What a Skill is, in practice
It's a documented procedure that an AI can follow consistently — how to draft a reply to an angry customer, how to structure a weekly report, how to classify a support ticket. The difference from "writing a good prompt" is that a Skill is documented and reused; a one-off prompt gets lost in a conversation's history.
Why it matters that it belongs to the team, not to one person
If only one person knows how to "prompt well", the company depends on them just as much as it used to depend on that same person doing the task by hand. Documenting the Skill moves that knowledge out of someone's private chat and into a resource anyone on the team can use the same way.
How to start documenting one
Pick a task that's already repeated and currently varies depending on who does it. Write down the steps, the quality criteria and a couple of examples of a good result and a bad one. It doesn't need to be perfect from day one — it needs to exist and be improvable.
Who maintains it and when it gets updated
Like any internal process, a Skill goes stale if no one reviews it. Assign an owner, even an informal one, and update it when the real process it describes changes — not whenever someone happens to remember.
The end goal
It's not that your team "knows AI" — it's that the AI supports them with the same judgment, no matter who wrote the Skill. When that works, bringing someone new onto the team stops meaning rebuilding the knowledge from scratch.