This checklist is used by the agent to self-score every specification before planning begins. All items must PASS before proceeding. Fix and re-score if any item fails.
- Every feature mentioned in the prompt has at least one user story.
- Every user story has at least one measurable acceptance criterion (not vague like "works correctly").
- All happy-path flows are fully described end-to-end.
- At least the most likely edge cases and error states are enumerated per user story.
- No user story uses "should" without an explicit rationale for why it is optional.
- No requirement can be interpreted in more than one way without an explicit decision documented in
decisions.md. - All referenced entities (users, roles, objects, states) are defined in the spec or in a glossary.
- No requirement references external documents or context that is not available in the repo.
- No two requirements contradict each other.
- Terminology is consistent throughout the spec (no synonyms for the same concept).
- User stories are consistent with the governing principles in
constitution.md.
- Every acceptance criterion is verifiable by a test or a deterministic check.
- No acceptance criterion requires subjective human judgment to evaluate.
- The spec does not prescribe implementation details that belong in the plan, not the spec.
- Every API interaction specifies both success responses and error states.
- No API contract leaves authentication, authorization, or rate limiting undefined if applicable.