Module 1Free45 - 60 min

New to Claude Code? Start with Module 0: Claude Code Basics

Write a Production-Ready PRD with AI

AI is most valuable when it helps you think better, not when it does all the thinking for you.

Most PMs spend 4 - 8 hours writing a PRD. Half that time is staring at a blank page, the other half is chasing down edge cases reviewers will flag. This module flips the process. Instead of writing then reviewing, you will think first with AI, then generate. The PRD Generator MCP server acts as a senior PM sitting across from you — asking the hard questions, surfacing blind spots, and drafting the document only after you have done the real thinking.

What You Will Learn

@-Mentions for Full Project Context

Reference product docs, research, and architecture so Claude understands your domain before asking its first question.

Socratic Questioning

Sharpen your problem statement through a structured Q&A that exposes gaps in your thinking before they reach stakeholders.

Multiple Strategic Approaches

Generate 2 - 3 distinct approaches and compare trade-offs before committing to a direction.

Multi-Perspective Feedback

Get instant reviews from an engineer, exec, researcher, and QA lead — before your actual review meeting.

Two Ways to Use This Module

Teacher Mode

6-lesson interactive course on AI-assisted PRD writing. Exercises, quizzes, and explanations — no prior knowledge needed.

Say: "I want to learn how to write PRDs with AI"

Usage Mode

6 production-ready MCP tools. Generate, validate, and review PRDs immediately — skip the teaching.

Say: "I need to write a PRD. Can you help me?"

Course Structure

Welcome to PRD Generation

15 min

What AI-assisted PRDs are, why they matter, and what you'll build in this course.

Context & Socratic Questioning

20 min

@-mentions, context loading, and the Socratic Q&A technique that sharpens your thinking.

PRD Structure & Templates

20 min

Templates for feature launches, API integrations, and redesigns. Learn when to use each.

Generating & Validating PRDs

20 min

Generate a full PRD from your answers, then validate it for completeness with automated scoring.

Multi-Perspective Review

20 min

Get feedback from Engineer, Designer, and QA perspectives — before your real review meeting.

Edge Cases & Polish

15 min

Surface edge cases specific to your product, iterate on feedback, and export a production-ready PRD.

Traditional vs AI-Partnership

See how each step of PRD writing changes when you bring AI into the process.

Research

Before

Dig through docs, Slack, and tickets manually

With AI

@-mention relevant docs, Claude surfaces patterns

Problem Framing

Before

Write, rewrite, rewrite again

With AI

Socratic Q&A until the problem is crisp

Requirements

Before

Brainstorm alone, miss edge cases

With AI

AI generates candidates, you curate

Validation

Before

Wait for the review meeting

With AI

Instant multi-perspective feedback

Edge Cases

Before

Reviewers find them for you (too late)

With AI

AI suggests them before you share

Final Document

Before

Copy-paste from scattered notes

With AI

Structured generation from your answers

The Four Core Techniques

These four techniques turn Claude from a content generator into a thinking partner.

Full Context via @-Mentions

Before writing a single word, give Claude the context it needs. The better the context, the sharper the questions and output.

What to @-mention:

Product strategy docs or OKRsUser research findingsExisting PRDs for related featuresTechnical architecture docsCompetitor analysis

Example prompt

@product-strategy.md @user-research-q4.md @api-architecture.md

I need to write a PRD for a new notifications center.
Can you review these docs and help me think through the approach?

Pro tip: Do not dump everything. Pick 3 - 5 docs that are directly relevant. Too much context creates noise, not clarity.

Socratic Questioning for Clarity

The PRD Generator asks 10 questions, but the real magic is in the follow-ups. A good PRD starts with a crisp problem statement — and most PMs skip this.

Problem Clarity

"Who exactly is affected?" "How do you know this is a problem?"

Vague problems lead to vague solutions

Solution Validation

"What alternatives did you consider?" "Why this approach?"

Prevents solutioning without exploring options

Success Criteria

"How will you know this worked?" "What does failure look like?"

Forces measurable outcomes, not vibes

Constraints

"What cannot change?" "What dependencies exist?"

Surfaces blockers early, not mid-sprint

Strategic Fit

"How does this connect to company OKRs?" "What are you saying no to?"

Ensures the feature is worth building

Pro tip: If you cannot answer a question clearly in conversation, you cannot write it clearly in a PRD. Use the struggle as signal.

Generate Multiple Approaches

Do not jump to the first solution. Ask Claude to generate 2 - 3 strategic approaches, then pick the best one.

Start Broad

"Give me 3 different ways to solve [problem]"

Best for: Greenfield features, unclear direction

Compare Trade-offs

"Compare build vs buy vs integrate for [need]"

Best for: Technical decisions, vendor selection

Phased Thinking

"What is the MVP vs v2 vs full vision?"

Best for: Large features that need scoping

Example prompt

I am considering three approaches for the notifications center:
1. Real-time WebSocket push
2. Polling with smart batching
3. Email digest only

Help me compare these on: user experience, engineering effort,
scalability, and time-to-ship.

