An automated multi-agent workflow that iteratively drafts and reviews professional Product Requirement Documents through a specialized writer-reviewer feedback loop.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install prd-iterative
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install prd-iterative using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is PRD Iterative Agent?
PRD Iterative is a high-performance workflow within the Openclaw Skills ecosystem designed to transform vague product ideas into professional-grade documentation. By orchestrating two specialized AI agents—a Writer and a Reviewer—this skill ensures that your Product Requirement Documents (PRDs) undergo rigorous scrutiny and refinement.
This skill is built for developers and product managers who require more than just a single-shot AI response. It simulates a real-world product team environment where a Writer produces content and a Reviewer identifies logical gaps, UX issues, and technical inconsistencies. By using Openclaw Skills like this, teams can ensure their documentation is production-ready and technically sound before the first line of code is even written.
PRD Iterative Agent Use Cases
- Generating a complete, multi-version PRD from a brief product description or voice summary.
- Refining documentation through multi-stage professional reviews using specific AI models like Kimi or MiniMax.
- Creating platform-specific requirements for iOS, Android, or Web applications with automated technical adjustments.
- Converting competitor links or research articles into structured product outlines through automated content scraping.
How PRD Iterative Agent Works
- Information Gathering: The agent initiates by confirming the product name, target platform (Web/App), preferred AI models, and the number of iteration rounds.
- Material Pre-processing: If the input includes URLs or competitive analysis links, the agent scrapes the content to serve as context for the writing phase.
- Writing Phase: The Writer Agent generates the first version of the PRD, covering core features, information architecture, and user personas.
- Review Phase: A separate Reviewer Agent audits the draft for logic, completeness, and MVP feasibility, producing a detailed issue list.
- Feedback Loop: The Writer addresses the review comments in a subsequent version, repeating the cycle for the specified number of rounds (default is 3).
- Final Delivery: Once the rounds are complete or the reviewer issues a pass, all versions and a final document are saved to the user's local directory.
PRD Iterative Agent Setup
To use this feature within Openclaw Skills, no complex installation is required beyond ensuring your agent environment is active. You can trigger the workflow using natural language:
# Trigger the skill by stating:
"Help me write a PRD for [Product Name]"
# Or for specific multi-round refinement:
"Run PRD rounds"
You will be prompted to confirm your output path and model preferences (e.g., MiniMax for writing and Kimi for reviewing).
PRD Iterative Agent Data Schema & Taxonomy
The skill organizes its output into a structured directory using a strict versioning taxonomy to track the evolution of the requirements. Files are generated in Markdown format for easy integration into existing developer workflows.
| File Name Pattern | Description |
|---|---|
{product-slug}-prd-v{N}.md |
The PRD draft for iteration round N. |
{product-slug}-prd-v{N}-review.md |
The review report containing pass/fail status and issue lists for round N. |
{product-slug}-prd-v{final}-final.md |
The polished, final delivery version of the document. |
{product-slug}-prd-v{final}-final-review.md |
The final approval report from the Reviewer agent. |
PRD Iterative Agent Advanced Features
- Role-Based Model Specialization: Assign different LLMs for the Writer and Reviewer roles (e.g., high-speed models for drafting and reasoning-heavy models for auditing).
- Automated Web-to-PRD Conversion: Built-in capability to analyze competitive links or online articles as source material for the documentation.
- Customizable Iteration Depth: Configure between 2 to 5 rounds of refinement depending on project complexity and precision requirements.
- Contextual Platform Awareness: Automatically adjusts technical descriptions and interaction flows based on whether the target is a Mini-program, iOS, Android, or Web platform.
SKILL.md
Loading
Related Openclaw Skills
moses-stamp: MO§ES™ Governed Output
A document-level governance tool that embeds cryptographic integrity hashes and provenance metadata directly into AI-generated outputs.
Smart Compact
Smart Compact is a context preservation skill for Openclaw Skills that uses a 4-phase progressive strategy to rescue vital information before triggering context compression.
Smart Compact
Smart Compact is a four-phase progressive strategy for Openclaw Skills that scans, extracts, and preserves critical agent data before performing context compression.
Agent Harness Doctor
An automated diagnostic and repair utility designed to harden AI agent environments through multi-dimensional scoring and prioritized P0 fixes.
ppt-maker
A sophisticated AI skill designed to generate, preview, and export structured PowerPoint presentations from text or Markdown inputs.
1688 Shopkeeper
An AI-powered assistant for sourcing products on 1688 and automating store management across multiple Chinese e-commerce platforms.
Featured*
