smb-onboard
anthropics/knowledge-work-plugins · Agent Skill
Helps SMB owners get started by connecting tools, demonstrating value, and gathering business context for personalized assistance.
Jump to install ↓Overview
| Publisher | |
|---|---|
| Repository | anthropics/knowledge-work-plugins ↗ |
| Skill folder | smb-onboard |
| Stars | 23,908 |
| Bundled files | 4 |
| License | Apache-2.0 |
| Last updated | May 12, 2026 |
| Links | Skill source ↗ · Repository ↗ |
Mozgi check
What we found in the skill files. Details come with a subscription.
Installation
A skill is a folder with a SKILL.md file. Pick your agent and run the commands in a terminal (macOS, Linux or Git Bash on Windows) from your project root.
Copy the prompt into Claude, Codex or another assistant: it will review the skill and install it for you.
Install the “smb-onboard” skill from mozgi.io (https://mozgi.io/skills/smb-onboard/) for me. 1. Download the archive: https://mozgi.io/skills/smb-onboard/download If it fails, take the skill folder from the source: https://github.com/anthropics/knowledge-work-plugins/tree/main/small-business/skills/smb-onboard 2. Read SKILL.md and the other files. Check the skill is safe: no hidden commands, no sending data elsewhere, no scripts doing something other than described. If anything looks suspicious, stop and tell me. 3. Put the “smb-onboard” folder into your agent's skills folder: .claude/skills/ for Claude Code, .agents/skills/ for Codex, .cursor/skills/ for Cursor, .gemini/skills/ for Gemini CLI. Inside the project if we are in one, otherwise in the home folder (~/.claude/skills/ etc.). 4. Make sure <skills folder>/smb-onboard/SKILL.md exists. 5. Briefly tell me what this skill does and offer to try it on my task. If you cannot do a step yourself, tell me what to click or type.
git clone --depth 1 --branch main https://github.com/anthropics/knowledge-work-plugins.git /tmp/knowledge-work-plugins mkdir -p .claude/skills cp -r /tmp/knowledge-work-plugins/small-business/skills/smb-onboard .claude/skills/smb-onboard
Restart Claude Code — the skill appears in the list; call it manually with /smb-onboard. For all projects use ~/.claude/skills.
git clone --depth 1 --branch main https://github.com/anthropics/knowledge-work-plugins.git /tmp/knowledge-work-plugins mkdir -p .agents/skills cp -r /tmp/knowledge-work-plugins/small-business/skills/smb-onboard .agents/skills/smb-onboard
Codex looks for skills in .agents/skills from the current folder up to the repo root; call one with /skills or $smb-onboard. For all projects use ~/.agents/skills.
git clone --depth 1 --branch main https://github.com/anthropics/knowledge-work-plugins.git /tmp/knowledge-work-plugins mkdir -p .cursor/skills cp -r /tmp/knowledge-work-plugins/small-business/skills/smb-onboard .cursor/skills/smb-onboard
Cursor picks the skill up on start; call it with “/” in Agent chat. Cursor also reads .claude/skills and .agents/skills. For all projects use ~/.cursor/skills.
git clone --depth 1 --branch main https://github.com/anthropics/knowledge-work-plugins.git /tmp/knowledge-work-plugins mkdir -p .gemini/skills cp -r /tmp/knowledge-work-plugins/small-business/skills/smb-onboard .gemini/skills/smb-onboard
Gemini CLI offers to activate the skill when a task matches — you approve it. For all projects use ~/.gemini/skills.
Read SKILL.md and any companion files before deciding whether to install. Click a file to preview it.
SKILL.md contents
Original author text · read-onlySMB Onboard
Quick start
Four moves: connect two tools → run one recipe → capture business context → set a weekly rhythm. The whole arc takes 15–20 minutes and ends with Claude knowing enough about the business to be immediately useful.
User: "get me started"
→ Assess what's already connected; pick the best 2 tools to connect first
→ Guide connection of each tool (one at a time)
→ Run one recipe against live data to prove value
→ Ask 5 business questions one at a time; store answers to persistent memory
→ "Each Monday, say 'weekly check-in' — I'll pull your numbers and flag anything urgent."
Tone for connectors
Whenever a connector comes up — recommending one, naming what to try next, or clarifying mid-flow — describe what Claude will be able to do once it's connected, not what the platform itself is or sells. Owners already know what HubSpot, QuickBooks, Gmail, and Calendar do; they don't need a product pitch from us.
- Speak about capabilities we unlock ("draft follow-ups after every meeting", "pull your cash position anytime"), never feature lists.
