Acceptance criteria standard
The contract between Definition and the test gate. Every acceptance criterion (AC) must be concrete enough that a test agent generates an executable test from it without interpretation. Ambiguity here becomes defects downstream, at agent speed.
The rules
- Given / When / Then, disciplined toward Specification by Example — each AC pins a concrete example, not a general intention.
- Three examples per rule: a positive case, a negative case, and a boundary case. Insurance is full of boundaries (renewal dates, backdating, currency rounding, policy states) — name them explicitly.
- Testable, not aspirational. An AC a test can't check isn't an AC.
- Ambiguity lint. These words are banned in an AC because they can't be tested: appropriate, quickly, fast, robust, user-friendly, etc., and so on, handle gracefully. A PFD-review skill (to follow) flags them; the Definition of Ready gate blocks on them.
Example
Rule: A mid-term adjustment (MTA) recalculates premium pro-rata from the effective date.
- Positive — Given a 12-month policy at £365 When an MTA adds £100 annual cover effective on day 100 Then the additional premium is £100 × 265/365 = £72.60.
- Negative — Given an MTA with an effective date before the policy start When it is submitted Then it is rejected with "effective date precedes cover start".
- Boundary — Given an MTA effective on the final day of cover When it is submitted Then one day of additional premium is charged.
Why this matters most in RAPID
In the old world, an ambiguous AC cost a clarification conversation. In RAPID, an agent will confidently build something from it — plausibly wrong. Precision at Definition is what lets the rest of the pipeline run without a human in the loop on every step. This standard is enforced at the Definition of Ready gate and consumed directly by the test strategy.