> For the complete documentation index, see [llms.txt](https://notara-1.gitbook.io/notara-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://notara-1.gitbook.io/notara-docs/skills.md).

# Skills

A skill is a reusable AI prompt + tool configuration + context binding with a slug. Skills let you encode your team's best practices and workflows into the agent — so instead of re-explaining how to write an engineering digest every week, you define a skill once and invoke it by name.

## What Is a Skill?

Each skill has:

| Field             | Description                                                   |
| ----------------- | ------------------------------------------------------------- |
| **Name**          | Human-readable label                                          |
| **Slug**          | URL-safe identifier for invocation (e.g., `eng-error-digest`) |
| **Description**   | What the skill does (used by the agent to match intent)       |
| **Prompt**        | The instruction fragment injected when the skill is active    |
| **Bound Tools**   | Connectors this skill has access to                           |
| **Bound Context** | Libraries or specific documents this skill pulls from         |

When a skill is active, its prompt is appended to the agent's system context, its tools are added to the available set, and its context is injected alongside the channel's default context.

> **On Free:** a workspace can have up to 5 skills, written in the app or imported as `SKILL.md` or a zipped folder. Copies of built-in library skills count toward the 5, and only the workspace's own skills reach your AI over MCP. Bound tools need the paid platform. See [Free plan](/notara-docs/billing/free-plan.md).

## 48 Platform Skills

Notara ships 48 platform skills that cover common team workflows out of the box:

* **Engineering**: `eng-error-digest`, `eng-pr-review`, `eng-oncall-brief`, `eng-incident-report`
* **Growth**: `mkt-competitor-teardown`, `mkt-campaign-brief`, `growth-weekly-digest`
* **Operations**: `ops-meeting-notes`, `ops-weekly-report`, `ops-action-items`
* **Customer Success**: `cs-account-health`, `cs-onboarding-check`, `cs-churn-signal`
* **Research**: `research-company`, `research-market-size`, `research-trend-brief`
* ...and many more

Platform skills have a `null` tenant — they're available to every workspace. You can fork any platform skill to customize it for your team.

## The Skill Shelf

Navigate to **Skills** in the dashboard to see the skill shelf — a card grid showing all skills available to your workspace, including platform skills and any skills your team has created or forked.

Each card shows:

* **Name and description**
* **Slug** (for direct invocation)
* **Readiness state**: green (all required tools connected), amber dot (needs a tool), grey (platform key managed)
* **Schedule** (if the skill is used in an automation)
* **Connector marks**: icons showing which connectors the skill uses

## Skills and Permissions

Skills interact with the permission system in one direction: **they can only tighten permissions, never loosen them**.

If a workspace has `linear_create_issue` set to `requires_approval`, a skill cannot change it to `open`. A skill can raise `open` tools to `requires_approval` within its scope (e.g., a skill that's cautious about writes), but it cannot grant permissions the tenant hasn't authorized.

This design ensures that authoring a skill (which requires `editor` role) can never be used to grant elevated tool access. Granting elevated access always requires `admin`.

## Groups and Skill Scoping

Skills are group-scoped. A skill created in your engineering group is available to:

* The engineering group's Slack channels
* The engineering group's MCP endpoint

It's not automatically available to your marketing group's channels. To share a skill across groups, either create it at the workspace level (available to all groups) or fork it into each group separately.

## Invocation

The agent selects skills based on semantic matching — if your request sounds like what a skill is designed for, the agent activates it automatically. You can also invoke a skill explicitly:

```
@Notara use the eng-error-digest skill
```

For external agents using the MCP endpoint, skills are accessible as MCP prompts via `prompts/list` and `prompts/get`.

→ [Creating and forking skills](/notara-docs/skills/creating-skills.md)\
→ [Using skills](/notara-docs/skills/using-skills.md)