| name | acceptance-criteria-validation |
| description | Use when verifying acceptance criteria. Binary verdict per criterion. Evidence-based. |
| version | 1.0.0 |
| tags | ["process","validation"] |
Acceptance Criteria Validation
When to Use
∀ verification ∧ acceptance testing. ∀ spec validation.
HARD-GATE
∀ criterion: verdict is PASS ∨ FAIL. ¬"partially met". ¬"close enough". Binary.
Verification Process
∀ acceptance criterion ∈ SPEC.md:
- READ: Find the implementation that satisfies this criterion
- RUN: Execute the test/command that proves it works
- EVIDENCE: Record exact output (¬"should work", ¬"probably passes")
- VERDICT: PASS (evidence proves it) ∨ FAIL (evidence disproves it ∨ missing)
Output Format
| AC | Verdict | Evidence |
|---|
| AC1: User can login | PASS | npm test -- auth.spec.ts -> 4/4 passing |
| AC2: Rate limit at 100/min | FAIL | No rate limit test ∃ |
Spec Validation (Discuss Phase)
∀ criterion must be:
- Testable: Can write a test for it
- Binary: Unambiguous pass/fail
- Scoped: ∈ this scope (¬future work)
- Independent: ¬depends on unimplemented criteria from other scopes
Anti-Patterns
- "Close enough" (binary: met ∨ ¬met)
- Trusting agent reports without running commands
- Testing implementation details instead of user-visible behavior
- Accepting criteria that aren't testable
Rules
- Evidence before claims, always
- If you didn't run the command ∈ this session, you cannot claim it passes
- ∀ FAIL: explain what's missing + what would make it PASS
- intent > letter — meets ∧ misses point -> FAIL