Take on real work — overdue invoices, leaky pipelines, month-end close. Build reusable AI skills to clear it, and take them with you when you leave.
The first training simulation for AI workflow engineering.
No videos. No lectures. You do the work.
You don't watch lessons about automation — you're handed a messy business and asked to automate it. One real task, played out here: describe the skill, test it against real data, catch what it missed, ship it.
Built on Anthropic's AI Fluency framework — and the AI is real: every skill is built and run live by Anthropic's Claude models.
Every task ends with a reusable skill you engineered yourself — a real file that runs in your real AI tools:
Every task is a real business problem: you engineer the skill, the AI runs it live, and four review dimensions decide whether it ships.
Accounts receivable is a mess and the controller needs answers. Build a skill that surfaces every overdue, unpaid invoice from the ledger — your first reusable command.
Which accounts are about to walk? Design your own churn-scoring rubric across three data sources — and get graded on how well-reasoned and consistent your rubric is.
The company already has a standard lead-scoring formula. Build the skill that applies it faithfully from a bundled reference doc instead of inventing your own weights.
A signed offer needs to go out the day the candidate says yes. Build the workflow that generates it from prepped data — with a mandatory HR approval gate before anything sends.
A world is a complete simulated business — its own tasks, data, stakeholders, and a themed collection of creatures to earn. Finish a world's tree and you've genuinely learned its domain.
Step in as the newest automation engineer at a mid-market software company. Build the skills that run all eight departments — finance, hiring, sales, support, and legal.
Join a regional hospital network and automate the operations behind patient care — patient flow, staffing, claims, and compliance under real clinical constraints.
More worlds are on the roadmap — and your company could be one of them. We build private worlds for teams →
Every skill you master unlocks a collectible creature in that world's habitat — and a perfect run earns the rare shiny variant. Completionists, you've been warned.
Bring the training simulation to your team. We build a private world on your actual departments, workflows, and data — so your people learn AI where it counts: on the work they already do.
A safe, simulated copy of your operations — realistic data, your stakeholders, your edge cases. Nobody practices on production.
Every task ends in a working automation your team exports into the AI tools they already use. The training pays for itself in shipped workflows.
Four review dimensions score every deliverable, and completion is visible per person. You see fluency grow — not just seat time.
Company worlds are visible only to your team, with seat-based access. Your data and workflows never appear on the public platform.
Think of it as a flight simulator for your actual workflows — the hours count.
No subscription — a skill tree is something you finish, so we don't charge you rent on it. Pricing is per world.
Every world's opening tier, no card required.
That world's entire tree, yours for good.
A private training simulation built on your business.
Yes — if the simulation makes you do the work. In Skill Tree you write real prompts and skill specs, run them against real business data, and get scored on structure, automation quality, prompt quality, and efficiency. You draft, iterate, and improve exactly the way working engineers do — no videos, no multiple choice.
The craft of turning business problems into reliable, reusable AI workflows: describing a task precisely, giving a model the right data and tools, adding review gates where judgment matters, and composing skills into larger automations. It is the skill set behind titles like AI engineer and automation engineer — and it is what every task in Skill Tree teaches.
No. You build skills by describing what they should do — clearly, precisely, with edge cases handled. If you can write a good brief, you can play. Developers still get plenty of depth: later tiers cover multi-agent design, orchestration, and state management.
Real. Skills are built and executed by Anthropic’s Claude models against real data files, and grading is a mix of deterministic checks and AI review. When your skill works, it is because you engineered it well — not because the simulation let you win.
Every world’s first tier is free with an account. A one-time unlock opens that world’s full skill tree — no subscription, nothing recurring. Finish the tree and it stays yours.
Yes — that is the point. Every skill you build exports as a real, installable command or workflow you can run in actual AI tools at your actual job. Think of it as a flight simulator where the hours count toward your license.
Yes. Teams can run the public worlds as-is, or we can build your company its own private world — a training simulation of your actual departments, data, and workflows, visible only to your organization with seat-based access. Your people learn AI fluency on the work they already do, and every skill they build exports into the tools they use on the job. Reach us at [email protected] and we will scope it with you.
Free to start. Built by you. Yours to keep.
Play free