Chalk Labs
Need help deploying jarvis-onboarding?
Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/johannsenlum-jarvis-onboarding
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/jarvis-onboarding
What's inside
- Skill name
jarvis-onboarding- Description
- The first-run experience for Jarvis. A guided, five-chapter conversation with tap-to-answer questions that builds the user's second brain live as they answer, sets up routines, and records skipped questions so onboarding can be resumed later. Use on first contact, or when the user says "onboard me", "finish onboarding", "set me up", or wants to add a new job, company or life area.
- License
- MIT
- Instructions
- 293 lines in SKILL.md · view on GitHub
Files (4)
- SKILL.md
- onboarding-json.md · references/
- telegram.md · references/
- templates.md · references/
SKILL.md (first lines)
# Jarvis onboarding
Onboarding is the user's first real experience of Jarvis. It should feel like meeting a sharp new
chief of staff who is already getting to work, not like filling in a form. Five short chapters, about
eight minutes, and every answer turns into something they can see.
## When to Use
- `me/onboarding.json` in the knowledge vault is missing, or its `status` is not `complete`.
- The user asks to (re)do or finish onboarding, or to add a new job, company, client list or life area.
- Resume mode: onboarding is complete but `pending` is not empty and the user agreed to finish it.
## The experience rules
**Chapters, with a progress line.** Open every chapter with one line like:
```
◆◆◇◇◇ Chapter 2 of 5 · Your world
```
(filled diamonds = chapters reached). Then one short, human sentence, then the questions.
**Tap, don't type.** Ask with your question tool (Claude Code's question tool, Hermes `clarify`). Options
go in its choices, not in the question text. Design for at most 4 questions per call and 2-4 choices per
question plus the free-text "Other" row. Recommended choice first. With no question tool, ask in plain
text with numbered options and accept free text.
**Open questions: one at a time, with a picker when you can.** Never send a numbered list of open
questions to answer in one reply. Ask each one on its own, one short line, and wait.
- If you can suggest likely answers, ask it with the question tool instead: the suggestions are the
choices and the user types anything else in "Other". Sources for suggestions: what the installer
saved, the existing-folders scan (client and project folder names), a connected calendar or contacts
(frequent names), and earlier answers. Example: "Who are your current clients?" multi-select
`Brightlabs`, `Nomi`, `Kopi Co` (from folder names), Other to add more.
- No suggestions available: ask in plain text, one question, and accept a free-text answer. Mention
"or skip".
- Batch only choice questions (up to 4 per call). Open questions never share a message.
**Show, don't tell.** Where your question tool supports option previews (Claude Code), use them for the
choices that change how you behave. Each preview is a short sample of what that choice feels like
(see Chapter 1). No previews available: put a four-to-six-word taste in the choice label instead.
**Build live.** After each chapter, create what the answers fill straight away (pages, folders, settings)
and show it as a small growing tree, only the new lines marked `+`:
```
knowledge/
+ work/northwind-studio/clients/brightlabs/
+ relationships/people/sam.md
```
One line under it on what that unlocks ("Mention Sam or Brightlabs any time and I'll have the context.").
The user should watch their second brain appear.
**React like a person.** One specific line in response to an answer, never generic praise ("Three
retainers and a half marathon. Busy year, I'll protect your mornings."). No lectures, no filler, no
exclamation-mark enthusiasm. Use their name now and then, not every message. Match the tone they chose
from the moment they choose it.
**Skip anything, pause any time.** Say once, in the opening: *"Skip anything (tap Other, type skip) and
say 'pause' whenever you like. I'll pick up where we left off."*
- A skip ("Skip for now", "skip", "later", "not now", or a timeout) is never re-asked in this session.
Add it to `pending` in `me/onboarding.json` (see `references/onboarding-json.md`). Nothing about a
skipped question is guessed or invented.
- "Pause": save, say how to resume ("say *finish onboarding*"), stop.
