What problem does it solve? Writing PRDs and test cases separately often produces documents that drift apart, contain vague test cases, or assert behaviors the code never implements. This Skill runs a structured six-phase collaboration that produces a story-driven PRD and a code-grounded atomic test-case document as a matched pair, plus human-readable HTML review versions kept strictly 1:1 with the Markdown sources. ## Core Features & Use Cases - Dual-document generation: Produces four artifacts per requirement — PRD Markdown and test-case Markdown (fact sources for AI) plus two templated review HTML files (for humans), validated by a scripted MD-to-HTML 1:1 check. - Code-grounded test cases: Every assertion carries a file:line code reference; task-type cases specify exact task names, round counts, verbatim prompts, and per-round expected results. - Adversarial review loop: Runs at least three independent review agents (code reconciliation, coverage completeness, executability and evidence honesty) before freezing documents. - Use Case: After splitting a feature into tasks, ask the agent to write the PRD and test cases; it reads the codebase first, confirms each user story and sample test case with you, then delivers the four frozen documents. ## Quick Start Ask the agent to write the PRD and executable test cases for your already-scoped task, and confirm each stage as it walks through requirement confirmation, code reading, story drafting, and test-case design.
