Kumu / Stage 0 · part 2

Stage 0, part 2 — Set up your workspace

Every course · do it once Part 1 is the GitHub account

Stage 0 has two parts, and this is the second. Part 1 is the account itself: create it, or sign in to the one you already have, and learn add → commit → push. This page is what you do next — build the repository every deliverable in every course here commits into.

Objective

One public repository, named for you, built to the portfolio-repo standard, with your instructor invited as a collaborator. That is the whole stage. No analysis happens here.

It belongs to the course, not to any one case or project, for a practical reason: the workspace has to be finished before the first deliverable is written, not half-built alongside it. Every engagement afterwards commits into it, and it is the one artifact here that outlives the course — a repository stood up properly at the start is a portfolio by the end of term; one assembled on the way to a deadline is a folder of homework. Build it once and every course on this site uses the same one.

Part 1 first: the account, and the three verbs

This page assumes you can already save work to a repo, so it does not re-teach the mechanics. If you cannot yet, that is Stage 0, part 1, and its Git mechanics walkthrough covers local versus remote, add → commit → push in the desktop app and on the command line, editing files straight on github.com, and the fixes for the mistakes everyone makes the first time. If you already have a GitHub account you can use that — one account, one portfolio, every course. If you don't, create one there and sign up with your hawaii.edu address: that is what unlocks GitHub Education.

Learn — why it is organized this way

The repository is organized by capability and engagement, never by course. A folder named for a class stops meaning anything the moment the class ends, and a reader who opens week3/final_v2.xlsx learns nothing about what you can do. capabilities/<capability>/ — marginal-analysis, pricing-power, fx-hedging, whatever you turn out to be able to do — names a capability; docs/, data/, and analysis/ hold the work that proves you exercised it.

Two mechanical facts about Git decide most of what goes wrong in this stage, and both are cheaper to know now than to discover later:

Git tracks files, not folders. An empty directory does not survive a push — it simply is not there when you look at github.com. Every directory in the skeleton needs at least one file, and a one-line README.md saying what belongs there is more useful to a reader than an empty folder anyway.
History is permanent. A file committed once stays in the repository's history even after you delete it. That is why .gitignore goes in before the first workbook, rather than after the first accidental commit of a ~$ temp file.

Do — set up the workspace

Do — check what you built

If you used the starter prompt, you handed a mechanical job to a model, which is the right call — and it means the verification is yours. Three things to check before you commit, each of which a generated skeleton gets wrong often enough to be worth looking for:

No folder is named after a course, a term, or a week. Models default to the shape they have seen most, and most public student repositories are course-shaped.
Every directory contains something. Otherwise Git silently drops it and your skeleton has holes you will not notice until a deliverable has nowhere to go.
The placeholder files are actually placeholders. A model asked for a portfolio will cheerfully invent a biography. Committing generic AI-written filler as your bio is obvious to any reader, and it is the first thing anyone sees.

AI boundary for this stage: everything here is a means to the work rather than the work itself, so AI may build all of it and you verify. That flips the moment your course starts — briefs, analyses, memos and reflections are yours, first draft included.

Deliverable

What	Where it goes
Portfolio repo, public, named for you	`github.com/{you}/{firstname-lastname}`
Bio + engagement index	`README.md`
AI conventions	`AGENTS.md` (+ `CLAUDE.md` pointing at it)
Resume + prompt log, started	`RESUME.md` and `prompt-log.md` at the root — rough is fine on day one
Exclusion rules	`.gitignore`
Instructor access	`adamwstauffer` invited as a collaborator

Repo is public — the URL opens in a private browser window without logging in
Named for you (firstname-lastname), not for a course
README.md holds a real 3–6 sentence bio and the start of an engagement index
AGENTS.md written in your own words; CLAUDE.md is the one-line pointer
RESUME.md and prompt-log.md exist at the root — rough is fine
.gitignore filters Office and OS temp files
Every skeleton directory holds at least one file
adamwstauffer invited as a collaborator
At least two commits, each with a message that says what changed