how to use one skill in every ai tool

Most people who use AI seriously now use more than one tool. A chat assistant for thinking and writing, a coding agent in the terminal, another inside the editor, maybe a second assistant because a colleague prefers it.

Skills were designed as files in a folder, and each tool reads its own folder. So the skill you rely on exists in one place, a copy of it in a second, and nothing at all in the third. The copies drift. You fix a step in one and forget the others. After a few months you have three slightly different versions of how you work.

the three ways people handle it

Copy it into every tool. Works on day one. Every improvement then has to be made several times, and it won't be. This is how most people's skills quietly go stale.

Symlink one folder into several. Neat, if all your tools run on one machine and all read skill folders. It doesn't reach a browser assistant, a second computer or a teammate, and it breaks quietly when a tool changes where it looks.

Keep the skill in one place every tool reads from. One copy, one address, and each tool fetches it when needed. This is the only version that stays true as the number of tools grows.

what "one place" has to do

For a single source to work across tools, it needs to:

• serve the skill to any tool you connect, not only the ones that read a particular folder;
• always serve the current version, so an edit reaches every tool at once;
• work across machines and accounts, because your tools don't all live on one laptop;
• keep the skill's author and history, so you can tell which state you're following.

A plain link can do most of this on its own. A skill published at a stable address, with the raw markdown available at the same address, can be handed to any AI that can read a URL: paste the link and it has the skill.

how innernet does it

innernet is memory that every AI tool you connect reads over MCP, the open protocol most AI tools now support. Connect it once in each tool, and they all read and write the same memory: your projects, your decisions, your preferences, and your skills.

Your own skills are pages in that memory. Edit one anywhere, from any connected tool or the dashboard, and every tool reads the new version next time.

Skills other people keep are links on the community page. Paste one into any connected AI and it loads there, at whatever version its maker keeps now. Nothing is copied, so nothing drifts.

Any AI that can read a URL can use a published skill even without the connection, by fetching the link.

when copying is fine

If you use one tool, copying a skill into its folder is the simplest thing and you should do it. The case for one source starts at the second tool, and becomes hard to argue with at the third, or the first time you share a skill with someone else.

browse skills you can use anywhere →

questions

can i use claude skills in chatgpt?

A skill is a plain markdown document, so its instructions work in any capable assistant. ChatGPT does not read Claude's skill folders, but it can use a skill handed to it as a link or through a connected memory such as innernet, which serves the same skill to both.

how do i keep skills in sync across ai tools?

Keep one copy in a place every tool reads, instead of a copy per tool. Publishing a skill at one stable link, or keeping it in a shared memory that each tool connects to over MCP, means an edit reaches every tool at once.

do cursor and codex support skills?

Support for the skill format has spread to several coding agents and editors, and details change quickly. Check your tool's documentation for where it reads skills from. A skill served over MCP or by link works regardless of folder conventions.

what is the best way to use the same skill on two computers?

Don't store it only on either computer. Keep it in a synced or hosted location both machines' tools can reach, such as a published link or a memory service your AI tools connect to.

innernet is user-owned memory across AI tools, reachable over MCP. Connect it once and every tool reads the same projects, decisions and skills. innernet.live

Checked 17 september 2026.