A request such as “refunds within 30 days” leaves important behavior unstated: the time boundary, partial refunds, money arithmetic, and rejection reasons. Specification authoring turns that intent into explicit, reviewable conditions that a pipeline can check. A working prototype can help reveal rules, data shapes, and edge cases before the final feature is designed. section 1.5 section 1.4.1
Write examples in a form the project’s verification pipeline can run. Given–When–Then scenarios can capture both successful outcomes and reasons for rejection. The examples are a contract for the failures considered, not proof that every failure has been found. section 1.5 section 1.4.1
Feature: Refund policy
Rule: Time-bounded, single-order refunds
Example: Within 30 days is allowed
Given an order "O1" purchased at 2025-08-01T10:00:00Z
When a refund of 25.00 USD is requested at 2025-08-28T09:59:00Z
Then the refund is approved and remaining refundable equals 75.00 USD
Example: After 30 days is rejected
Given an order "O2" purchased at 2025-08-01T10:00:00Z
When a refund is requested at 2025-08-31T10:00:01Z
Then the request is rejected with reason "expired"
Example: Partials must not exceed the original
Given prior refunds of 90.00 USD on a 100.00 USD order
When a refund of 20.00 USD is requested
Then the request is rejected with reason "exceeds-remaining"The refund scenarios make approval, expiry, and the limit on partial refunds concrete. section 1.5
Use an iterative authoring loop:
- Prototype to discover the real rules and checks; keep the prototype simple and treat it as disposable. section 1.5
- Specify observable behavior and boundary cases, then add application constraints such as money representation and idempotency strategy. section 1.5
- Ask an agent for focused, verifiable artifacts and additional corner cases. Review its proposals against human intent rather than accepting them as the specification. section 1.5 section 1.4.1
- Keep asking what the contract omits. A generated test suite may appear exhaustive while sharing its author’s blind spots. section 1.4.1
Checking a refund boundary
From the refund scenarios, an agent can be asked to generate Cucumber step definitions, a pure refund policy, and property tests probing 29d23h59m, 30d00m00s, and 30d+ε. The request also calls for a rationale about time zones and rounding. These outputs still need review against the intended policy. section 1.5 section 1.4.1
Even a refined specification can omit real input: an ISBN validation case may account for hyphens but miss spaces. Add the missing case, check the intended standard, and continue looking for unstated assumptions. section 1.4.1

