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

# Skills

> What a skill is, where skills come from, how the AI Companion decides to use one, and the limits that apply.

A skill is a saved set of instructions the AI Companion follows when a request matches. It is
plain Markdown: a name, a one-line description, and a body of steps. There is no separate model,
runtime, or tool behind a skill. The companion reads the body inside its own turn and does what it
says with the tools it already has.

Skills are how a team turns a workflow that worked once into one that works every time, for
everyone. "Monday pipeline review", "pre-call brief for enterprise accounts", "log a referral the
way we do it" are all skills.

## Where skills come from

| Source    | How it is created                                                                                       | Editable                            |
| --------- | ------------------------------------------------------------------------------------------------------- | ----------------------------------- |
| Authored  | Written in the Skills tab with the **New Skill** form.                                                  | Yes                                 |
| Generated | The companion drafts it from a conversation when you say "save this as a skill".                        | Yes                                 |
| Imported  | Pulled from a GitHub link or an uploaded file in the [Agent Skills format](/ai-companion/agent-skills). | Yes                                 |
| Built-in  | Seeded from Salesfinity's starter catalog when a team first activates.                                  | Yes                                 |
| Platform  | Shipped with the product for the companion's own use, such as the Skill Creator guide.                  | No, and not shown in the Skills tab |

Every skill except platform skills belongs to the team. Any member can see, edit, deactivate, or
delete it in the Skills tab.

## How the companion uses a skill

<Steps>
  <Step title="It sees the catalog">
    At the start of every conversation the companion is given the name and description of every
    active skill on the team. That is all it sees at this point, which is why the description
    matters so much.
  </Step>

  <Step title="It matches a request">
    When your request aligns with a skill's description, the companion loads the skill's full
    body. It prefers a matching skill over improvising a workflow from scratch.
  </Step>

  <Step title="It follows the body">
    The steps run inside the companion's normal loop: plan, call tools, ask for approval on
    writes, reply. If the body introduces steps the companion did not anticipate, it re-plans
    around them.
  </Step>

  <Step title="It tells you if the skill could not load">
    If a skill is missing or fails to load, the companion says so and handles the request from
    first principles rather than pretending.
  </Step>
</Steps>

You can also invoke a skill directly by naming it: "run the Monday pipeline review".

## The Skills tab

Open the companion and switch to **Skills**. From there you can:

* **Create** a skill with the form: name, description, and content.
* **Ask the companion** to create one with you. The tab offers "Help me create a skill together",
  which starts the guided flow described in [Creating skills](/ai-companion/creating-skills).
* **Import** a skill from GitHub or a file.
* **Edit**, **deactivate**, or **delete** any skill. A deactivated skill is kept but hidden from
  the companion; use it to pause a skill without losing it.

## Limits

| Field       | Limit             |
| ----------- | ----------------- |
| Name        | 200 characters    |
| Description | 500 characters    |
| Content     | 50,000 characters |

The content limit is generous on purpose, but a skill that approaches it is usually two skills.
Around 1,500 words is a good ceiling for a body the companion can follow reliably.

## Skills and routines

A skill is instructions; a routine is a trigger. When a skill describes something that should run
on a schedule, "every Monday at 9am", wire it into a [routine](/ai-companion/overview#routines)
whose instruction invokes the skill. The routine's write policy controls what the skill may
change while running unattended.

## Related

<CardGroup cols={2}>
  <Card title="Creating skills" icon="pen" href="/ai-companion/creating-skills">
    The structure of a good skill and the guided creation flow.
  </Card>

  <Card title="Importing agent skills" icon="file-import" href="/ai-companion/agent-skills">
    Bring in skills written in the open Agent Skills format.
  </Card>
</CardGroup>