Hat	Stage	Upstream gate	Output	Unblocks
QA	QA	All relevant BDDs approved	qa/outputs/test-cases/PRD-NNN-cases.md	/test-spec

When to Use

All relevant BDDs for a PRD are approved; user wants detailed test cases.

Prerequisites

• Every .feature file referenced has an approved entry (BDD-{NNN}-{cap}) in .workbench-state/approved.json.

Protocol

1. Gather approved BDDs for the PRD — cross-reference approved.json with qa/outputs/bdd/.
2. Expand each scenario. Scenario Outline Examples rows → individual test cases.

3. Write qa/outputs/test-cases/PRD-{NNN}-cases.md — table columns:

   TC ID | Title | Feature | Scenario | Preconditions | Test data | Steps | Expected result | Priority | Type | Automation candidate | Notes

4. Append Automation coverage summary — totals, P0/P1 automated, P2 manual, no-flagged candidates.

Column Rules

Column	Values
TC ID	TC-{NNN} zero-padded per PRD
Priority	P0 (must-automate), P1 (should), P2 (manual ok)
Type	functional, regression, smoke, performance, security, a11y
Automation candidate	yes, no, manual-only
Steps	numbered, imperative, one action per step
Test data	JSON or k/v; synthetic only

Output Frontmatter

id: TC-set-{NNN}
status: draft
created: {today}
owner: {gh-user}
epic: {EPIC_ID}
prd: PRD-{NNN}
source_features:
  - PRD-{NNN}-{capability-slug}.feature

Do Not

• Use production data in Test data.
• Skip the automation-coverage summary.
• Mark everything yes for automation — visual / regulatory / judgment cases stay manual-only.

Source

skills/test-cases-gen/SKILL.md