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

# Skills, Plugins & Connectors

> Extend your agent with marketplace packages. Skills teach, plugins add runtime capabilities, connectors wire to external systems. Install with one command.

## Three Ways to Extend Your Agent

ArgentOS has three extension types, each serving a different purpose:

<CardGroup cols={3}>
  <Card title="Skills" icon="graduation-cap">
    **Teaches** the agent how to do something. Injected as prompt instructions. No code runs.
  </Card>

  <Card title="Plugins" icon="plug">
    **Adds new tools** to the gateway runtime. Runs inside the gateway process.
  </Card>

  <Card title="Connectors" icon="link">
    **Wires to external APIs** with auth and health checks. Runs as an isolated subprocess.
  </Card>
</CardGroup>

| Type           | What It Does                                       | Runs Where                      | Example                           |
| -------------- | -------------------------------------------------- | ------------------------------- | --------------------------------- |
| **Skills**     | Teaches the agent how to do something              | Injected as prompt instructions | Deep Research Pro, SEO Audit      |
| **Plugins**    | Adds new tools to the gateway runtime              | Inside the gateway process      | Think Tank, VIP Email Alert       |
| **Connectors** | Wires to external APIs with auth and health checks | Isolated subprocess             | HubSpot, Google Workspace, Linear |

## Installing Packages

```bash theme={null}
# Browse the marketplace
argent marketplace search hubspot

# Install a skill
argent marketplace install deep-research-pro

# Install a connector
argent connectors install aos-hubspot
```

Or browse visually at [marketplace.argentos.ai](https://marketplace.argentos.ai).

## Skills

A skill is a `SKILL.md` file — markdown instructions that tell your agent how to use an API, follow a workflow, or handle a specific task. No code runs. The agent reads the instructions and uses its existing tools to accomplish the task.

### SKILL.md Format

```yaml theme={null}
---
name: my-skill
description: What this skill does and when to use it.
---

# My Skill

Instructions for the agent go here.

## When to Use
- User asks about X
- User wants to do Y

## How It Works
Step-by-step instructions, API endpoints, examples.
```

<Info>
  **Required fields:** `name` and `description` in the YAML frontmatter. The description is what triggers the skill — make it specific.
</Info>

### Creating a Skill

<Steps>
  <Step title="Create the skill directory">
    ```bash theme={null}
    mkdir my-skill && cd my-skill
    ```
  </Step>

  <Step title="Write your SKILL.md">
    Add the frontmatter and instructions as shown above.
  </Step>

  <Step title="Package and upload">
    ```bash theme={null}
    tar czf my-skill-1.0.0.tar.gz my-skill/
    # Upload at marketplace.argentos.ai/submit
    ```
  </Step>
</Steps>

## Plugins

A plugin is TypeScript code that runs inside the gateway. It registers new tools via `api.registerTool()`, hooks into agent events, and has full access to the runtime API.

Plugins live at `~/.argentos/extensions/` and use an `argent.plugin.json` manifest:

```json theme={null}
{
  "id": "my-plugin",
  "name": "My Plugin",
  "description": "What it does",
  "version": "1.0.0"
}
```

<Tip>
  Plugins are more powerful but carry more risk — they execute code inside your gateway process. See [Plugin Development](/tools/plugins) for the full guide.
</Tip>

## Connectors (AOS)

Connectors are production-grade integrations with external APIs. They run as isolated subprocesses with their own:

* **Authentication** — service keys resolved through the secret store
* **Health checks** — `doctor` command validates connectivity
* **Risk tiers** — `readonly-external`, `readwrite-external`
* **Capability discovery** — commands and resources declared in `connector.json`

Once installed, a connector's actions become agent tools automatically.

### Available Connectors

<CardGroup cols={2}>
  <Card title="Google Workspace" icon="google">
    Gmail, Drive, Calendar — 15 commands
  </Card>

  <Card title="HubSpot CRM" icon="hubspot">
    Contacts, deals, companies, tickets
  </Card>

  <Card title="QuickBooks" icon="calculator">
    Invoices, payments, financial reports
  </Card>

  <Card title="Linear" icon="diagram-project">
    Issues, projects, workflow management
  </Card>

  <Card title="Firecrawl" icon="fire">
    Web scraping and content extraction
  </Card>

  <Card title="Mailchimp" icon="envelope">
    Email campaigns and audience management
  </Card>

  <Card title="WordPress" icon="wordpress">
    Content publishing and page management
  </Card>

  <Card title="Google Places" icon="location-dot">
    Local business search and discovery
  </Card>
</CardGroup>

## Marketplace Security

Every package in the marketplace goes through three-layer security:

<Steps>
  <Step title="VirusTotal scan">
    Scanned by 70+ antivirus engines.
  </Step>

  <Step title="ArgentOS AI Safety">
    25 prompt injection patterns, secret detection, dangerous code checks.
  </Step>

  <Step title="Manual Review">
    The ArgentOS team reviews before publishing.
  </Step>
</Steps>

<Tip>
  Packages with a "VT Scanned" badge link directly to the VirusTotal scan report.
</Tip>

## Submitting to the Marketplace

<Steps>
  <Step title="Sign in">
    Sign in with GitHub at [marketplace.argentos.ai/submit](https://marketplace.argentos.ai/submit).
  </Step>

  <Step title="Upload package">
    Upload your `.tar.gz` package.
  </Step>

  <Step title="Fill details">
    Fill in display name, description, category, version, tags.
  </Step>

  <Step title="Submit for review">
    Submit — security scanning runs automatically. If scans pass, the team reviews and approves.
  </Step>
</Steps>

See the [Marketplace](https://marketplace.argentos.ai) for the full catalog.