Shared AI skills sessions: learn, test and build a reusable skill together

3D Open Website is opening conversations for shared skills sessions across our growing European collaboration network. The idea is simple: experts and beginners bring practical questions, explain useful methods and improve small, reusable skills together. You do not need to be an expert to contribute a valuable example or reveal an assumption that a tool missed.

What do we mean by a skill?

A human skill is an ability built through practice. An AI agent skill is a reusable package of task instructions and supporting resources for a compatible tool. These are related, but not identical. A good session can develop both: people understand the task better, while a written method becomes easier to repeat and review.

Vendor formats differ. For example, Anthropic describes agent skills as modular instructions, metadata and optional resources. That does not mean any downloaded skill is safe or that every tool will run the same package. We check the intended environment and permissions before trying a contribution.

What can beginners and experts each bring?

A beginner can bring a confusing real task, an attempted solution and the point where it stopped working. An expert can explain decisions, show a failure case and help define a reliable check. A business participant can describe the actual constraints: who uses the output, which data is private and what a useful result looks like.

The session should not become a competition in prompt length or technical vocabulary. A plain-language explanation that helps another person complete a task is more useful than a polished demo nobody can reproduce.

A proposed session structure

1. Choose one task. For example, turn an approved brief into a website content checklist.
2. Show one starting example. Use invented or explicitly permitted data.
3. Explain the method. Show the inputs, instructions and expected output.
4. Try a different example. Ask another participant to use the method.
5. Inspect the result. Record missing facts, errors and points requiring human judgement.
6. Improve and document. Save the useful instructions, checks and limitations.

This is a proposed working structure, not a published timetable or a promise that every session has the same length. Dates, format, participation arrangements and any terms will be confirmed separately with participants.

A small skill brief you can prepare

Before the conversation, write six lines: task, intended user, required inputs, expected output, checks and stop conditions. For a multilingual content checklist, the inputs might be an approved business brief and the required languages. The output might be a list of missing content and translation questions. A stop condition might be “the source does not state the price, so do not invent one”.

Add one normal example, one incomplete example and one case where the method should refuse or ask for clarification. A reusable skill is not ready merely because it produced one impressive answer.

How do we review the skill?

• Can another person follow it without a private explanation?
• Does it distinguish observed facts from assumptions?
• Does it say when to stop and ask a human?
• Are its sources, licenses and author contributions clear?
• Does it require more access than the task needs?
• Can the output be checked in a different example?

A code-related contribution needs a review appropriate to the change. A design contribution needs readable text and usable controls, not just an attractive screenshot. MDN’s accessibility learning material is one starting resource; it is not replaced by a single AI check.

Share methods, not secrets

Bring a synthetic example instead of a customer database. Do not share API keys, passwords, personal records or work covered by client confidentiality. Remove identifying details before screen sharing. If a workflow sends mail, publishes content or changes access, replace that step with a demonstration or an explicitly reviewed action.

Before reusing or distributing a contribution, agree attribution and the relevant license or permission. Joining a conversation does not transfer ownership of your work or promise that a skill will be published.

How to express interest

Use the partner introduction form and mention shared skills sessions in the message. Tell us whether you want to learn, demonstrate a method, help review skills or bring a business task. Describe your current level honestly and suggest one topic. A website or portfolio is optional.

Read the open network invitation and the guide to agentic workflow responsibilities. We are collecting introductions and shaping the sessions; no session booking or certification is issued by the form.