify what should control factual claims.
- 4. Constraints. Define scope, risk, style, and approval limits.
- 5. Output. Specify the deliverable and acceptance test.
| Starting prompt | Improved prompt | Why it is stronger |
|---|---|---|
| Write a summary of this report. | For the operations director, summarize the attached report using only its contents. Return five findings, each with the supporting page, business impact, and one open question. Do not recommend a decision where the report lacks evidence. | The revision names the audience, source boundary, structure, evidence location, and uncertainty rule. |
The five-ingredient prompt workflow
Move from intent to a testable result
five-ingredient prompt workflow
Five stages turn an intention into a reviewable output
Audiobook description
The figure shows a five-step path from left to right. Step 1 is task: Name the work and why it is needed. Step 2 is context: Supply background that changes the answer. Step 3 is evidence: Identify what should control factual claims. Step 4 is constraints: Define scope, risk, style, and approval limits. Step 5 is output: Specify the deliverable and acceptance test. A reminder below the path says to run the prompt, review the result, and revise only the part that failed.
Practice, review, and improve
Fix the cause of failure instead of adding random words
If the output fails, identify the missing ingredient before rewriting the whole prompt. A single precise addition is easier to evaluate than five simultaneous changes.
Google describes prompt design as an iterative practice and covers clear instructions, context, constraints, response format, examples, decomposition, and testing.
The Academy production brief connects the chapter method, demonstration, and exercise to a reusable business practice.
Pause and apply
Reflection questions
- Which ingredient is most often missing from your current prompts?
- What source of truth should control one repeated task?
- Which output format would make review fastest for the intended user?
Sources and further reading
Chapter 2 endnotes
- OpenAI. Model Guidance — Prompting Best Practices, 2026. Official OpenAI documentation.
- Google AI for Developers. Prompt Design Strategies, 2026. Official Google guidance.
- Agent Worker Academy. Prompting for Business Results Production Brief, 2026. Academy blueprint.