tpo-acceptance-criteria-template

Write acceptance criteria using the Given/When/Then format so each criterion describes an observable behavior. Cover positive, negative, and edge cases, and link every criterion to a requirement so developers and QA can verify it objectively.

Acceptance criteria for AI Developer Agents should include functional behavior, UX states, data validation rules, security and permission rules, and performance targets. The developer must return an implementation summary, test results, evidence, and known limitations for review.

Positive cases verify expected behavior with valid inputs, negative cases verify correct handling of invalid inputs like missing fields or unauthorized users, and edge cases verify behavior under unusual conditions like empty states, large datasets, or concurrent updates.

When criteria fail, mark the decision as Rejected, list the failed AC IDs, and specify required fixes. Use a rework prompt that constrains the developer to fix only the failed criteria, avoid unrelated changes, and provide evidence that each failed criterion now passes.

Include security criteria whenever a feature involves permissions, private data, or secrets, and performance criteria whenever response time, reliability, or large inputs affect the user experience. These non-functional criteria prevent silent failures that functional tests miss.