---
title: 'Skills'
description: 'Create reusable AI instructions and control their visibility.'
---

# Skills

A skill supplies instructions; it does not execute server code. Open **AI → Skills** to find available skills and inspect their source and visibility.



## Create a Studio skill

1. Choose **New Skill**.
2. Enter a descriptive Name and Description so the agent can identify its purpose.
3. Select personal, project, or space visibility as appropriate.
4. Write the Markdown Content and inspect **Preview**.
5. Save and attach it to the intended automation when it must always be loaded.

For a catalog review skill, specify which fields to inspect, how to report missing values, and that the agent should cite record IDs. Keep instructions independent of credentials and private infrastructure details.

## Attachment and discovery

Attached skills are loaded for the automation. Other visible skills can be discovered as needed. Skill visibility does not broaden data permissions. A skill telling the agent to call a tool does not attach or authorize that tool.

## Source ownership

Studio-created and code-defined skills have different maintenance paths. Edit version-controlled skills in their source and sync the deployed handler; do not rely on a UI edit to override code ownership. See [define skills in code](/docs/ai/skills/define-in-code).

The content limit is 50,000 characters. Prefer focused instructions with explicit completion criteria over an oversized collection of unrelated procedures.


---

[View full sitemap](/docs/sitemap.md)