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

# Skills

> Reusable, named instructions you write once and attach to any agent.

A skill is a named instruction you write once and reuse. Instead of re-typing "keep the tone plain" or "never invent a source" into every agent, you save that behaviour as a skill and attach it. Every AI step in an agent then follows it.

Skills live on the **Skills** tab of the [Agents](/agents) area, at `/app/workflows`. Agents are also called workflows in a few places in the app; they're the same thing.

<Note>
  Agents are in **Beta** and review-gated. A skill only shapes how an agent writes. It never publishes, sends, or approves anything on its own. AI-generated work still waits for your review before you use it.
</Note>

## The four built-in skills

FixAEO ships four skills out of the box. They carry a **Built in** badge:

* **Humanizer** — makes generated writing feel natural and specific instead of robotic.
* **Improved Copywriting** — sharpens clarity and persuasion without adding claims the evidence doesn't support.
* **Match Brand Voice** — keeps wording aligned with your connected brand context.
* **Citation Guard** — keeps factual claims traceable to the sources the agent was given.

You can't edit, archive, or delete a built-in skill. To make one your own, **Duplicate** it. The copy is a normal custom skill you can change freely.

## Create your own

Click **New skill** and fill in four things:

<Steps>
  <Step title="Name">
    A short label, up to 80 characters. This is what you'll pick from the list later.
  </Step>

  <Step title="Category">
    One of **Writing**, **Brand**, **Evidence**, **Quality**, or **Custom**. Category only helps you filter the list.
  </Step>

  <Step title="Description">
    A "when to use it" line, up to 500 characters. It helps you and your teammates choose the right skill later.
  </Step>

  <Step title="Instructions">
    What the skill should actually do, up to 12,000 characters. Be concrete about tone, evidence, claims, and output. A focused instruction is easier to reuse than a sprawling one.
  </Step>
</Steps>

Use **Preview instructions** to see how the skill will read to the agent before you save it.

<Tip>
  Keep each skill to one job. A "plain, human tone" skill and a "never invent a statistic" skill are easier to mix and match than one skill that tries to do both.
</Tip>

## Manage your skills

The Skills tab has category chips and a **Search skills** box so you can find a skill by name, description, or instruction text. Each custom skill has an actions menu:

* **Duplicate** — make an editable copy of any skill, built-in or custom.
* **Archive skill** — hide a skill you're not using without losing it. **Restore skill** brings it back. Archived skills sit behind the **Archived** filter.
* **Delete skill** — a permanent removal. You type the skill's name to confirm. It's removed from current agent drafts, but published agent versions and past runs keep their own frozen copies.

### Version history

Every time you save a change, FixAEO records a new immutable version. Open **History** on a skill to see each version with its date and instructions; the latest is marked **Current**. Older versions are kept, not overwritten, so you can always see what an earlier run followed.

## Attach a skill to an agent

Open an agent in the builder and use the **Skills** button (labelled **Skills** in the toolbar, with a count when some are attached). This opens **Skills for this workflow**, where you:

* Choose up to **10** skills.
* Set the order they run in — skills apply top to bottom on every AI step.
* Remove ones you no longer want.

<Note>
  Only the agent's owner can change its attached skills. Editors and viewers can see them. See [Agents](/agents) for how sharing and roles work.
</Note>

You can also switch individual skills off before a manual run, without changing the saved setup.

### Save an assistant reply as a skill

When an agent or the [AI Marketer](/ai-marketer) writes something you'd want to reuse as an instruction, use **Save as skill** on that reply. The response becomes the new skill's instructions, so you don't have to re-write it by hand. Pick a name and category, and it lands in your Skills library ready to attach.

## Skills freeze at run time

Editing a skill later never rewrites the past. A run keeps the exact version of every skill it used, and a published agent version keeps the versions it was published with. So:

* Changing a skill does **not** change results you've already produced.
* To make a change take effect in an agent, publish a new version of that agent so it picks up the current skill versions.

This is what keeps your history honest — you can always trace what any past result was told to do. Skills are also readable through FixAEO's read-only public API, so you can review them outside the app.

<Warning>
  Building, organizing, and saving skills is broadly available. **Running** an agent needs an active paid plan — including any run that uses your skills. Your saved work stays available either way. See [Plans and limits](/plans-and-limits).
</Warning>

## Related

<CardGroup cols={2}>
  <Card title="Agents" icon="robot" href="/agents">
    Build an agent, then attach your skills to its AI steps.
  </Card>

  <Card title="Templates" icon="grid-2" href="/agents/templates">
    Ready-made agents to start from instead of a blank canvas.
  </Card>

  <Card title="Scheduling and automation" icon="clock" href="/agents/automation">
    Run a published agent on a schedule, a webhook, or in bulk.
  </Card>

  <Card title="Resources and review" icon="shield-check" href="/agents/resources">
    Brand kits, knowledge bases, and the approval checkpoints your runs pause at.
  </Card>

  <Card title="AI Marketer" icon="message-bot" href="/ai-marketer">
    The conversational sibling — and where "Save as skill" often starts.
  </Card>

  <Card title="Plans and limits" icon="credit-card" href="/plans-and-limits">
    What's free, what needs a paid plan, and the run limits.
  </Card>
</CardGroup>
