QA Prompts: Templates That Actually Save Time
A good prompt isn’t a magic spell — it’s a well-scoped task. Break it into parts and you can assemble any request in a minute with a predictable result.
The formula
Role + Context + Task + Format + Constraints.
- Role — “You’re a senior mobile QA engineer”.
- Context — product, feature, stack, what’s already done.
- Task — one, narrow, with a verb: “generate”, “find gaps”, “turn into”.
- Format — table / JSON / Gherkin / list. The model holds structure well.
- Constraints — region, versions, “no duplicates”, “negatives only”.
You don’t always need all five, but the more you give, the sharper the answer.
Seven working templates
1. Test cases from a feature
You’re a senior QA. Feature: [description]. Constraints: [platform, region, limits]. Generate test cases grouped by: positive / negative / boundary / security / network. Format — a table: # / steps / expected result. No duplicates.
2. Regression checklist
Build a regression checklist for [module]. Only what actually breaks when this area changes. Group by criticality. Short, one item per line.
3. Log / stacktrace analysis
Here’s a log [paste]. Find the root error, group the repeats, explain unfamiliar exceptions in plain words, and suggest what to check first.
4. Requirements review
Read these requirements [paste]. Find: contradictions, ambiguous wording, invalid assumptions, missing scenarios (errors, offline, permissions). Ask me questions where it’s unclear.
5. Bug report from notes
Turn my raw notes into a bug report: [notes]. Format: title / steps / actual / expected / environment / severity. If data is missing, flag what needs clarifying.
6. Test data and boundary values
Give a test-data set for the field [type: date / money / name]. Include: valid, boundary, invalid, attacks (injections, unicode, emoji), and local formats for [region].
7. Manual case → automated test
Convert this manual case into a Playwright test with Page Object: [case]. Selectors by data-testid, add explicit waits, no hardcoded delays. Point out where the test is fragile.
Techniques that raise quality
- “Ask me questions.” Add “before answering, ask 3 clarifying questions” — the answer gets much sharper.
- Iterate, don’t restart. Don’t start over — critique: “cases 3 and 5 duplicate, add an expired-session scenario”.
- Show a desired example (few-shot). Give one case in the format you want; the rest will match.
- Pin the format explicitly. “Table only, no intro” saves cleanup time.
- Role narrows the tone. “You’re a security paranoid” yields a different case set than “you’re a junior QA”.
Anti-patterns
- Too broad — “test the app”. The model spreads thin across the surface.
- No context — it’ll invent your product and get it wrong.
- Blind trust — run the code, verify facts and numbers, a confident tone proves nothing.
- Real secrets in the prompt — no credentials, tokens, or user PII.
For how to fit this into a daily process, see How QA Should Work with Claude.