**Keep the machinery out of sight.** The user should see a conversation, not file edits.
- Save answers with the `jarvis_onboarding` tool (a `patch`), never by editing `me/onboarding.json`
with file tools (that prints raw JSON diffs). Create pages with `jarvis_write`. Don't announce saves.
- Save at the end of each chapter (`status: "in_progress"`, `current_step`), or every two or three
answers in a long chapter, so a dropped chat loses little.
- After an answer, reply in one message: a short reaction, anything it just created, and the next
question. No extra turns in between.
**Every answer does something visible.** When an answer names things (clients, people, a company),
create their pages right away and show them in the same message as the next question:
```
+ work/twiss/clients/brightlabs/overview.md
+ work/twiss/clients/nomi/overview.md
```
Then one line on what it unlocks. Answers that only set preferences get a one-line reaction instead.
**Never ask for keys, tokens or passwords in chat.** Tell the user where to put them.
## Procedure
Knowledge vault: the path in your identity ("Your second brain"), or `jarvis_status`. Read `SCHEMA.md`
there first. Folder layouts per answer are in `references/templates.md`. Follow them exactly.
**Resume mode.** If `me/onboarding.json` has `status: in_progress`, continue from `current_step`
(open with "Welcome back, we were on chapter N."). If it is `complete` with `pending` items, ask only
those, in step order, removing each from `pending` once answered.
**Came from the installer?** The `npx` wizard may already have saved name, role, tone, autonomy and work
type (`user.*`, `work`, and `current_step`). Never ask those again: confirm them in one line in the
opening and continue from `current_step`.
### Getting ready (silently, before the opening)
In one go, with no commentary, gather what lets you offer pickers instead of blank questions:
- `python3 <Jarvis dir>/bin/scan_folders.py --time-limit 20` (names and counts only, no contents):
likely client, project and company folder names. Keep the result for Chapter 2 and Chapter 5.
- Company hints: the domain of `git config user.email` (e.g. `twiss.io` → "Twiss"), and folder names.
- Is Obsidian installed (`/Applications/Obsidian.app`)?
### Opening (before Chapter 1)
If Obsidian is installed, start by offering (question tool): "Want to watch your second brain grow
while we talk?" `Yes, open it in Obsidian`, `Not now`. Yes: run
`open "obsidian://open?path=<url-encoded vault path>"`. If that doesn't show the vault (first time),
tell them: Obsidian → "Open folder as vault" → pick `knowledge`. Suggest the graph view.
Then three short lines, in the chosen tone if known:
1. Who you are and what you're about to do: "I'm <assistant name>. Give me eight minutes and I'll know
your work, your people and what matters to you, and I'll set up your mornings."
2. Everything is plain files they own, in `knowledge/`, readable in Obsidian.
3. The skip / pause rule.
### Chapter 1 · Meet (step `hello_role`)
Skip the questions already answered by the installer. Otherwise one batch:From JohannsenLum/jarvis, MIT.
More from JohannsenLum/jarvis
- 0.570score
jarvis-connections
JohannsenLum
Connect Jarvis to the user's services, one at a time, from the recommended set only - Google (Gmail, Calendar, Drive), Apple (Reminders, Notes, iMessage, Find…
SkillAFreshness: AClaude Code +9 - 0.565score
jarvis-models
JohannsenLum
Run a specific job on a specific AI model when the user asks - "use GPT Astra for this", "use Claude Opus for the proposal", "have Sonnet summarise these" -…
SkillAFreshness: AClaude Code +9 - 0.565score
brain-consolidate
JohannsenLum
Nightly "sleep" for Jarvis - read the day's conversations (Telegram, terminal, voice, dashboard), file durable facts into the vault, capture commitments…
SkillAFreshness: AClaude Code +9 - 0.565score
jarvis-settings
JohannsenLum
Change Jarvis's role (chief of staff, executive assistant, thinking partner, coach, life manager), tone (warm, formal, direct) or autonomy (ask-first…
SkillAFreshness: AClaude Code +9 - 0.565score
jarvis-frameworks
JohannsenLum
Run one of Jarvis's frameworks with the user (the 168-hour week budget, AIOO, declarations, deal cards) and file the results in their vault.
