generating-test-caseslisted

# Generating Test Cases Produce the canonical test-case set into `.claude/qa/<story-id>.md`. Generate once; later presentation or format changes never regenerate. Requires a confirmed context block. If absent, run `capturing-story-context` first. ## 1. Assess complexity first Classify the story per `../references/complexity-and-volume.md`, then **state the tier, the reason, and the resulting band** before generating anything. Write it to the artifact. QA may override the tier. An override stands for the rest of the run. ## 2. Understand before designing From the story, AC, navigation, and dependencies, establish: what the feature does, who uses it, what must exist first, what changes afterwards. Do not convert acceptance criteria into test cases one-for-one. That produces a set that misses everything the AC did not think to say. ## 3. Derive with techniques Apply `../references/technique-catalog.md`. Where a situation matches a technique, use it and record the technique on the case. Techniques bound the count; unguided generation does not. Cover, where applicable: functional positive and negative, business rules, validation, boundaries, dependency states, workflow and state transitions, roles and permissions, integration failure modes, data persistence, error handling, and realistic user behavior. Security cases come from `../references/manual-security-scope.md` — that scope, and no wider. ## 4. Write cases to schema Every case carries all ten fields of `../refe