> ## Documentation Index
> Fetch the complete documentation index at: https://langchain-5e9cc07a-preview-ramona-1774553401-08f2388.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Skills

> Use skills to give your agents access to specific capabilities.

Skills are reusable capabilities that provide specialized workflows and domain knowledge to your agent. Each skill is stored in the agent's long-term memory at `memories/skills/<skill-name>`. The skill's name and description is loaded when the agent starts. Based on this info the agent can decide to use the skill. The full skill file is only loaded when the agent determines it is relevant to the current task. Any referenced additional resources may be loaded by the agent if they become relevant.

Using skills can help:

* Save on token usage by only providing context relevant to the current task.
* Prevent the agent from having too much context in the system prompt, which can lead to hallucinations and incorrect responses.

<Info>
  Fleet skills are built on [Deep Agents](/oss/python/deepagents/skills) and follow the [Agent Skills specification](https://agentskills.io/specification). For details on skill structure, the `SKILL.md` format, and authoring best practices, see the [Deep Agents skills documentation](/oss/python/deepagents/skills).
</Info>

## Create a skill

You can create a skill two ways:

* **With AI**: Use natural language to describe the skill and the agent will create it for you. You can also add additional resources. Any additional files must be referenced in `SKILL.md` for the agent to be aware of them.
* **Manually**: Create a skill with a `SKILL.md` file.

<Note>
  Skills are **private** to the agent they belong to. Each agent has its own set of skills stored in its long-term memory, and no other agent can access them.
</Note>

<Tabs>
  <Tab title="With AI">
    In [Fleet](https://smith.langchain.com/agents), select the agent to which you want to add the skill. In the chat, prompt the agent to create a skill. For example:

    <Prompt description="Create a skill that helps the agent use the web to research a topic.">
      Create a skill that helps the agent use the web to research a topic. Use when asked to research a topic, person, company, technology, event, or any question that requires gathering and synthesizing information from the web. Covers news lookups, competitive analysis, background research, and fact-finding tasks. Prefer `tavily_web_search` for most queries.
    </Prompt>
  </Tab>

  <Tab title="Manually">
    <Steps>
      <Step title="Select the agent">
        In [Fleet](https://smith.langchain.com/agents), select the agent to which you want to add the skill.
      </Step>

      <Step title="Create the skill">
        In the graph view, navigate to the **Skills** section and click <Icon icon="pencil" /> **Create**.
      </Step>

      <Step title="Enter the skill details">
        1. Enter the skill name.
        2. Add a short description of the skill.
        3. Write the instructions for the skill.
        4. Click **Save Changes**.
      </Step>
    </Steps>
  </Tab>
</Tabs>

## Edit a skill

1. In [Fleet](https://smith.langchain.com/agents), select the agent to edit.
2. Click the <Icon icon="pencil" /> **Edit Agent** icon.
3. In the graph view, navigate to the **Skills** section and select the skill you want to edit.
4. Update the skill name, description, or instructions.
5. Click **Save Changes**.

## Delete a skill

1. In [Fleet](https://smith.langchain.com/agents), select the agent to which the skill belongs.
2. Click the <Icon icon="pencil" /> **Edit Agent** icon.
3. In the graph view, navigate to the **Skills** section.
4. For the skill you want to delete, click the <Icon icon="trash" /> **Delete skill** icon.

***

<div className="source-links">
  <Callout icon="edit">
    [Edit this page on GitHub](https://github.com/langchain-ai/docs/edit/main/src/langsmith/fleet/skills.mdx) or [file an issue](https://github.com/langchain-ai/docs/issues/new/choose).
  </Callout>

  <Callout icon="terminal-2">
    [Connect these docs](/use-these-docs) to Claude, VSCode, and more via MCP for real-time answers.
  </Callout>
</div>