Wirasm/prp

What it solves

It addresses the gap between high-level product requirements (PRDs) and the technical context an AI coding agent needs to write production-ready code. By providing a structured "Product Requirement Prompt" (PRP), it prevents the AI from guessing implementation details, reducing the need for multiple correction cycles.

How it works

PRP extends a traditional PRD by adding AI-critical layers: precise file paths, library versions, codebase patterns, and executable validation commands. It is implemented as a set of "Agent Skills" (slash commands) for Claude Code that automate the software development lifecycle:

• Planning: Generates PRDs and detailed implementation plans.
• Execution: Implements the plan, validates the code via commands, and opens Pull Requests.
• Review: Performs comprehensive code reviews and loops through fixes until the code is clean.
• Automation: The /prp-loop command can run the entire pipeline (plan $\rightarrow$ implement $\rightarrow$ review) headlessly.

Who it’s for

Developers, product teams, and engineering teams using Claude Code who want to move from simple AI demos to shipping production-grade features autonomously.

Highlights

• Autonomous Loop: A headless pipeline that handles the full cycle from feature description to a reviewed PR.
• Structured Artifacts: Maintains a project-specific store outside the repo for PRDs, plans, and research reports.
• openzeppelin-style orchestration: Supports parallel development using git worktrees for different PRD phases.
• Integrated Validation: Requires executable commands to verify work, ensuring the AI doesn't just write code but proves it works.