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

# Skills

> Reusable instruction packages for Cerebrum. They help the agent recognize a task, load the right guidance, and follow a repeatable workflow while working with you.

## Skill Types

<CardGroup cols="3">
  <Card title="Personal" icon="user">
    Skills attached to your user account. Use them for private preferences, personal workflows, or individual coding standards.
  </Card>

  <Card title="Workspace" icon="users">
    Skills shared with everyone in the workspace. Use them for team conventions, project workflows, and shared operational playbooks.
  </Card>

  <Card title="System" icon="cloud">
    Skills provided by Ardor. They teach Cerebrum how to work with built-in Ardor capabilities and supported platform workflows.
  </Card>
</CardGroup>

## How Skills Work

Cerebrum discovers skills by reading their metadata and descriptions. When a you ask for something, the agent compares the request with available skill descriptions and activates the most relevant skill automatically.

<Steps>
  <Step title="You ask for a task">
    For example, "prepare a deployment checklist" or "use our API review process."
  </Step>

  <Step title="Cerebrum matches a skill">
    The agent checks skill metadata, especially the description trigger in the YAML frontmatter.
  </Step>

  <Step title="The skill is applied">
    Cerebrum follows the skill's instructions and uses any included references, scripts, or assets when they are useful.
  </Step>
</Steps>

## Package Structure

Every skill is stored as a structured package:

```text theme={null}
my-skill/
|-- SKILL.md
|-- references/
|-- scripts/
`-- assets/
```

| Path          | Required | Purpose                                                                                                                    |
| ------------- | -------- | -------------------------------------------------------------------------------------------------------------------------- |
| `SKILL.md`    | Yes      | Contains YAML frontmatter with the skill name and trigger description, followed by step-by-step instructions for Cerebrum. |
| `references/` | No       | Stores supporting documentation, policies, examples, and reference material.                                               |
| `scripts/`    | No       | Stores executable scripts that automate repeatable work.                                                                   |
| `assets/`     | No       | Stores templates, code examples, snippets, and reusable files.                                                             |

### SKILL.md

`SKILL.md` is the entry point for every skill. It contains metadata and the instructions Cerebrum should follow.

```md theme={null}
---
name: api-review
description: Use when reviewing API design, endpoint naming, request validation, or response contracts.
---

# API Review

1. Check that endpoint names follow the workspace naming conventions.
2. Verify request and response schemas.
3. Look for missing validation and error states.
4. Summarize risks before suggesting implementation changes.
```

<Note>
  The description should be specific enough for Cerebrum to know when the skill applies.
</Note>

## Ways to Use Skills

<AccordionGroup>
  <Accordion title="Automatic activation">
    Cerebrum activates a skill automatically when your request matches the skill's YAML frontmatter description.

    For example, if a skill description says it should be used for API reviews, asking "review this API contract" can activate that skill without any extra command.
  </Accordion>

  <Accordion title="Manual activation">
    You can explicitly ask Cerebrum to use a skill by name, either through a slash command or directly in text.

    Examples:

    * `/api-review check this endpoint design`
    * `Use the api-review skill for this service`
  </Accordion>
</AccordionGroup>

## When to Create a Skill

Create a skill when you want Cerebrum to follow the same guidance repeatedly:

* Team engineering standards
* Code review checklists
* Deployment or incident workflows
* Documentation templates
* Project-specific context that should be reused across sessions

<Note>
  The best way to do this is asking Cerebrum to create skill and describe the workflow you want to bake in.
</Note>