Getting Started (V1)

V1 long-form archive. Pre-precision-pass version, preserved for engineers who prefer the dense narrative. New (V2) version: ../getting-started.html.

Install the Devkit (Once per Machine)

git clone https://github.com/amit-t/ai-devkit ~/Projects/Tools-Utilities/ai-devkit
cd ~/Projects/Tools-Utilities/ai-devkit
./install.zsh
source ~/.zshrc

You now have three global commands: init.wb, join.wb, update.wb (plus .dev and .cly variants to force Devin or Claude).

Requirements

gh CLI authenticated against GitHub (HTTPS or SSH with custom hostname — both supported).
git, rsync, python3, zsh.
devin CLI (default). claude CLI as fallback.

Initiate a Workbench (First Collaborator, E.g. QA)

mkdir ~/workbenches/wb-example && cd ~/workbenches/wb-example
init.wb

Devin asks: workspace label, epics in scope, repos to clone, roles, figma refs, optional MCPs, target GitHub org (defaults to your login). It creates a private repo from the ai-workbench template under your org, clones your service + automation repos into repos/, seeds project.conf and EPIC-PIPELINE.md, writes .github/CODEOWNERS with you listed, commits and pushes.

Share the resulting URL with your counterpart.

Join an Existing Workbench (Second Collaborator, E.g. Dev)

cd ~/workbenches
join.wb https://github.com/<your-org>/wb-example

Devin asks what additional repos you want cloned for your part of the work, appends them to project.conf, adds you to .github/CODEOWNERS, commits, pushes.

Pull Template Updates Later

cd ~/workbenches/wb-example
update.wb

One-way sync: fetches only template-owned paths (skills, scripts, CLAUDE.md, AGENTS.md, aliases.sh) from the ai-workbench upstream. Never touches your PRDs, specs, BDDs, etc.

First Session in a Workbench

cd ~/workbenches/wb-example
source aliases.sh    # add this line to ~/.zshrc for next time
claude .             # or: devin .

Try: /epic-intake EPIC-001 to pull your first epic and start a PRD.