Main goal
Write prompts that produce useful, constrained, and testable outputs instead of vague generic text.
Prompt Engineering Cheat Sheet
A practical QA-oriented prompt guide built around the exact frameworks and prompt habits we teach in the AI curriculum.
Best default
State the role, task, context, constraints, and output format explicitly.
Use it for
Test cases, bug summaries, code reviews, requirement analysis, and AI-assisted documentation.
Pair it with
Use this with AI basics, AI test generation, and LLM evaluation.
Prompt Anatomy
A strong prompt is structured, not just wordy.
Role: You are a senior QA engineer Task: Generate regression scenarios Context: Checkout flow with coupon logic Constraints: Include positive, negative, boundary cases Output: Markdown table
Zero-Shot vs Precise Prompting
Zero-shot is faster; precise prompting is stronger for repeatable QA work.
Zero-shot: "Write test cases for login" Precise: "Generate 15 login test cases with priority, preconditions, steps, and expected result"
Role-Based Prompting
Give the model a professional stance that matches the task.
You are a senior SDET reviewing Playwright code. Focus on flakiness, locator stability, weak assertions, and maintainability.
STAR Framework
Useful when the task needs situation, target, action, and result clarity.
Situation: Checkout flow changed Task: Identify regression risk Action: Generate focused test scenarios Result: Return a prioritized list
CLEAR and CRISP
These frameworks help keep prompts specific and readable under pressure.
CLEAR: Context, Limitations, Expectations, Action, Result CRISP: Context, Role, Intent, Scope, Presentation
RICE POT
This QA-specific framework is useful for structured test design prompts.
Role Input Constraints Examples Process Output Tone
Context Templates
Reusable context blocks improve consistency across a team.
Application: e-commerce Users: admin, buyer Critical module: checkout Testing focus: coupon, address, payment, confirmation
Structured Output
Ask for predictable shapes so the output can be reviewed or reused programmatically.
Return JSON with: - title - priority - steps - expectedResult - riskArea
High-Value QA Prompt
This is the kind of production-style prompt that usually works well.
You are a senior QA engineer. Analyze the following requirement. List assumptions, risks, missing details, and 20 test scenarios. Group results into positive, negative, boundary, and security.