US-0982

Search v2

SHIPPED

AI drafted the designLast verified 4 min ago

User story

searchindexing

Channels & integrations · auto-detected from this story

In-app

Acceptance criteria

2 / 2 verified

Each scenario is run as a Playwright test by the QA agent

Scenario 1: Typo-tolerant match Verified

Givena misspelled query

Whenthe operator searches

Thenrelevant orders still appear in the results

Scenario 2: Results stay fresh Verified

Givena newly created order

Whenit is searched within a second of creation

Thenit appears in the results

Spec quality · craft learnings applied

7/10 passing

Generalisable rules the agent applies to every spec — they grow from review feedback & flaky scenarios, and act as a self-improving linter.

Assert one outcome per scenarioAtomicity

Each scenario asserts a single outcome

One trigger per scenarioAtomicity

Every scenario has exactly one trigger

Keep implementation out of Given/When/ThenClarity

Scenarios describe behaviour, not implementation

Make every outcome machine-checkableTestability

Outcomes are concrete and observable

Bound time-sensitive outcomesTestability

Time-sensitive outcomes carry a bound

Pair every happy path with a failure pathCoverage

No failure or edge-case scenario — add the unhappy path

Happy-path-only specs ship bugs; the failure modes are where the real spec lives.

Assert the negative spaceCoverage

No scenario states what must not happen

Say what must NOT happen (no double-charge, no email when muted) — bugs hide in the unstated.

State actor, capability and benefitClarity

Story is missing a clear actor/benefit

“As a [role], I want [capability], so that [benefit]” — a missing benefit means the goal is unclear.

One promise per storyScope

The story makes a single promise

Keep the slice thinScope

2 scenarios — a thin, shippable slice

Learnings applied · from the library