agentleFS
ManceRayder42ManceRayder42/jarvis-os/skills/setup-for-me/SKILL.md

setup-for-me

Skill5 starsChanged 43 days agoMITView on GitHub

Do the Jarvis OS setup for the user instead of sending them to the setup page — create the memory, switch skills, install and connect tools, and hand back only the steps a person must do. Use when the user says "set me up", "set it up for me", "finish the setup", "configure Jarvis", "what's left to set up", "תגדיר לי", "תסיים את ההתקנה", right after the plugin was installed, or when the session-start note says the memory isn't set up yet.

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 3 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Every how is a complete, absolute command: run it exactly as given, unquoted, from any directory.

    line 28
  2. Distinctive:

    Never ask the user to paste a key, token or password into the chat, and never put one in a command line.

    line 40
  3. Very specific:

    Never run a --from-clipboard command yourself, and never read the clipboard.

    line 47
  4. Distinctive:

    Don't open the setup page unless they ask for it — the point is that they don't have to.

    line 60
  5. Very specific:

    Don't install anything marked ask without their yes.

    line 62
  6. Distinctive:

    Don't retry a failed install in a loop.

    line 63

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Setup, done by you

Everything on the /jarvis-setup page has a command. Run them yourself; the person should only do what genuinely needs a person.

The tool: node "${CLAUDE_PLUGIN_ROOT}/setup/agent.mjs" <command>. It drives the same setup server the page uses, so results are identical. Add --json to status, plan, skills or tools for machine-readable output.

Steps

  1. Read the plan: plan --json. Every item has who:
    • agent — run its how command now, no need to ask.
    • ask — it installs software on their computer. List these in one message ("I'd install GitHub CLI, qmd and ffmpeg — OK?") and run the ones they approve. One question for the batch, not one per tool.
    • person — a key, a sign-in, or an app only a human can operate. Collect these for the end. Items with recommended: false or optional: true are not done by default: list them once at the end under "Also available", one line each. An item's then is a follow-up for the person (e.g. an OAuth sign-in via /mcp) — pass it on with that item. Every how is a complete, absolute command: run it exactly as given, unquoted, from any directory. A how starting with ! is for the person.
  2. Do the agent items, starting with setup if the memory doesn't exist (it also connects the recommended free tools, so don't connect those again). Run independent ones back to back without narrating each.
  3. Run plan again to confirm what actually changed. Report from the command output, never from what you expected to happen.
  4. Hand over the person items as a short numbered list, each with the exact thing to type.

Keys and sign-ins — never through you

  • Never ask the user to paste a key, token or password into the chat, and never put one in a command line.
  • For a connection that needs a key, give them the plan's how line as is: they copy the key, then type it in Claude Code — ! node "<plugin root>/setup/agent.mjs" connect <id> --from-clipboard. The key is read from the clipboard and never printed. (In a normal terminal, without the flag, it asks with typing hidden.)
  • Never run a --from-clipboard command yourself, and never read the clipboard.
  • Sign-ins (gh auth login, notebooklm login, /mcp for OAuth connections) are the person's: give the exact command.

Report

Three short parts: what you did (from the output), what they need to do (numbered, copyable), and anything that failed with its exact error. Then remind them of the one habit: say "done" at the end of a session.

Don't

  • Don't open the setup page unless they ask for it — the point is that they don't have to.
  • Don't install anything marked ask without their yes.
  • Don't retry a failed install in a loop; show the error and the manual line the tool printed.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.