ba-prd-skills_skill

Overview

This skill helps AI agents produce context-aware, expert-driven, and token-efficient Product Requirements Documents (PRDs) organized into a clear folder structure. It blends proven product discovery techniques with automated repository reconnaissance and templates to speed PRD creation. The outputs are optimized for both human reviewers and AI parsing to minimize token costs while preserving clarity.

How this skill works

The skill starts by scanning the repository to detect project type, tech stack, architecture patterns, and documentation gaps, then saves a concise preliminary summary. It uses adaptive, expert-informed questioning (Marty Cagan, Teresa Torres, George Biddle) to refine problem definition, assumptions, and success metrics. Finally, it generates a token-efficient PRD following a structured template and places artifacts in a consistent folder layout for tracking and version control.

When to use it

• Starting a new project and you need a scoped, outcome-focused PRD
• On an existing codebase to assess feasibility and propose changes
• When preparing stakeholder reviews or sprint planning with clear success metrics
• To standardize PRD output across teams for easier automation and review
• When you need to reduce AI token usage while keeping PRDs comprehensive

Best practices

• Run repository reconnaissance before asking discovery questions to ground recommendations
• Focus on outcomes over features; define measurable success criteria up front
• Capture assumptions and risks with explicit validation plans and quick tests
• Use the provided token-efficient template: short summary, bulleted requirements, and semantic markers
• Track PRDs in git, update via pull requests, and archive superseded documents

Example use cases

• Generate a PRD for user authentication in a Next.js + Prisma project using a prefilled template
• Assess whether to migrate to Next.js 16 app router by scanning the repo and listing architectural implications
• Create prioritized opportunity maps and testable assumptions for a new product idea
• Produce a compact PRD for vector search integration with links to technical specs and research
• Standardize PRD naming, folder layout, and status tracking across multiple features