What problem does it solve? Vague acceptance criteria like "the feature works" cannot be verified by tests, so implementations ship unproven and regressions go undetected. This Skill turns feature ideas into PRDs where every criterion is a concrete, machine-checkable assertion tied to an automated test pipeline. ## Core Features & Use Cases - Structured PRD generation: Asks 3-5 clarifying questions with lettered options, then produces a full PRD with user stories, numbered functional requirements, non-goals, success metrics, and open questions, saved to tasks/prd-[feature-name].md. - Assertable acceptance criteria: Every criterion follows an act-then-observable-result pattern tagged by verification layer ([feature], [unit], [vitest], [browser]), with a divergent-data rule for filter and search assertions. - Verification checklist and spec files: Each PRD ends with per-change coverage, test data seeding instructions, browser end-to-end steps, and suite gates, plus a spec file created under specs/features/ with Status: planned. - Use Case: A product manager says "spec out a task priority feature." The Skill asks clarifying questions, then writes a PRD where each story names exact test files, seed data counts, and dev-browser click-paths so an agent can later prove the feature works. ## Quick Start Ask the assistant to write a PRD for your feature, for example: "Write a PRD for adding priority levels to tasks with filtering and badges."