Write acceptance criteria with AI
Describe the observable result of a story in testable scenarios, from the normal path to a meaningful failure or agreed performance limit.
To write acceptance criteria with AI, start with one slice of a feature. A guest purchase is easier to specify than the entire checkout redesign. The deliver-acceptance-criteria skill turns the agreed slice into scenarios that describe the starting conditions, a single event and what should follow.
A tester can use the result to check whether a story is ready for sign-off. A product manager can read the same criteria and spot an unresolved rule, such as whether a shopper can retry after a declined payment without entering the basket again.
What it does
- Describe the successful outcome from a user’s point of view.
- Separate invalid data from a failure in a dependent service.
- Make an agreed non-functional requirement observable and measurable.
- Remove implementation assumptions that do not affect acceptance.
- Flag conflicting requirements before they become test cases.
Install and use
- Install the complete skill folder in a compatible agent. No external account is needed. Provide a scoped user story, confirmed rules and exclusions.
- Ask, “Write acceptance criteria for applying a valid voucher and for trying an expired one. Leave undefined rules as open questions.”
- Read each condition with a tester. Check that the expected outcome follows from the story and can actually be observed.
FAQ
Source
Made by Jonathan Prisant (Product on Purpose). Original skill: deliver-acceptance-criteria, published on GitHub under the Apache-2.0 licence.
View on GitHub