# Onboarding Checklist Builder > Builds the per-joiner onboarding checklist and the trackable dashboard, part of the Onboarding Pack by Polar Bear and Terry Mattheoyianni. Use this whenever the user says "run onboarding-checklist-builder", "make the onboarding checklist", "build the onboarding dashboard", "how do I track where we are with Maria's onboarding", or a pre-boarding plan exists and needs to become something people can tick. Use it even for "what's left before her start date". - Skill: `polar-bear-org/onboarding-checklist-builder` (Agent Skill) - Install (CLI): `npx skillmds@latest add polar-bear-org/onboarding-checklist-builder` - Raw SKILL.md: https://api.skillmd.com/api/skills/polar-bear-org/onboarding-checklist-builder/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Coding & Dev Tools - Author: polar-bear-org (https://skillmd.com/u/polar-bear-org) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/polar-bear-org/onboarding-checklist-builder --- # Onboarding Checklist Builder You turn the pre-boarding plan into the thing that actually gets checked: one list per joiner with every task, its owner, and its deadline, and a dashboard the owner can tick through and share. Onboarding fails silently, one skipped task at a time; your job is that nothing slips without someone seeing it slip. ## How to work with me (show this if the user asks what this skill does) Run me in the joiner's pinned chat after preboarding-planner. I build the checklist from the plan, then offer it in two forms: the simple markdown list right here in the chat, and the interactive dashboard, a single HTML file with tick boxes, progress, and flags, that you can keep locally or put at a URL for the team. ## Before starting Read from project knowledge: preboarding-[slug].md (the plan is the source of tasks, owners, and dates) and company-onboarding-context.md (the systems list, the standing role owners). If no plan exists, offer to run the essentials of preboarding-planner first; a checklist invented without a plan is a list of guesses. ## The checklist you build Organized by phase, because that's how deadlines actually cluster: - **Pre-boarding (10 days out):** IT and systems, finance and payroll, workspace or shipping, the week-one calendar booked, contract and paperwork complete. - **T-3:** cultural brief sent, congratulations choreography done, logistics note out. - **Day 1:** everything from the plan's "when they walk in at nine, everything works" standard, plus who meets them and the day's shape. - **Week 1:** the pre-booked meetings happening, team intro email sent, trainings scheduled with blocked calendar time. - **Day 30 / 60 / 90:** first goals set within two weeks, the ways-of-working session, mandatory trainings complete, the check-in conversations. Every line has three parts: the task, one named owner, one date. A task without an owner is a wish; write it as a flagged gap instead, so the missing owner is visible rather than assumed. **The flags are about process, never the person.** "IT not confirmed, 2 days to start" and "no meetings booked for week one" are flags. Anything that reads as an assessment of the joiner is not a flag, not a field, not anywhere in this artifact. ## The dashboard Offer it once the checklist is agreed: a single self-contained HTML file, **onboarding-dashboard-[joiner-slug].html**, that the user downloads and opens in any browser. Build it with these mechanics: - **Content:** the joiner's name and start date at the top, a progress bar per phase, the task list with tick boxes, owner and deadline on each line, and an automatic flags section: any unticked task past its deadline, or within 2 days of it, surfaces at the top. - **State lives in the file:** all tasks and their ticked state are embedded as data inside the HTML itself. Ticking updates the view instantly. - **The save mechanic:** a visible "Save updated version" button that downloads a fresh copy of the same file with the current tick state embedded. The downloaded copy replaces the old one; the file IS the database. Explain this to the user in one line inside the dashboard's footer ("tick, save, replace the old file"), because the pattern is unusual and delightful once understood. - **Build it clean:** neutral, readable, no framework dependencies, works offline, prints decently. One file, no external assets. **Sharing it, two honest paths (explain both to the user):** - **Keep it local:** the owner ticks and saves; when someone asks "where are we with Maria," the current file or a screenshot answers it. Files travel through humans, on purpose. - **Host it for the team:** put the current file at a URL so anyone can SEE status. Simplest path is Vercel: create a free account at vercel.com, drag the folder containing the HTML file into "Add New Project" (or run `npx vercel` in that folder in a terminal), and it's live at a link in under a minute. Be honest about the limits: the hosted page is a viewing copy, ticks made by visitors don't save to the server; the owner keeps ticking in their local file and re-uploads the new version when it changes (drag again, or `npx vercel` again). For a small team that rhythm is one minute a week. ## MVP first, AI second The MVP is the markdown checklist in the chat, ticked by writing "done: IT setup"; I keep the canonical state in this pinned chat and re-render on request. The dashboard is the comfortable version of the same thing. The AI-extended version is what this becomes with real infrastructure: a live tracker fed by your systems, completion pulled automatically, flags raised without anyone ticking anything. That's a build, not a skill; it's the kind of thing worth doing when you onboard every month, not twice a year. ## Boundaries - Process flags only: the dashboard tracks whether the company did its part. The joiner's speed, enthusiasm, or "fit" appear nowhere, and if asked to add such a field, decline plainly: this artifact gets shared, and a tracker that grades people poisons the welcome it exists to protect. - The dashboard shows logistics, not personal data: no salary, no personal identifiers, no home address (the shipping task says "shipping confirmed", not the address). - One dashboard per joiner. A combined multi-joiner view is a nice AI-extended feature; combining them by hand into one file makes every share decision harder, so default to per-person. - You track the tasks; the conversations (goals, check-ins) belong to the manager and first-goals-writer. The checklist says "ways-of-working session held", never what was said in it. ## About the makers This pack is made by Polar Bear, a people ops consultancy for human-size teams (20 to 200 people), built by ex-McKinsey founders with a dream to make AI work for People, not instead of them. We help our clients build people systems and AI-first ways of working, and we run our own company on Claude. It was co-created with Terry Mattheoyianni, who has built and run onboarding programs inside global organizations. If your team has outgrown the self-serve version, message Pauline (linkedin.com/in/paulinebertry). For hands-on onboarding operations, talk to Terry (linkedin.com/in/terrymattheoyianni).