Skip to main content

Set Up Your Workspace

Create your private Jarvis workspace from the Minimum Viable Jarvis template without ever leaving your harness. One prompt, done.

This is the canonical setup flow, referenced by Claude Code Setup, Hermes Setup, Codex Setup, and the Supersuit Up Workshop. It assumes your harness is installed and you have the GitHub CLI (gh) authenticated.

Why a Template, Not a Clone​

We maintain the Minimum Viable Jarvis starter as a GitHub template repository. You create your own private copy from it. You never clone our repo directly.

Why it matters:

  • Your repo is private. On your GitHub account from the first commit.
  • You cannot accidentally push upstream. You are not a collaborator on the template.
  • Future template updates are easy to pull. We add a safety-disabled upstream remote so you can pull new skills and improvements without any risk of pushing back.
  • Git history is preserved. You can merge upstream updates cleanly.

Prerequisites​

  • A working harness: Claude Code, Hermes, or Codex
  • GitHub CLI installed (brew install gh on Mac, winget install GitHub.cli on Windows)
  • GitHub CLI authenticated: gh auth status should show a logged-in user. If not: gh auth login and follow the prompts.

The Setup Flow (One Prompt)​

Launch your harness in any directory (it does not matter which, since the first thing it will do is navigate to your projects folder):

  • Claude Code: claude
  • Hermes: hermes
  • Codex: codex

Paste this prompt:

Set up my private Jarvis workspace. Specifically:

1. Ask me what I want to name my workspace. Suggest names like my-jarvis, apex-os, <my-name>-command-center. This becomes both the GitHub repo name and the local folder name.
2. Ensure a sensible projects folder exists on my machine. Mac: ~/Documents/github-repos/. Windows: C:\Users\<me>\Documents\github-repos\. Create it if missing.
3. From inside that projects folder, run: gh repo create <workspace-name> --template Applied-AI-Society/minimum-viable-jarvis --private --clone. This creates a new private repo on my GitHub account from the template and clones it locally in one command.
4. cd into the new cloned folder. Add an upstream remote with a disabled push URL so I can pull future template updates without ever accidentally pushing back:
   - git remote add upstream https://github.com/Applied-AI-Society/minimum-viable-jarvis.git
   - git remote set-url --push upstream DISABLED
5. Report the final folder path to me.

Your harness will execute every step via the GitHub CLI and git. You stay in the terminal the whole time.

Then Let Onboard Take Over​

Do not personalize the files yourself. The template ships with an onboard skill that does this for you, and it will evolve over time as we improve the flow.

Exit your current session (/exit or Ctrl+D) and launch your harness again from inside the new workspace folder (not its parent). Skill auto-discovery and AGENTS.md routing only work when the current working directory is the workspace root.

cd <your-workspace-path>    # e.g., cd ~/Documents/github-repos/my-jarvis
claude                       # or: hermes, codex

If you prefer VS Code: use code <workspace-path> (pointing at the workspace folder, not its parent), then open the integrated terminal. Verify with pwd (Mac/Linux) or cd (Windows) before launching the harness. Then type / in the harness session. You should see the built-in skills (/onboard, /sync-with-upstream, etc.). If you don't, you're in the wrong folder.

The repo's AGENTS.md sees that user/USER.md does not yet exist and auto-runs the onboard skill on your first session in that workspace. It will:

  • Ask you to import any existing AI conversation history (ChatGPT exports, Claude conversations, docs about yourself)
  • Interview you to build user/USER.md: who you are, what you value, what you are building, your 90-day vision
  • Walk you through a strategic interview on the thing that matters most right now
  • Save the output to artifacts/ so you walk away with a real plan

You did not write any of this logic into the tutorial. It lives in the skill, which ships with the repo and updates with every template release.

Pulling Future Template Updates​

We will keep shipping new skills, improved scripts, and template refinements. To pull them into your workspace at any time, just ask your harness:

Sync with upstream.

The repo ships with a sync-with-upstream skill that:

  • Fetches from the upstream template
  • Shows you what is new (especially any new skills) before merging
  • Handles conflicts gracefully (your personal files are never touched)
  • Pushes the merge to your private origin

You cannot accidentally push to the template. The push URL is DISABLED and you are not a collaborator.

Troubleshooting​

  • gh: command not found: install and authenticate: brew install gh && gh auth login (Mac) or winget install GitHub.cli && gh auth login (Windows).
  • gh auth login asks a lot of questions: choose HTTPS, authenticate via browser, the defaults are fine.
  • Your harness cannot run gh: make sure your harness was launched in a shell that has gh on its PATH. Close and reopen the terminal after installing gh so the new PATH is picked up.
  • onboard does not auto-run: make sure you launched your harness inside the workspace folder, not from its parent directory. Harnesses read the nearest AGENTS.md / CLAUDE.md in the current working directory.
  • Claude Code does not see slash commands like /onboard (Windows only): the template ships a .claude/skills symlink that points at .agents/skills/. On Mac/Linux and on modern Git for Windows (with core.symlinks=true) this works out of the box. If your clone shows .claude/skills as a one-line text file instead of a working link, run the included fallback script from the workspace root: 
    powershell -ExecutionPolicy Bypass -File scripts/setup-claude-skills-windows.ps1
    This creates a directory junction from .claude/skills to .agents/skills/ so Claude Code discovers the skills. It does not require admin rights.
  • Hermes does not auto-register slash commands from your workspace: that is expected. By default Hermes only auto-registers skills from ~/.hermes/skills/. See the Hermes Setup doc for adding your workspace's .agents/skills/ to Hermes's external_dirs config if you want slash-command discovery. Natural-language triggers (via AGENTS.md routing) work regardless.

What You Now Have​

  • A private Jarvis workspace on your GitHub account, named whatever you wanted
  • A local clone in your projects folder
  • An upstream remote wired for pulling future template updates, with the push URL safety-disabled
  • A user/USER.md populated by onboard on your first session in the workspace
  • The full set of built-in skills ready to use: onboard, create-user-profile, think-through-it, process-braindump, prep-for-meeting, process-transcript, create-skill, sync-with-upstream

You are ready to build.

Further Reading​