> ## Documentation Index
> Fetch the complete documentation index at: https://docs.usecadence.gg/llms.txt
> Use this file to discover all available pages before exploring further.

# Skills

> Reusable recipes your bots follow.

A **skill** is a written recipe for a task: a name, a short description of when to use it, and the steps. Bots read a skill when the task in front of them matches its description.

Skills live in a library every bot in your space shares. Write a skill once, and every bot can use it.

## Write a skill

1. Open any bot's settings, then **Advanced**, then **Skills**.
2. Click **New skill**. The editor opens with a template.
3. Fill in the name, the description and the steps, then save.

A skill uses the `SKILL.md` format: a short header with `name` and `description`, then the steps in plain Markdown.

```markdown theme={null}
---
name: weekly-sales-summary
description: Use when asked for the weekly sales summary.
---

1. Open the sales sheet and read last week's rows.
2. Total sales by product.
3. Reply with the top five products and the week's total.
```

Names can be up to 80 characters, and a skill up to 100,000 characters.

## Built-in skills

Cadence includes **Interrogate**, which reviews a change, pull request or plan adversarially and reports problems without fixing them. Built-in skills are read-only and show a **Built-in** badge.

## Bots can write skills too

When a bot works out a good way to do something, it can save it as a skill with its `skill_create` tool, and improve it later with `skill_update`. You see and edit those skills in the same list.

## Bring skills from another assistant

The **Import an agent** flow can bring skills from another assistant, such as OpenClaw or Hermes. Imported skills join the shared library.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.