🖥 The Full-Stack Master Reference

You're setting me up on a brand-new project and I want you to explain things simply. Do these
steps IN ORDER and DON'T build any features until I approve.
1) Look at anything already in this folder. Then ask me up to 4 short questions to learn: what
   I'm building, who it's for, the main features I want to start with, and any tools I already
   use. Wait for my answers.
2) Recommend the SIMPLEST tech stack that does the job (no over-engineering). Explain it in
   plain English, name the exact tools and where it will be deployed, and give one main pick
   plus one simpler alternative. If the tools I already use fit, say so. Then wait for me to
   say "go".
3) After I approve, create two Markdown files that reference each other:
   - CLAUDE.md: the permanent project brain. Most important rules at the top. Include the
     stack, the exact install/run/build/deploy commands, folder + code conventions, and a
     "Hard Rules" section. End with: "Current tasks and progress live in RUNBOOK.md."
   - RUNBOOK.md: the living task list. Start with a "RESUME HERE ->" line. Turn the features I
     named into small, independently testable slices in build order; each slice gets a
     checkbox, a one-line goal, the files it touches, and how I'll verify it. Tag slices
     P1/P2/P3. Add an "IDEAS / LATER" section. Near the top put: "Stack, commands, and rules
     live in CLAUDE.md - read it first."
4) Tell me in one sentence: from now on any new feature or idea goes into RUNBOOK.md (under
   IDEAS / LATER if we're not building it yet) - never straight into the code.
5) Show me both files and the single next thing to do. Don't build yet - wait for my go.

This is a project I've already built. Create a CLAUDE.md so you have context every session
without me re-explaining. First actually read the project - main files, config, package.json /
firebase.json if present, and the folder structure. Don't guess; base it on what's really here.
Then create CLAUDE.md in the root, most important rules at the top, including: what this project
is (1-2 plain sentences), the real stack you found, the exact run/build/deploy commands from the
config, the folder layout and conventions I follow, and a "Hard Rules" section with a couple of
blank bullets for me to fill in. End with: "Current tasks live in RUNBOOK.md." Show me the file
and flag anything you weren't sure about so I can correct it.

Read CLAUDE.md and RUNBOOK.md, then follow the "RESUME HERE" line. Plan the next unchecked P1
slice only (tell me the files and approach first), then build just that slice and show me it
working. Once I confirm, check it off in RUNBOOK.md with a one-line note, then commit with a
clear message, push, and deploy. Then stop and tell me the next slice.

Add this to the "IDEAS / LATER" section of RUNBOOK.md so we don't forget it, but don't build it
now: [describe the feature]. If it's something we should do next, turn it into a proper slice
with a checkbox, goal, files, and how I'll verify it - otherwise just log it.