Skip to content

new-project.md

Latest commit

 

History

History
72 lines (48 loc) · 3.84 KB

new-project.md

File metadata and controls

72 lines (48 loc) · 3.84 KB
Raw

New project

Set a project up with flow init, and choose where its tickets are kept.

Table of contents

Set up a project

Type flow init in the project's folder. With no git repository, it makes one. It asks for the ticket prefix, offering the folder's name, and Enter keeps it:

Ticket prefix, so its tickets read shop-1, shop-2: shop

In an empty folder it then writes the project's files at once:

$ flow init --prefix shop
.flow/: a new flow branch, sharing no history with the code, checked out
.gitignore: .flow/ added
wrote: .claude/settings.json
wrote: .gitignore
wrote: .uncommitted-include
wrote: CLAUDE.md

set up: ~/code/shop is on entry 1. Nothing in the code is committed: CLAUDE.md, .gitignore and .claude/ wait for your next commit.
A plugin or skill switched off on this computer is switched on here with: claude plugin enable <id> --scope project, or flow skills on <name>

The last line is for a plugin or skill your computer's setup switched off: switch it on here once the project uses it.

CLAUDE.md holds what every session in this project needs. /flow:groundwork fills in its ## Project section, and sessions add the rest. Commit the files when you choose.

Where the tickets live

  • The project's flow branch, the default: a branch in the project's own repository, sharing no history with your code, checked out at .flow/. A ticket never gets merged into the code, and everyone who can read the repository reads the tickets.
  • Your Flow home, with flow init --private: only you read them, and flow sync carries them to your other computers.

On the branch, every change to a ticket is committed and sent to GitHub by itself. A new ticket gets its number only once GitHub has it, so offline, flow new refuses.

A public repository makes flow init ask, since the branch would publish the tickets:

This repository is public. Where should its tickets live?
  1. Your Flow home: private, on all your machines
  2. The project's flow branch: PUBLIC, anyone can read them
Type 1 or 2: 1

flow store says where the tickets are, and flow store private or flow store branch moves them. A move to your Flow home leaves the branch as it was, and prints how to delete it.

A folder that already has files

flow init opens the setup session, a Claude Code session that reads the project's rule files, docs, code and Claude Code memory before anything changes. Where the folder holds no rules for an agent, it asks first. -y answers yes.

The session writes every change into one form, migration.md, and stops. Nothing changes before you approve it.

  • What moves into Flow's files is counted: 9 rules → CLAUDE.md. It takes no box: edit the new file under files/ before you say go.
  • The code wins over the docs. A doc the code contradicts becomes a ticket.
  • Anything working against Flow has a box under ❌ Removed: works against Flow's rules.
  • A plugin or skill your computer's setup switched off is switched on where the code uses it: supabase plugin: this project uses @supabase/supabase-js.

Start claude again once the session ends, so the new rules load. flow restore project undoes the whole setup: Install.

On your other computer

Clone the project, and type flow init in it. Flow sees it is already set up, and asks nothing: it checks the flow branch out at .flow/, or links .flow/ to the tickets in your Flow home. A teammate's computer does the same.