For the complete documentation index, see llms.txt. This page is also available as Markdown.

Build your own & share with your team

Turn the way you analyse your studio into a reusable skill, then publish it so your whole team's AI runs it the same way.

The best way you've found to look at your studio shouldn't live only in your head. Turn it into a skill — a saved playbook — and share it with your team, so everyone connected to your studio runs it the same way.

You don't need to write code. You describe what you want; the AI writes the skill; you review and publish it.

Build a skill in three steps

1. Get the analysis right once. Work through the questions with your AI assistant until the answer is exactly what you want — the right time window, your own definition of terms, your studio's language.

2. Ask the AI to save it as a skill. When you're happy, say:

"Turn this into a reusable skill called 'Monday Check-in' so I can run it every week."

The AI captures the steps and saves the skill to your studio. New skills land as a draft first, so nothing goes live to your team until you've looked it over.

3. Review and publish. Read the draft, run it once to confirm it does what you meant, then publish it. From that moment it's available to everyone connected to your studio.

Sharing with your team

A published skill is org-wide: every AI assistant connected to your studio — yours, your manager's, a coach's — sees the same skill and runs it the same way. You don't send anything around; publishing is the sharing.

This is how a studio builds its own playbook: your "Monday Check-in", your "End-of-month review", your "New member 30-day follow-up" — each one captured once and run by anyone.

You build and manage all of this from your Skills page in the operator app — the same marketplace where you add built-in and paid skills. Sharing a skill beyond your own studio — listing it for other operators on the marketplace — is rolling out over time.

Keeping skills good

  • Test before you trust. Run a new skill on a period you already understand, and check the answer matches what you know to be true.

  • Save a few examples. You can attach known questions-and-answers to a skill so the AI can check itself against them over time. This is the same quality-check the built-in Eval Runner uses.

  • Retire what you outgrow. If a skill stops being useful, mark it deprecated — it stays in the record but drops out of everyday use.

Want to go deeper?

Skills can do a lot more — load reference notes, follow a strict procedure, build on each other. Studio Ignition, our paid staged diagnostic, is a large example of a multi-step skill set. If you (or a technical teammate) want to author skills in depth, the developer tools reference documents the skill tools in full.

Last updated

Was this helpful?