Skills

The team's know-how, written once

A skill is a way of doing things (a procedure, a tone, a template, a script) that your agents and the chat apply. One person writes it, the whole organization benefits.

The format

An open format: Agent Skills

A skill is a folder with a SKILL.md file: a name, a description that tells the model when to use it, then the instructions. Files placed next to it travel with it.

  • Scripts, references and templates are copied into the agent's workspace at launch: it reads and runs them as needed.
  • A skill written for Claude Code works as is in Appstrate, and the other way round.
  • Import a ZIP, a GitHub repository or an .afps package. Appstrate adds versions and sharing.

One skill, three places where it serves

Written once

  • In its home spacethe only one that can edit it
  • Published as a versionfrozen, comparable, restorable
Skillsupport-tone 1.2.0

Used by

  • Agentseach at a pinned version
  • The chatautomatic, manual or strict
  • Other spacesshared, then enabled by each

Catalog and sharing

Shared with those who need it, enabled by those who decide

Every skill has a home space, the only one that can edit it. It is shared with a space or a person, then each space chooses to enable it.

  • Sharing needs a published version. Outside its home space, the latest published version runs.
  • Disabling a skill deletes nothing: its settings come back when you enable it again.
  • The Catalog shows everything available, "By space" or in "Discover" mode.

In the chat

The chat uses them too

Each conversation has its skills mode, and a space administrator can enforce some for everyone.

  • Automatic: the assistant sees the space's skills and loads the ones that fit.
  • Manual or strict: you pick up to 5 skills; in strict mode it cannot load any other.
  • An administrator enforces up to 3 published skills in the space's chat.

Versions

Each agent picks its version

The draft can be edited freely. A published version never changes: you compare it, restore it, and each agent pins the one it uses.

  • "Create a version": patch, minor or major.
  • "Compare" and "Restore" from the version history.
  • Through the API, a run can test a skill's draft without republishing the agent.

What it changes

For your teams

  • Write your method once, by chatting with the chat or in a short form.
  • Bring a skill you already use in Claude Code.
  • A colleague shares a skill with you: it shows up on your side, ready to use.

For the organization

  • A consistent tone and procedures from one agent to the next, and in the chat.
  • Frozen versions: you know what ran, you compare, you restore.
  • A home space that controls changes, spaces that choose what they enable.

Frequently asked questions

What is the difference between a skill and an agent?
An agent does a task: it has its instructions, tools and model. A skill is reusable know-how that several agents and the chat apply, such as a procedure or a writing tone.
Do a skill's scripts run?
Yes, in the run's sandbox: the skill's folder is copied there, and the agent reads or runs what it needs with its file and shell tools.
Can I bring a skill from Claude Code?
Yes: a SKILL.md folder in the Agent Skills format works as is. Import it as a ZIP or from a GitHub repository.
Who can see and edit skills?
Builder and Administrator roles edit them; Operator and Viewer read them; Runner does not see them. Only the home space edits a skill.

Let's see what it would look like on your side.

Try it on your own in a few minutes, or take thirty minutes with us to scope a first use case.

To try it

Try it for free

Create a cloud account and describe your first agent.

For a guided start

Let's talk about your case

Thirty minutes with us to scope a first use case and its gain.

To audit

Read the code

Apache 2.0. Your technical team can read it all, install it all, keep it all.