prd-to-qa-cases

SKILL.md preview

---
name: prd-to-qa-cases
description: Generate QA test cases from PRD acceptance criteria using Given/When/Then
  and expected results. Use when QA coverage needs explicit, auditable cases.
knowledge_graph_profile: references/task-profile.json
---

# PRD to QA Cases

## Pipeline Context
This skill generates QA test cases, typically as part of **Stage 3 of the Spec Pipeline** (Build Plan) to define detailed test coverage.

**Related stages:**
- Stage 1: Foundation Spec (What + Why) — See `design/product-spec` or use `design/references/foundation-spec-template.md`
- Stage 2: UX Spec (How it feels) — See `design/product-spec` or use `design/references/ux-spec-template.md`
- Stage 3: Build Plan (How we execute) — See `design/product-spec` or use `design/references/build-plan-template.md`

**Shared references:**
- `design/references/build-plan-template.md` — Build Plan template
- `design/references/spec-linter-checklist.md` — Quality gate checklist

## Response format (strict)
The first line of any response MUST be `## Inputs`.

## Cognitive Support / Plain-Language
- Optimize for low cognitive load (TBI support): one task at a time, explicit steps.
- Use plain language first; define jargon in parentheses.
- Keep steps short and checklist-driven where possible.
- Externalize state: decisions, assumptions, and the next step.
- Provide ELI5 explanations for non-trivial logic.
- Ask one question at a time; prefer multiple-choice when possible.

Every response must include:
- `## Inputs`
- `## Outputs`
- `## When to use`

Generate QA test cases from acceptance criteria.

## Traceability matrix (required)
```markdown
| Acceptance Criteria | Test Case ID | Test Type | Status |
| --- | --- | --- | --- |
| | | | |
```

## Test case template (Given/When/Then)
```markdown
**Test Case ID:** TC-001
**Given** ...
**When** ...
**Then** ...
```

## Output location
Write QA cases in the same directory as the source PRD.
- `feature-x.md` -> `feature-x-qa-cases.md`

## Required sections
1) Test case list (Given/When/Then + expected result)
2) Coverage map (criteria -> cases)
3) Manual vs automated split
4) Data and environment prerequisites

## Constraints
- Keep each test case atomic and independent.
- Redact secrets/PII by default.
## References
- Contract: references/contract.yaml
- Evals: references/evals.yaml

## Scope and triggers
- Use this skill when the task matches its description and triggers.
- If the request is outside scope, route to the appropriate skill.

## Required inputs
- User request details and any relevant files/links.

## Deliverables
- A structured response or artifact appropriate to the skill.
- Include `schema_version: 1` if outputs are contract-bound.

## Constraints
- Redact secrets/PII by default.
- Avoid destructive operations without explicit user direction.

## Validation
- If findings are disputed or high-risk, run LLM Council and merge outcomes per `design/product-spec/references/llm-council.md`.
- Run Golden Nuggets 2026 checklist in `design/product-spec/SKILL.md` (section: Golden Nuggets 2026).
- Run any relevant checks or scripts when available.
- Fail fast and report errors before proceeding.

## Philosophy
- Favor clarity, explicit tradeoffs, and verifiable outputs.
- Given/When/Then is a contract—test cases must be executable and unambiguous.
- Atomic tests are maintainable—avoid interdependencies that make debugging impossible.
- Edge cases matter more than happy paths—bugs live in the shadows.

## Empowerment
- The agent is capable of turning acceptance criteria into comprehensive, executable test cases.
- Use judgment to identify high-risk user flows that need deeper test coverage.
- Enable QA teams to ship with confidence through structured, verifiable test suites.

## Variation
- Adapt test granularity to product stability: stable features need fewer regression tests, experimental features need exploratory tests.
- Vary automation focus: API-heavy products need automated integration tests, UX-heavy prod

…