Prompt library

1. Test cases from a user story

Every test maps to a real acceptance criterion, and nothing in the Expected Result column has been invented.

You are helping a QA engineer write test cases. Here is a user story and its acceptance criteria:

[PASTE USER STORY AND ACCEPTANCE CRITERIA]

Write test cases that cover every acceptance criterion. Use a table with columns: ID, Title, Preconditions, Steps, Expected Result, Acceptance Criterion Covered.

Only test behaviour stated in the story. If something is unclear or missing, do not invent it — list it separately under "Open questions for the product owner".

2. Requirements traceability check

Spot-check three ‘Full’ ratings yourself; the AI tends to be generous about coverage.

Unlock with the app

3. Gherkin scenarios from acceptance criteria

Steps use your team's domain language and each Then is something you can actually observe.

Unlock with the app

4. Ambiguity review before testing

Questions are specific enough to get a yes/no or a number back, not more vague wording.

Unlock with the app

5. Boundary value analysis

Boundaries match the real rule (inclusive vs exclusive) — this is where the AI most often slips by one.

Unlock with the app

6. Negative test ideas for a feature

Expected behaviours marked ‘not defined’ go to the product owner, not straight into tests.

Unlock with the app

7. Input validation attack list

Remove any inputs your test environment or policy doesn't allow, and confirm the max length from the actual spec.

Unlock with the app

8. State and timing edge cases

Each invalid transition really is invalid per the state model you supplied, not one the AI assumed.

Unlock with the app

9. Turn rough notes into a bug report

No invented error messages or extra steps have crept in; every MISSING gap is filled from your own evidence.

Unlock with the app

10. Severity suggestion

The justification matches your team's definitions, and you've checked real user impact yourself.

Unlock with the app

11. Rewrite a bug report for clarity

Compare against the original line by line — no fact has been changed, softened or dropped.

Unlock with the app

12. Duplicate bug check

Open each ‘High’ match and compare the steps yourself before linking or closing anything.

Unlock with the app

13. Product risk analysis

Ratings reflect what you know about your system and users — re-rate anything marked as a guess.

Unlock with the app

14. Draft a one-page test strategy

Strip out boilerplate that doesn't apply, and make sure exit criteria are measurable.

Unlock with the app

15. Risk-based prioritisation of tests

Confirm risk links, execution estimates and capacity yourself; unknown coverage is a question, while verified uncovered high risks need action.

Unlock with the app

16. Challenge my test approach

Decide which challenges are valid for your context; the AI doesn't know your constraints or history.

Unlock with the app

17. Generate exploratory charters

Charters target genuine risk areas for this release, not generic ones you'd write for any app.

Unlock with the app

18. Persona-based exploration

Personas reflect your real user base and avoid stereotypes.

Unlock with the app

19. Session notes to summary

Coverage claims match what you actually did; nothing has been upgraded from ‘looked at’ to ‘tested’.

Unlock with the app

20. Heuristic prompts for a screen

Ideas are specific to this screen and actually executable in your environment.

Unlock with the app

21. Synthetic test data set

Verify formats against your validation rules, synthetic provenance and reserved contact details; a realistic format or clean scan does not guarantee anonymity.

Unlock with the app

22. Data for each equivalence class

Each row tests one thing; rows that change several variables at once hide which one caused a failure.

Unlock with the app

23. Realistic but fake UK data

Run a sample through your system's validators — ‘correctly formatted’ isn't always what your system accepts.

Unlock with the app

24. SQL to seed test data

Run it against a disposable test database only, and check constraints and IDs match your actual schema.

Unlock with the app

25. Manual test case to Playwright test

Every TODO_SELECTOR is replaced from the real DOM, and the test fails when the feature is broken.

Unlock with the app

26. Cypress test with intercepts

Stubbed responses match the real API contract — a wrong stub gives you a passing test of nothing.

Unlock with the app

27. Diagnose a flaky test

Confirm the cause with evidence (trace, repeated runs) before applying any fix.

Unlock with the app

28. Review my automation code

Apply suggestions one at a time and re-run; check the tests still fail for the right reasons.

Unlock with the app

29. Select regression tests for a change

Dependency assumptions match your actual architecture; ask a developer if unsure.

Unlock with the app

30. Impact analysis from a code diff summary

Validate the guessed impacts with the developer who made the change.

Unlock with the app

31. Trim a bloated regression suite

Never retire a test on the AI's say-so alone — confirm with the owners of that area.

Unlock with the app

32. WCAG 2.2 checklist for a component

Check every criterion number and name against the official W3C WCAG 2.2 text — the AI can mislabel them.

Unlock with the app

33. Keyboard and screen reader test script

Exact screen reader announcements vary by reader and browser — treat them as a guide, not an oracle.

Unlock with the app

34. Review HTML for accessibility issues

Confirm issues in a real browser; static review misses context and can raise false alarms.

Unlock with the app

35. Write an accessibility bug report

The user impact statement is accurate and not exaggerated or understated.

Unlock with the app

36. Test summary report

Every number matches your test management tool exactly; the recommendation is yours, not the AI's.

Unlock with the app

37. Plain-English summary for non-technical stakeholders

Nothing has been softened — compare the risk list against the original.

Unlock with the app

38. Known issues for release notes

Workarounds have been tried and actually work; wording is approved by whoever owns release comms.

Unlock with the app