prd

The PRD skill is an essential tool for developers and product managers working within the Ralph autonomous agent framework. It streamlines the scoping process by transforming vague feature requests into structured, actionable documentation. By initiating an interactive clarifying dialogue, the skill ensures that objectives, scope, success criteria, and technical constraints are clearly defined before implementation begins. This structured approach prevents scope creep and provides the necessary context for AI coding agents like Amp or Claude Code to execute tasks effectively.

The skill produces standardized Markdown-based PRDs, which serve as the foundation for the Ralph development loop. These documents detail core functionality, user stories with verifiable acceptance criteria, and specific design or technical considerations. It is designed to bridge the gap between initial product ideas and machine-executable tasks, ensuring that every user story is scoped to fit within a single AI context window for maximum reliability.

• Automatically generates comprehensive PRDs based on user input and iterative clarifying questions.

• Structures documentation with sections for goals, user stories, functional requirements, success metrics, and open questions.

• Enforces best practices for acceptance criteria, ensuring every UI-related task includes verifiable browser-testing requirements.

• Facilitates the transition from planning to execution by preparing files for conversion to the Ralph task system (prd.json).

• Promotes consistency in documentation, making it easier for junior developers or other AI agents to interpret complex feature requirements.

• Helps identify non-goals and scope boundaries, effectively managing developer focus and project velocity.

• Use this skill when kicking off new features, projects, or refactoring efforts to align intent and execution.

• Inputs should be a high-level feature description; the skill will ask 3-5 critical questions to refine the PRD.

• Outputs are saved to the tasks/ directory as markdown files, ready for review or conversion.

• Designed specifically for the Ralph ecosystem; ensure you follow up with the companion Ralph conversion skill to enable autonomous agent execution.

• Acceptance criteria should remain explicit and unambiguous; avoid vague language like "works correctly" in favor of specific validation steps.