software-tester-skills

Requirements & Design Testing (Static Testing)

Reviewing requirements, user stories, designs or specs before coding (shift-left static testing) — finding ambiguity, missing acceptance criteria, testability gaps, contradictory rules, and NFR omissions; INVEST checks, walkthroughs, and requirement-quality scoring.

A defect found in requirements costs ~1×; the same defect found in production costs 30–100×. Review before anyone writes code.

Quality checklist (apply to every requirement)

AttributeQuestion
UnambiguousCould two engineers implement it differently? Highlight weasel words: fast, easy, appropriate, etc., should, as needed, user-friendly, robust, similar to.
TestableIs there an observable pass/fail? Has a number/threshold?
CompleteError cases, empty state, permissions, limits, undo, audit, i18n, a11y, data retention?
ConsistentContradicts another rule/screen/API? Same term = same meaning?
FeasibleCan it be built/measured within constraints?
TraceableHas an ID, source, owner?
AtomicOne requirement per statement (no "and/or" bundles)?

User-story review: INVEST + 3C

Independent · Negotiable · Valuable · Estimable · Small · Testable. Card, Conversation, Confirmation (acceptance criteria).

Question generators (use on every story)

  • Who can do this? Who must not? What happens for each role?
  • What if input is empty / huge / duplicated / concurrent / offline / slow / partially failed?
  • When does it start/expire/retry? Time zones? DST?
  • How much: volume, latency, size, rate limit, cost?
  • What next: what state is the data in afterwards? What is logged/notified?
  • What changed: which existing behavior, reports, integrations or migrations does this touch?

Testability review of the design

Observable state (logs, IDs, events)? Seams for stubbing time/network/payments? Deterministic IDs/clocks? Feature flag/kill switch? Idempotent operations? Test data creatable via API?

Output: review findings table

| # | Ref | Type (ambiguity/gap/conflict/NFR) | Quote | Question / Suggested rewrite | Severity |

Example rewrite: "Search should be fast" → "p95 search latency ≤ 300 ms for 1M records at 50 concurrent users."

Techniques

Walkthrough · technical review · inspection (roles: author, moderator, reader, scribe) · perspective-based reading (user, tester, ops, security) · prototype/wireframe review · Three Amigos.

bdd-gherkin-testing, acceptance-testing, test-strategist, test-design-techniques, threat-modeling