- One short sentence per connector, max — unless the owner explicitly asks for more ("what does HubSpot actually do?"), in which case answer that directly.
- This rule applies to every step below.
Workflow
1. Welcome and assess. Greet the owner briefly. Check which connectors are already active. If a ## Business context block already exists in the owner's CLAUDE.md or memory, read it first — then skip to the return-session path: show the existing profile, ask what's changed, update only the fields that changed. Do not re-interview from scratch.
2. Pick two functions, then check what the owner uses. Ask: *"What are your biggest day-to-day headaches — money, customers, scheduling, or getting organized?"* Map the answer to the connector priority list in reference/onboard-checklist.md ↗.
Name the two functions we want (e.g. "a place to track customers and deals" and "your inbox") — not the platform features. One short sentence each, max. Then ask whether the owner uses a supported tool for each.
For each function, branch:
- Owner uses a supported connector (e.g. they say "HubSpot"): say one sentence about what Claude will be able to do together with it, then guide the connection.
- Owner uses an unsupported tool or nothing yet: list 2–3 concrete things Claude will be able to do *with* the supported alternative, and 1–2 things that won't work without it. Then let the owner decide whether to switch or add it. Do not push.
Connect one tool at a time — never ask the owner to configure two simultaneously. See reference/gotchas.md ↗ for the failure pattern this replaces.
3. Run one recipe to prove value. Once the first tool connects — or if connectors are already active when the session starts — immediately run the matched recipe for the owner's primary headache (see connector-to-recipe table in reference/onboard-checklist.md ↗). Narrate what Claude is doing and why — this is the "aha" moment. Do not skip it to get to the interview faster. For a worked example of the full arc, see reference/examples/happy-path.md ↗.
4. Interview the owner. Ask the five questions from reference/onboard-checklist.md ↗, one at a time, conversationally. Wait for the full answer before moving to the next. If the owner seems pressed for time, compress to three: industry, headaches, tools — but never fewer.
5. Store context. Show the owner the full profile before writing. Wait for explicit approval. Write the block to the Cowork session memory directory under the heading ## Business context using the exact format in reference/onboard-checklist.md ↗. If a memory file already exists, update only the ## Business context section — do not touch other content. Confirm: *"Saved. Every skill from here will know your business."*
6. Set the weekly cadence. Propose: *"Each Monday, just say 'weekly check-in' and I'll pull a snapshot of your numbers, flag anything urgent, and remind you what's due."* If they prefer a different phrase or day, store it in the profile. If tools are connected, name one skill the owner can try right now. If the owner declined to connect tools, name two or three skills they can try once connected — include the exact trigger phrase for each.
Approval gates
- Show context before writing. Display the full owner profile draft before storing it. Wait for explicit approval.
- Never overwrite existing context silently. If a
## Business contextblock already exists, show current vs. proposed before writing any changes. - Never connect a tool on the owner's behalf. Guide; do not act. Connector auth is always owner-initiated.
Reference
- reference/onboard-checklist.md ↗ — interview questions, connector priority matrix, recipe selection, context storage format
- reference/gotchas.md ↗ — Good / Bad patterns for pacing, tool selection, and context storage
- reference/examples/happy-path.md ↗ — worked example: retail shop owner, first session end-to-end
Frequently asked questions
What does the smb-onboard skill do?
Helps SMB owners get started by connecting tools, demonstrating value, and gathering business context for personalized assistance.
How do I install smb-onboard?
Open the “Installation” block on this page and pick your agent — Claude Code, Codex, Cursor or Gemini CLI — for ready-made commands that copy the skill folder to the right place. Or copy the prompt from the “Via assistant” tab and let the assistant review and install the skill.
Which AI models work with smb-onboard?
Skills are plain Markdown instructions (the Agent Skills format), so smb-onboard works with any agent that reads SKILL.md: Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, OpenCode and more. The model loads the instructions only when a task matches the skill's description.
Is smb-onboard free?
The skill is published openly by anthropics under the Apache-2.0 license. The skill itself costs nothing — you only pay your AI provider for the tokens you use.
Is it safe to install skills?
A skill is a set of instructions your agent will follow, so read SKILL.md and its companion files before installing. Scripts in the skill folder only run if your agent runs them — keep an eye on what it does and do not grant unnecessary permissions.
How is a skill different from an MCP server or a plugin?
An MCP server or plugin gives the model new tools — code that runs and returns a result. A skill gives it knowledge and process: how to approach a task, which steps to follow and what good output looks like. It needs no server or API key.