Turn an idea into a live product in one weekend with a SPEC, a PLAN and one /goal per milestone.
Most weekend projects die on Saturday afternoon, half built, because nobody wrote down what "done" meant. This guide is for people who have an idea and a free weekend, and who know Claude Code well enough to run a prompt. You will spend Friday evening turning the idea into a short spec and a plan, Saturday building one milestone at a time with /goal, and Sunday deploying and launching. The attached kit has every prompt and template, so you fill in blanks instead of starting from nothing.
What you will have at the end
A SPEC.md with 3 to 7 requirements, each with an acceptance check, and a written list of what you are not building.
A PLAN.md that splits the work into milestones, each ending in a committed, tested state.
A /goal prompt per milestone that keeps Claude working until a test proves the milestone is done.
A live URL that passes a smoke test, and a launch checklist you ticked by hand.
Before you start
Claude Code installed and signed in. New to it? Do the free start guide first.
A stack you have shipped with before. A weekend is the wrong time to learn a new framework.
An account with a host you know how to deploy to.
The attached weekend-kit/ folder. Copy SPEC-template.md and PLAN-template.md into an empty project folder before you start.
A rough schedule:
When
What
Output
Friday evening
Interview, spec, plan
SPEC.md, PLAN.md
Saturday
Build, one goal per milestone
Tested commits
Sunday morning
Review and deploy
Live URL
Sunday afternoon
Launch
First users
Step 1: Let Claude interview you into a spec
Open your project folder, start claude, and paste the interview prompt from 1-interview-prompt.md, with your own idea on the first line:
I want to build: a booking page where yoga students can reserve a spot in a class from their phone.
I have one weekend. Interview me until you could write the spec, then write it to SPEC.md using the structure in SPEC-template.md.
Rules for the interview:
- Ask a few questions at a time, multiple choice where you can, with your recommended answer marked.
- Push me to cut scope. For every feature, ask whether the first user needs it on day one.
- Ask how a user would know each feature works, so every requirement gets an acceptance check.
- Ask what I will not build this weekend, and write that down too.
- Ask which stack and host I already know. Prefer boring choices I have used before.
- Stop asking when you can fill every section of the template. Then write SPEC.md and show me a summary.
Our prompt adds four rules that matter for a weekend:
Cut scope on every question. Your first list will be too long. "Does the first user need this on day one?" shrinks it fast.
An acceptance check for every requirement. "Students can book a class" is a wish. "Booking a class with a valid email returns a confirmation and lowers the free spots by one" is something a test can prove. You need that on Saturday, because every /goal will point at these checks.
A "Not this weekend" list. Writing down what you will not build stops scope creep on Saturday, both yours and Claude's.
A boring stack. An hour fighting a new tool is an hour not spent on the product.
When SPEC.md exists, read it slowly and fix anything wrong by hand. You want 3 to 7 requirements with ids and acceptance checks, and real commands in the "Checks" section. Commit it. This file is the contract for the weekend: every goal, review and test points back to it.