Skip to content

OffchainLabs/arbitrum-docs / .agents/skills/prd

Generate Product Requirements Document via multi-agent workflow. Use for significant features, new systems, or complex changes that warrant formal specification. NOT for bug fixes, small features, or quick iterations. Triggers on "create PRD", "write requirements", "spec this out", "define the project".

PRD Generation

Orchestrate the multi-agent workflow for comprehensive PRD creation.

When working on Arbitrum projects, use @arbitrum-best-practices for domain-specific patterns.

When to Use This Workflow

USE /prd for:

  • New features with multiple components
  • System redesigns or major refactors
  • External-facing APIs or SDK features
  • Cross-team initiatives requiring alignment
  • Features with security/compliance implications
  • Complex implementations with multiple integration points

DON'T USE /prd for:

  • Bug fixes (just fix them)
  • Simple features with clear, limited scope
  • Internal tooling tweaks
  • Quick iterations on existing features
  • Exploratory spikes (use /research instead)

If unsure, ask: "Would I regret not having written requirements for this?" If no, skip the PRD.

Core Flow

research → requirements → specs → design
   (1)         (2)         (3)      (4)

This produces an Implementation Design Document that defines WHAT to build and HOW.

Full Workflow

/context-prep        → Project Orientation (run first!)
research             → Domain Research (researcher)
requirements         → Requirements Discovery (product-manager)
specs                → Technical Specifications (product-manager)
design               → Implementation Design (architect)
[skip execute]       → Implementation is separate
qa                   → PRD Validation (qa)
polish               → Terminology & Glossary (code-quality-engineer)
docs                 → Final Polish (docs)

IMPORTANT: Run /context-prep first to orient to the project.

Phase Details

Research

subagent_type: researcher

  • Analyze 3-5 similar solutions
  • Document anti-patterns and failure lessons
  • Identify domain-specific considerations

Requirements

subagent_type: product-manager

  • Iterative questioning (2-4 questions per round)
  • Extract functional requirements (FR-001, FR-002...)
  • Extract non-functional requirements (NFR-001, NFR-002...)
  • Define constraints, assumptions, out-of-scope

Specs

subagent_type: product-manager

  • Transform requirements into precise specifications
  • Define data models with types and constraints
  • Document API contracts (inputs, outputs, errors)
  • Write testable acceptance criteria

Design

subagent_type: architect

  • System design and component breakdown
  • Technology stack with justification
  • File paths and integration points
  • Security and scalability approach

QA

subagent_type: qa

  • Hostile falsification of complete PRD
  • Edge case generation
  • Traceability check (requirement → spec → design)
  • Risk assessment

Polish

subagent_type: code-quality-engineer

  • Terminology consistency audit
  • Glossary creation
  • Clarity improvements throughout

Docs

subagent_type: docs

  • Document structure review
  • Executive summary
  • Final formatting

Shared Document

All agents write to: PRD-[project-name]-[YYYY-MM-DD].md

Output

Final PRD includes:

  • Executive Summary
  • Domain Research
  • Requirements Specification
  • Technical Specifications
  • Implementation Design
  • Validation Report
  • Glossary
  • Unresolved Questions (even if "None")
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 465kUpdated Claude CodeCodexVerifiedInstall
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 465kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 318kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 318kUpdated Claude CodeCodexVerifiedInstall