Learn
Share skills with your whole team through a shared drive

To share skills with your whole team, put them in one folder on a shared drive that every person's agents can read. When someone improves a skill, everyone's agents use the new version the next time they load it, with no copies to pass around.
In Fabric, SKILL.md files are also protected from changes by agents, so the skills stay as the team wrote them even when agents have write access to the folder.
What a skill is
A skill is a packaged set of instructions for one kind of task. Under the Agent Skills format, which Anthropic developed and released as an open standard, a skill is a folder containing a SKILL.md file. That file starts with a name and a description, followed by instructions. The folder can also hold scripts, reference documents and templates.
Agents don't read every skill in full all the time. At startup they load each skill's name and description, and when a task matches a description they read the full instructions. That's why a team can keep dozens of skills on hand: the cost of having one available is a couple of lines.
Teams tend to write skills for the work they repeat. Typical examples are a brand voice guide for client emails, a checklist for preparing a monthly report, the steps for turning interview notes into a findings document, or a script that formats invoices the way the accounts team likes them.
Why copies of a skill drift
Skills usually start on one person's laptop. Someone writes a good one, it works, and a colleague asks for it. The file goes into a chat message or an email, and now there are two copies.
A month later there are five. The designer has the version from March. The account manager edited hers to mention a new client and didn't tell anyone. The founder's copy still refers to a pricing sheet that was renamed. When the client report comes out with the wrong tone, nobody can tell which copy the agent was following, because each person's agent was following a different one.
This is the same problem shared drives solve for documents, and it gets worse with agents because nobody reads a skill while it's being used. A person might notice an outdated document. An agent follows the outdated skill without comment. Your AI agents need a shared drive, not a group chat describes how work splits into copies as soon as a second person joins, and skills split the same way.
Set up one skills folder
Create a single skills folder on the shared drive, with one subfolder per skill. A small team's folder might look like this:
Keep each skill focused on one task, and write the description so an agent can tell when to use it. A vague description such as "helps with writing" will match too much or too little. Here's a short example of a SKILL.md:
markdown
The example refers to the team's other conventions, such as decisions.md and NEXT.md. How to hand off work between AI agents explains those files, and a skill is a good place to tell every agent to use them.
If your team is new to shared folders, how to set up a shared drive for your team covers the basics first.
Connect every agent to the folder
How an agent finds skills depends on the tool. Some look in a set location on your computer, and some let you point them at a folder. The approach that works with any agent is a line in the project's AGENTS.md or CLAUDE.md saying where the skills live, for example "Skills for this team are in the skills folder on the shared drive. Check it before starting a task."
In Fabric, there are two main ways in. The Fabric desktop app mounts Fabric as a drive on Mac, Windows and Linux, so agents on that computer read the skills folder like any other folder. Files take no disk space and stream when they're opened. Skills that include scripts suit this route, since the agent can run the script on the machine it's working on.
Agents without a filesystem, such as Claude or ChatGPT in the browser, connect through Fabric's MCP server, which can search the drive and read file contents. An agent connected this way can find the right skill and read its instructions. How AI agents connect to a shared drive compares these routes with the CLI and API.
Either way, there's one copy. When the account manager updates the brand voice skill on Thursday, the designer's agent uses the new version on Friday without anyone sending anything.
Keep agents from rewriting skills
A shared skills folder means many agents read the same instructions. That's useful, and it also means a bad edit reaches everyone. An agent that misreads a request, or one that has been misled by instructions hidden in a document it read, could rewrite a skill that every other agent relies on.
Fabric treats SKILL.md as a protected instruction file. Agents can read it, but they can't change it, even with write access to the folder. Edits to skills come from people. Your agent instruction files are a security risk explains why this matters and how one compromised agent could otherwise steer the rest.
Agents can still be useful when a skill needs work. If an agent notices that a skill refers to a renamed file, ask it to write a short note beside the skill, such as skills/monthly-report/suggestions.md, describing the change it would make. A person reads the note and edits SKILL.md.
The other files in a skill's folder, such as templates and reference documents, are ordinary files. If you want agents to use them without changing them, give agents reading access to the skills folder only. Agent Vault lets you set access per folder, space or project, with separate permissions for reading, creating, moving and deleting, and for how long the access lasts. Agent Vault is on the Pro and Team plans.
Track and undo changes
Skills change as the team learns what works. Fabric keeps earlier versions of files, so if a revised skill produces worse results, you can look back at what it said before and restore it. Deleted files go to a trash you can restore from.
It also helps to note why a skill changed. A one-line entry in the team's decisions.md ("Shortened monthly report sections to 150 words after client feedback") makes the history easier to follow than comparing versions.
Habits that keep a skills library useful
Give each skill an owner, the person who edits it and answers questions about it. Without one, small fixes pile up as suggestions and nobody applies them.
Review the library every month or so. Remove skills nobody uses and merge ones that overlap, since two skills with similar descriptions confuse agents about which to load.
Keep skills short. A skill that runs to several pages is harder to review and harder for an agent to follow. Put long reference material in a separate file in the skill's folder and point to it.
Build tools as skills plus files. A team CRM, for example, can be a folder of client files and a skill that explains how to update them. Every person and every agent then uses the same tool. Build a CRM your agents can use shows the pattern, and agent skills in Fabric covers storing skills in Fabric.
For the broader idea of a drive shared by people and agents, see what is a shared drive for AI agents?
Frequently asked questions
What is an agent skill?
A skill is a folder with a SKILL.md file that gives an AI agent instructions for one kind of task. It can also include scripts, templates and reference files.
How do I share Claude skills with my team?
Keep the skills in one folder on a shared drive and point each person's agents at it, either through the tool's settings or with a line in CLAUDE.md or AGENTS.md. Everyone then uses the same, current version.
Can different AI tools use the same skills?
Skills follow the open Agent Skills format, which many agent products support. A shared folder lets agents in different tools read the same skill files.
How do agents access skills stored in Fabric?
Agents on your computer read them from the mounted Fabric drive. Agents without a filesystem can search for and read them through Fabric's MCP server.
Can an agent change a shared SKILL.md file in Fabric?
No. Fabric protects SKILL.md and other instruction files such as CLAUDE.md and AGENTS.md from changes by agents, even ones with write access. People edit them.
How do I update a skill for everyone?
Edit the SKILL.md file in the shared folder. Every agent that reads the folder uses the new version the next time it loads the skill.
What if a skill change makes things worse?
Fabric keeps earlier versions of files, so you can look at the previous version and restore it.
How many skills should a team have?
As many as you use. Agents load only each skill's name and description until a task needs one, so a larger library costs little, but overlapping skills can confuse agents.
Related pages
You might be interested in:

How to hand off work between AI agents

Share skills with your whole team through a shared drive

How AI agents connect to a shared drive: MCP, CLI, API or a mounted drive

Let AI agents analyze large data files without downloading them

Build a CRM your agents can use (it's just a folder)

What Is a Shared Drive for AI Agents?

What Is a Virtual Data Room?

How to let clients proof and choose photos online