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

# Skills

> Slash-command workflows that give agents specialized, repeatable procedures.

A skill is a packaged workflow an agent can run: a markdown instruction set with a name, a description, and a detailed procedure. Skills are what make a "Backend Developer" agent actually behave like one.

## How agents get skills

Skills come bundled with **specializations**. When you create an agent with a specialization (say, Backend Developer), it automatically carries that role's skill set: `/backend-design`, `/schema-check`, `/secure`, `/test-plan`, `/optimize-query`, `/error-handling`, and so on.

* **Built-in**: Syndicate ships 100 skills across 20+ specializations. Browse them in the [Marketplace](/guides/agent-marketplace) or the [skills catalog](/reference/skills-catalog).
* **User-made**: [Agent Creator](/guides/agent-marketplace/agent-maker) generates new skill packages from a description.
* **Community**: packages imported from GitHub repositories or local folders.

An agent's current skills are listed in its configuration panel under **Skills (N)**.

## Using skills in chat

Type **/** in the chat input to open the slash menu with the agent's skills. Pick one to insert the command, add any arguments or context, and send. The agent follows the skill's procedure instead of improvising.

You can also just ask in plain language ("run a security audit on this folder"); agents reach for their skills when the request matches. The slash command is the explicit, reliable way to trigger a specific workflow.

## Suggested workflows

In the Marketplace, each built-in specialization shows a **Suggested Workflow**: a numbered sequence of its skills that represents how the role typically operates. For example, Atlas's workflow runs `/think` (pressure-test the idea), `/delegate` (break down and assign), `/status` (check progress), `/pipeline` (run reviews), and `/reflect` (retrospective).

## Skills vs MCP servers

|                | Skills                                        | MCP servers                                            |
| -------------- | --------------------------------------------- | ------------------------------------------------------ |
| What they add  | Procedures and judgment (how to do something) | Tools and data access (the ability to reach something) |
| Where they run | Inside the agent's prompt                     | As separate processes or services                      |
| Examples       | `/secure`, `/test-plan`, `/launch`            | GitHub, database, web search                           |

They compose: a skill like `/deploy-review` works better when the agent also has an MCP server that can read your CI system.