Skip to main content
Skills are reusable procedures for the Slate agent. Each skill is a package built around a SKILL.md file — instructions Cowork loads when a request matches, so the agent follows your procedure instead of improvising one.

Where to Find It

Go to Assets → Skills in the sidebar. The page has two tabs: Built-in skills and Custom skills.
Skills page showing the built-in skills gallery with category filters and enable switches

Skills — built-in gallery and custom skills

Built-in Skills

The built-in tab is a gallery of skills managed by Slate, filterable by category. Each card shows the skill’s name, category, and description, with a switch to enable or disable it for your workspace. Click a card to preview the full skill. From the preview you can:
  • Enable or disable the skill.
  • Customize as new skill — creates an editable, workspace-owned copy from the latest managed version (the managed original is disabled so they don’t overlap).

Custom Skills

The custom tab holds skills your workspace owns. Three ways to add one:
  1. Ask Cowork to create a skill for you.
  2. New skill — author one in the editor: give it a Name and Description, then write the SKILL.md from the provided template.
  3. Import — upload a .zip containing <skill-id>/SKILL.md (plus optional contract and scripts). Imported scripts always run sandboxed.
Each custom skill row has an enable/disable switch and a menu with Download (.zip), Re-import (.zip), and Delete (removes the skill and its stored archive).
A skill that fails validation shows Broken — re-import needed and drops out of the agent’s skill index until you re-import a fixed package. When any skills are broken, a Repair button appears in the header.

Editing a Skill

Click a custom skill to open its files. Markdown and scripts in imported skills are editable in place — pick a file, edit, and Save. Built-in skills are read-only.

Using Skills in Cowork

Enabled skills join the agent’s skill index — Cowork applies them automatically when a request matches. To force one, type /skill in the Cowork input box and pick it; the agent starts with that procedure and follows it.

What’s Next

  • Cowork — where skills get used
  • Memory — what the agent learns across runs
  • Recipes — full multi-step procedures with inputs and review gates