SkillAFreshness: AClaude Code +9 - 0.561score
brain-query
JohannsenLum
Answer a question from the user's knowledge vault with links to the pages used, and save valuable answers back as pages.
SkillAFreshness: AClaude Code +9 - 0.560score
jarvis-telegram
JohannsenLum
Choose how Jarvis appears in Telegram - one normal DM, topics inside the DM (Inbox, Work, Life, Briefings, General), or a group with forum topics - and set it…
SkillAFreshness: AClaude Code +9 - 0.560score
jarvis-import
JohannsenLum
Find existing folders on the Mac (Documents, Desktop, Downloads, iCloud Drive, Google Drive, Dropbox) with clients, projects and work data, ask the user which…
SkillAFreshness: AClaude Code +9 - 0.559score
decision-coach
JohannsenLum
Help the user make a decision using their own principles (me/principles.md) and chosen frameworks (me/frameworks/), then log it.
SkillAFreshness: AClaude Code +9 - 0.556score
write-proposal
JohannsenLum
Draft a client proposal or quote from a brief, using the client's page, services.md, pricing.md and past proposals.
SkillAFreshness: AClaude Code +9 - 0.553score
weekly-client-report
JohannsenLum
Draft a weekly status report per active client from project pages, meetings and log entries.
SkillAFreshness: AClaude Code +9 - 0.304score
jarvis-weekly-review
JohannsenLum
No description from the publisher yet.
SkillAFreshness: AClaude Code +9
Signals
- Stars
- 0
- Forks
- 0
- Open issues
- 2
- Last commit (this folder)
- Oct 1, 2026 (1 month ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/JohannsenLum/jarvis
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from JohannsenLum
- 0.584score
LinkedIn
JohannsenLum
Read LinkedIn profiles, companies, jobs and messages via your own browser session
ConnectorBFreshness: BLocal★ 4↓ 55/wkClaude Code +5 - 0.570score
Canvas API MCP
JohannsenLum
MCP server exposing the Canvas LMS REST API (1,116 endpoints) for AI assistants
ConnectorBFreshness: BLocal★ 3↓ 54/wkClaude Code +5 - 0.570score
jarvis-connections
JohannsenLum
Connect Jarvis to the user's services, one at a time, from the recommended set only - Google (Gmail, Calendar, Drive), Apple (Reminders, Notes, iMessage, Find…
SkillAFreshness: AClaude Code +9 - 0.565score
brain-consolidate
JohannsenLum
Nightly "sleep" for Jarvis - read the day's conversations (Telegram, terminal, voice, dashboard), file durable facts into the vault, capture commitments…
SkillAFreshness: AClaude Code +9 - 0.565score
jarvis-frameworks
JohannsenLum
Run one of Jarvis's frameworks with the user (the 168-hour week budget, AIOO, declarations, deal cards) and file the results in their vault.
SkillAFreshness: AClaude Code +9 - 0.565score
jarvis-models
JohannsenLum
Run a specific job on a specific AI model when the user asks - "use GPT Astra for this", "use Claude Opus for the proposal", "have Sonnet summarise these" -…
SkillAFreshness: AClaude Code +9
Done for you
Rather not set it up yourself?
Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.
See how we build AI agents →Read the case studies
Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.
- launches shipped
- 50+
- launches shipped
- media placements
- 500+
- media placements
- founders on every account
- 2
- founders on every account
Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block
Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.
Get jarvis-onboarding running in your business
Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.
Cite this page
jarvis-onboarding skill. Agent Mart. https://agentmart.dev/skills/johannsenlum-jarvis-onboarding (last updated 2026-10-24)