Multi-Perspective Feedback

Before sharing your PRD with stakeholders, get feedback from multiple perspectives. Claude can role-play different reviewers.

Engineering Lead

Feasibility, technical debt, scalability, timeline realism

VP Product / Exec

Strategic alignment, ROI, opportunity cost

User Researcher

User needs, usability gaps, accessibility

QA Lead

Edge cases, error states, testability

Example prompt

Here is my PRD draft. Review it from three perspectives:

1. As a senior backend engineer — flag technical risks
2. As VP Product — challenge the strategic rationale
3. As a user researcher — identify gaps in user understanding

Be specific and critical. I would rather fix problems now than in review.

Real-World Walkthrough

The full flow for a PM writing a PRD for an In-App Notifications Center.

Context Loading

2 min

@-mention your product roadmap, user research, and any existing technical spikes. Tell Claude what you are building and ask it to help you think through the approach.

Socratic Exploration

15 min

Claude asks about the problem and you clarify that 34% of approvals are delayed 48+ hours. It pushes on users, metrics, and constraints — you commit to "median response time under 4 hours" and discover a WebSocket scaling gap.

Approach Generation

10 min

Claude proposes 3 delivery approaches: WebSocket push, polling with batching, and email digest. You compare trade-offs and pick WebSocket with polling fallback.

PRD Generation

5 min

Claude generates a structured PRD from all the thinking above. The document includes all sections, pre-filled with real content from your conversation.

Multi-Perspective Review

10 min

Engineering flags WebSocket scaling risk. UX catches missing mobile experience. The exec reviewer questions ROI, prompting you to strengthen the metrics section with dollar impact.

Validation and Export

3 min

The PRD scores 90% (Grade A). You add a missing rollback trigger, then export to Markdown and paste into Notion. Total time: about 45 minutes for a PRD that would have taken 6 hours.

Best Practices

Do

Think first, generate second

Use the questioning phase to sharpen your thinking, not to outsource it.

Be specific with metrics

"Improve engagement" is worthless. "Increase 24-hr read rate from 45% to 80%" is actionable.

Challenge the AI's suggestions

If an edge case does not apply, say why. If a question feels off, push back.

Iterate on the output

The first draft is never final. Use validate_prd to find gaps, then fill them.

Save your best prompts

Build a personal library of prompts that produce great results.

Do Not

Accept the first output blindly

AI-generated does not mean production-ready. You are the PM, not Claude.

Skip the problem statement

It is tempting to jump to requirements. Resist. A crisp problem is 60% of a good PRD.

Use AI to avoid hard thinking

If you cannot explain the "why" verbally, the PRD will not be convincing.

Forget your audience

Engineers read PRDs differently than execs. Tailor the depth accordingly.

Generate PRDs for features you do not understand

AI cannot compensate for a PM who has not talked to users.

Troubleshooting

PRD feels generic, not specific to my product

Cause: Not enough context provided upfront.

Fix: @-mention 3 - 5 relevant docs before starting. Include user research, product strategy, and technical constraints. The more specific your input, the more specific the output.

Edge case suggestions are not relevant

Cause: Product description is too vague.

Fix: Be specific about your product type, user base, and technical stack. "B2B SaaS with 12K MAU, WebSocket-based real-time features" produces better edge cases than "a notification feature."

The process feels too long

Cause: Trying to answer every question perfectly in one pass.

Fix: It is OK to answer "TBD" and come back later. Generate the PRD with what you know, then use validate_prd to identify the gaps worth filling.

Stakeholders want a different format

Cause: Your team uses a custom PRD template.

Fix: Edit the templates in the templates/ directory to match your team's format. The MCP server will generate PRDs using your structure.

I do not know how to answer some questions

Cause: You need more research before writing the PRD.

Fix: That is actually the point. If you cannot answer "How will you measure success?", that is a sign you need to talk to your data team before writing the PRD. The questions expose gaps in your preparation.

Quick Reference

The six MCP tools included in this module.

list_templates

Shows 3 PRD template types (feature launch, API integration, redesign).

Start here to pick the right template.

get_template

Returns the interactive questionnaire for your chosen template.

After picking a template.

generate_prd

Creates a full, structured PRD from your answers.

After answering questions.

validate_prd

Scores PRD completeness on an A - D scale.

After generation and before sharing with stakeholders.

suggest_edge_cases

Lists edge cases specific to your PRD type and product.

During or after PRD generation.

review_prd

Multi-perspective review from Engineer, Designer, and QA viewpoints.

After generation to catch blind spots before real review.

Download and Setup

Clone the module from GitHub, install dependencies, and add the MCP server to your Claude Code config.

View on GitHub

Terminal

git clone https://github.com/anmolgupta824/ai-native-pm.git
cd ai-native-pm/modules/module-1-prd
npm install && npm run build

Module 0 — Claude Code Basics

Module 2 — AI Image Generation