Validate ideas
before you build them.

Product Kit is a free, open-source toolkit that brings structured product thinking to Claude. 9 guided steps. 7 utility commands. 1 validated spec.

Building is easy. Building the right thing is hard.

Claude Code. Codex. Cursor. Windsurf. You can go from idea to working app in hours. But easier to build doesn't mean smarter to build.

Most people skip straight to code. No user research. No problem validation. No spec. The result?

Polished solutions for problems that don't exist.

Before you ask "Can we build it?" — ask "Should we build it?"

What if the same AI tool you use to build could first help you validate the idea?

How It Works

Product Kit guides you through 9 steps. Each step reads the previous one. Nothing gets skipped.

0Landscape
1Constitution
2Users
3Problem
4Assumptions
5Validate
6Solution
7Prioritize
8Spec

Plus seven utility commands — Clarify, Analyze, Bootstrap, Audit, Learn, Tech Review, and Stories — that you can run anytime.

Setup

Product Kit works two ways. Pick the one that fits your style.

No terminal required

Use Product Kit directly in Claude with the Cowork plugin.

  1. Download product-kit-plugin.zip
  2. In Claude, go to Plugins → + → Upload plugin
  3. Select the zip file
  4. Done — type /product-kit:users to start

What you get: All 9 guided steps plus utility commands (Clarify, Analyze, Bootstrap, Audit, Learn, Tech Review, Stories). Great for product thinking without leaving Claude.

For terminal users

Requires Node.js 18+ and Claude Code.

npm install -g productkit
productkit init my-project
cd my-project
claude

Then run the slash commands in order: /productkit.constitution, /productkit.users, etc.

Options: Use --existing to add to a current project, --minimal to skip constitution, or --artifact-dir docs/product to store artifacts in a custom directory.

CLI extras: The CLI includes everything in the plugin, plus terminal tools for power users: productkit status (artifact progress), productkit export (bundle all artifacts or --stories-csv for Jira/Linear import), productkit diff (changes since last commit), productkit doctor (project health check), and workspace support for multi-project setups.

The 9 Steps

Each step is a guided conversation. Claude asks the questions, pushes back on vague answers, and writes a markdown artifact when done.

0

Landscape

"Front-load the context so every future step is smarter."
/productkit.landscape

What Claude asks you

  • What does your company do?
  • Who are your customers today?
  • What's the competitive landscape?
  • What are your team and tech constraints?

What it produces — landscape.md

  • Mission & Vision
  • Company & Product Portfolio
  • Target Market & Domain
  • Team & Constraints
1

Constitution

"Make the hard tradeoffs before you write a line of code."
/productkit.constitution

What Claude asks you

  • What does your product stand for?
  • What will you never do?
  • What principles guide decisions?

What it produces — constitution.md

  • Vision
  • Core Principles
  • Non-Negotiables
  • Decision Framework
  • Prior Research & Decisions
2

Users

"Everyone" is not a user.
/productkit.users

What Claude asks you

  • Who specifically has this problem?
  • What's their day-to-day like?
  • How do they solve it today?

What it produces — users.md

  • Primary persona (specific, detailed)
  • Secondary persona
  • Key insight
3

Problem

"We need an app" → "What problem does that app solve?"
/productkit.problem

What Claude asks you

  • What specific pain are users feeling?
  • How often does it happen?
  • What's the root cause?

What it produces — problem.md

  • Problem statement grounded in evidence
  • Impact (frequency + severity)
  • Root cause
  • Scope (in + out)
4

Assumptions

What are you betting on without knowing it?
/productkit.assumptions

What Claude does

  • Reads your artifacts so far
  • Surfaces hidden risks
  • Identifies what you're taking for granted

What it produces — assumptions.md

  • Critical assumptions (low confidence, high impact)
  • Validation plan for top risks
  • Recommended next steps
5

Validate

Don't guess. Go find out.
/productkit.validate

What Claude does

  • Reads your assumptions
  • Generates interview scripts & survey questions
  • Creates CSV templates for data collection
  • Analyzes raw data when you return

What it produces — validation.md

  • Validated vs. invalidated assumptions
  • Evidence with quality assessment
  • Interview & survey instruments
  • Gates: blocks solution until critical assumptions are validated
6

Solution

NOW you can brainstorm what to build.
/productkit.solution

What Claude asks you

  • What approaches have you considered?
  • How does each trace to a user need?
  • What are the tradeoffs?

What it produces — solution.md

  • Options considered (with pros/cons)
  • Recommended solution for v1
  • What's deferred to v2+
  • Key risks
7

Prioritize

"What's the smallest thing you can ship that solves the core problem?"
/productkit.prioritize

What Claude does

  • Scores features: Impact x Confidence / Effort
  • Draws the v1 line
  • Justifies every cut

What it produces — priorities.md

  • Ranked feature table with scores
  • Must-haves vs nice-to-haves
  • What's deferred and why
8

Spec

Everything synthesized. Every claim traced. Ready to build.
/productkit.spec

What Claude does

  • Reads all 8 previous artifacts
  • Synthesizes into one document
  • Links every claim to its source

What it produces — spec.md

  • Overview, principles, users, problem
  • Solution with acceptance criteria
  • Risks, open questions, success metrics

Utility Commands

Run these at any stage to check your work.

Clarify

/productkit.clarify

Find contradictions across your artifacts. Run this anytime something feels off. Claude reads everything you've written and flags inconsistencies.

Analyze

/productkit.analyze

Get a score out of 10. Strengths. Gaps. Contradictions. Verdict: "Ready to build, or needs more work?"

Bootstrap

/productkit.bootstrap

Already have a codebase? Bootstrap scans your project and auto-drafts all missing artifacts in one session. Great for existing projects.

Audit

/productkit.audit

Compare your spec against the actual codebase. Surface gaps, scope creep, and unmet acceptance criteria. Get prioritized recommendations.

Learn

/productkit.learn

Index your knowledge/ directory into a compact knowledge-index.md summary. All other commands read this index instead of scanning raw files. Re-run whenever you add new research.

For teams, organize knowledge/ by type: knowledge/interviews/, knowledge/analytics/, knowledge/competitors/. In a workspace with multiple projects, shared research goes in the workspace-level knowledge/ directory.

Tech Review

/productkit.techreview

Review your spec against the actual codebase. Assess feasibility, estimate effort, and flag items that need engineering input — before you write stories.

Stories

/productkit.stories

Break your spec into user stories grouped by epic, with acceptance criteria, t-shirt estimates, and dependency mapping. Reads techreview.md for better estimates.

What You End Up With

After running the full workflow, your project contains:

my-project/
├── landscape.md      # Company & domain context
├── constitution.md   # Product principles
├── users.md          # User personas
├── problem.md        # Problem statement
├── assumptions.md   # Prioritized assumptions
├── validation.md    # Validation results & scripts
├── solution.md       # Chosen solution
├── priorities.md     # Ranked feature list
├── spec.md           # Complete product spec
├── techreview.md     # Technical feasibility review
├── stories.md        # User stories by epic
├── audit.md          # Spec vs codebase gap analysis
├── knowledge-index.md # Summary index of research
└── knowledge/        # Raw research files

Use --artifact-dir during init to store artifacts in a subdirectory (e.g. docs/product/) instead of the project root.

13 artifacts. 1 validated product idea. Hand spec.md to Claude Code and build the right thing.

See It in Action

Walk through the entire Product Kit workflow with a real example — building Streak, a habit tracker for people who keep starting and quitting. Extended dialogues, full artifacts, common mistakes, and tips for every step.

Start the Walkthrough

Ready to validate your idea?

Free. Open source. MIT license.