AI Product Development Toolkit

A collection of prompts, instructions, and command-style skills that help you use AI assistants (like Claude Code or Codex) to plan, build, and launch a web app in a structured way.

The project is MIT-licensed and organized into four main parts:

---

1. Planning Prompts

A set of 6 conversation-driven prompts (for any AI chat tool) that interview you first instead of instantly writing a document full of guesses. They ask a few targeted questions, confirm their assumptions with you, and only draft a plan once you agree. Each step feeds the next.

Step	Prompt → What you get
1	User research questionnaire → evidence-based research summary for your PRD
2	PRD generation → product requirements doc (goals, users, requirements, constraints)
3	UX and user flow → user journeys, screens, interface requirements
4	MVP scope → core hypothesis, problem to solve first, short in/out feature list
5	MVP build plan → full development plan OR a lean "ultra-lean" build spec
6	Test plan → test scope, test types, environments, entry/exit criteria

How to use: copy the prompt text (with the "Copy raw content" button), fill in bracketed placeholders, paste into any AI chat, and answer its questions. Most people start at step 2; do step 1 first if you can interview potential users. You can start at any step.

---

2. AI App Starter

Instructions (not code) your coding agent follows to build a Next.js + Supabase + Vercel web app with you, step by step, from idea to launch. You don't need to know the process — the agent leads and you answer in plain words.

• The starter is 100% instructions: stage-by-stage steps, permission rules, and test/CI requirements. The agent writes the app, its tests, and its checks itself.
• Setup: open a new empty folder in Claude Code (desktop or terminal) or Codex, paste a one-line clone command, then just say "what's next?" and "continue".
• If you already have a PRD or MVP scope from the planning prompts, the starter will ask for them and skip those interviews.

Safety & workflow features

• Can't touch production directly — it works only on your computer, with a local database and stand-in payment/email services. It never holds hosted keys (not even test keys), and you push and start production deploys yourself.
• Leads in plain language — asks only what the next piece of work needs, does every local step itself, hands you one command or a few clicks for the rest.
• Stays organized — one plan, one decisions file, one list of known issues; new ideas go into the plan rather than derailing work.

Requirements

• To start: Node.js, Git, Docker.
• Once the app goes online: GitHub, Vercel, and Supabase accounts.
• Tested: with Claude Code desktop up to the first pieces of a web shop, and on a live site with hosted staging/production.
• Not tested yet: Codex, and full launch.

---

3. Skills

If you use Claude Code or Codex, you can install the planning prompts, the AI App Starter, and a video maker as named commands (e.g. /prd, /ux, /mvp-scope, /nextjs-supabase-vercel). One line pasted into your agent installs them all (or you can install only a few by name).

Skill	Command	What it does
User research	/user-research	Plans a questionnaire, turns responses into PRD input
PRD	/prd	Interviews you, then drafts product requirements
UX & user flows	/ux	Maps screens and flows from your PRD
MVP scope	/mvp-scope	Picks the smallest first version worth building
MVP build plan	/mvp-plan	Full or ultra-lean build plan
Test plan	/test-plan	Drafts test scopes and entry/exit criteria
AI App Starter	/nextjs-supabase-vercel	Sets up an empty folder with the starter
Video	/remotion-video	Makes a short product video in code (Remotion), after you approve the idea + storyboard

Notes:

• In Codex, type $ instead of / (e.g. $prd).
• Skills run only when you call them; each planning skill runs its corresponding planning prompt unchanged.
• The video skill needs Node.js (and usually Python for sound). It installs Remotion itself after you approve its plan. There's a demo video embedded in the README.
• To update skills, paste the same install line again.

---

4. Agent Configurations

Ready-made working instructions for Claude Code, based on Anthropic's documentation, meant to be merged into your own setup (not replace it):

Config	What it adds
Claude Sonnet 5.5 rules (latest)	Opus 5.5 file + 3 short additions: no unrequested extras, stop after giving a plan, run a real check before saying done.
Claude Opus 5.5 rules	A CLAUDE.md for clear scope, concise replies, and up-to-date docs lookups (via Context7, a free docs tool).
Claude Opus 5 rules	Same idea, written for Opus 5.

---

Good to Know (from the README)

• AI output is a draft — review requirements, plans, and code before relying on them.
• Don't paste private information your AI tool doesn't need.
• Feedback via GitHub Issues (form asks which prompt/skill/starter you used); questions via Discussions; contributions welcome via CONTRIBUTING.