> ## Documentation Index
> Fetch the complete documentation index at: https://docs.a2v2.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Skills

> Give your agent structured reasoning for complex, multi-step tasks.

**Skills** give your agent structured reasoning for complex tasks. Where
[instructions](/building/instructions) shape *how* the agent sounds and what it's
allowed to do, a skill equips it to work through a harder, multi-step request in a
consistent, repeatable way.

Open **Expertise → Skills** from the agent sidebar. A new agent starts with **No
skills yet** — you add the ones your agent needs.

## Prerequisites

* An agent (see [Create an agent](/building/create-agent)).

## Add a skill

<Steps>
  <Step title="Open Skills">
    In the agent sidebar, go to **Expertise → Skills**.
  </Step>

  <Step title="Create a new skill">
    Select **New Skill** to open the builder and define the skill.
  </Step>

  <Step title="Describe the skill">
    Fill in the skill's details (see the field reference below) — at minimum a
    **Title** and **When to Use** so the agent knows when to apply it.
  </Step>

  <Step title="Create and test">
    Select **Create New Skill**, then try a relevant, multi-step question in the
    [Sandbox](/testing/sandbox) to confirm the agent reasons through it the way
    you expect.
  </Step>
</Steps>

<Frame>
  <img src="https://mintcdn.com/a2v2ai/ItYQQYTHsiiRmr-t/images/building/skills-new.png?fit=max&auto=format&n=ItYQQYTHsiiRmr-t&q=85&s=47a98886e9c874913390cec86c771d16" alt="The Create New Skill builder with Basic Information, Knowledge & Logic, and Constraints & Output sections." width="1200" height="710" data-path="images/building/skills-new.png" />
</Frame>

## Skill builder reference

A skill is organized into three parts:

<AccordionGroup>
  <Accordion title="Basic Information">
    | Field           | What it does                                                                                                                                  |
    | --------------- | --------------------------------------------------------------------------------------------------------------------------------------------- |
    | **Title**       | A short name for the skill (e.g. "Refund Eligibility Check").                                                                                 |
    | **Description** | A short summary of what the skill does.                                                                                                       |
    | **When to Use** | Defines when the agent should activate this skill — e.g. "When the user asks whether an order qualifies for a refund." (up to 300 characters) |
  </Accordion>

  <Accordion title="Knowledge & Logic">
    | Field                | What it does                                                                                                                               |
    | -------------------- | ------------------------------------------------------------------------------------------------------------------------------------------ |
    | **Domain Knowledge** | Structured knowledge the agent needs for this skill. Supports Markdown and is delivered to the agent as one coherent block (not split up). |
    | **Reasoning Steps**  | An ordered list of steps the agent should follow. Use **Add step** to build the sequence.                                                  |
  </Accordion>

  <Accordion title="Constraints & Output">
    | Field             | What it does                                                                           |
    | ----------------- | -------------------------------------------------------------------------------------- |
    | **Guardrails**    | Rules the agent must respect while using the skill. Use **Add guardrail** to add them. |
    | **Output Format** | How the answer should be structured (e.g. **Free Form**).                              |
  </Accordion>
</AccordionGroup>

<Note>
  **Skills vs. Instructions vs. Knowledge.** [Instructions](/building/instructions)
  set the agent's overall persona and rules; your
  [knowledge base](/knowledge-base/overview) supplies facts to retrieve; a **skill**
  packages a specific, multi-step procedure (when to use it, the knowledge it needs,
  the steps, and guardrails) so the agent handles that task consistently.
</Note>

## Tips

<Tip>
  * Reach for a skill when a request needs several reasoning steps — not just a
    lookup in your [knowledge base](/knowledge-base/overview).
  * Keep each skill focused on one task so it stays predictable, and re-test in the
    [Sandbox](/testing/sandbox) after adding or changing it.
</Tip>

## Related

<CardGroup cols={2}>
  <Card title="Write instructions" icon="pen" href="/building/instructions" />

  <Card title="Add knowledge" icon="database" href="/knowledge-base/overview" />

  <Card title="Test your agent" icon="vial" href="/testing/sandbox" />
</CardGroup>