create-prd

The create-prd skill acts as a product manager for your software development lifecycle, transforming ambiguous feature requests into structured, executable documentation. Designed for developers, engineering teams, and AI agents, this tool forces clarity by requiring a deliberate, multi-step planning phase before any code is generated. By prompting for the primary user, session persistence needs, and feature scope, it ensures that subsequent development is well-defined and measurable.

• Facilitates an interactive question-and-answer session to clarify ambiguous feature requirements and project scope.

• Generates standardized PRD templates featuring introductions, goals, user stories, acceptance criteria, functional and non-functional requirements, and technical considerations.

• Includes sections for design considerations, success metrics, and open questions to identify potential blockers early.

• Enforces clear, verifiable acceptance criteria rather than vague descriptions, ensuring that tasks are testable for both human developers and automation tools.

• Automates the initial project scaffolding by creating a structured markdown file in the prds/ directory.

• Provides a clear migration path to execution via integration with ralph-json-create-issues for rapid conversion into actionable JSON task lists.

• The skill is intended to be used at the start of any new feature or project iteration to ensure alignment before execution.

• Users should expect the agent to ask 3-5 clarifying questions; responding with specific choices like '1A, 2B' improves the quality of the resulting document.

• The output is optimized for readability and compatibility with issue trackers like GitHub or custom agent-based execution loops.

• Acceptance criteria are strictly checked for verifiability; avoid abstract goals and focus on concrete UI/UX or logic outcomes.

• If a feature request is unclear, the agent will prompt for a brief description before beginning the formal planning phase.

• The tool is best utilized alongside other agent-skills, specifically those involved in task decomposition and autonomous execution loops.