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

# Skills

> Reusable instructions the Mythos agent applies across all your projects — five built-in skills plus up to 100 of your own: written, imported, or generated.

A skill is a named, reusable set of instructions you give the Mythos agent once and reuse in every
project. Each skill is a small `SKILL.md` file with a name, a short description of when to use it, and
a body of instructions. When you build or edit a project, Mythos loads your skills and either applies
the right one automatically — based on its description matching what you're asking for — or runs the
one you pick from the chat with `/`.

The mental model: prompts are one-off; skills are the standing rules and recipes you don't want to
retype. "Always audit for accessibility before finishing" or "use our class-naming conventions" lives
in a skill, not in every prompt.

<Note>**Cost:** skills are part of your account — creating, importing, editing, and toggling them is
free. They take effect during a normal build (**3 credits**) or edit (**2 credits**), which is when
the agent runs. See [Credits & usage](/introduction/credits-and-usage).</Note>

## Why use it

* **Stop repeating yourself.** Rules you'd otherwise paste into every prompt — conventions, a review
  checklist, a house style — live in one place and apply everywhere.
* **Consistency across projects.** Skills are tied to your account, not a single project, so the same
  standards follow you into every new build.
* **The agent reaches for the right one on its own.** A skill's description tells Mythos when it's
  relevant, so a matching request can trigger it without you naming it.

## Built-in skills vs your own

There are two kinds, and they appear together in the Skills panel.

* **Built-in skills** ship with Mythos. There are five: **accessibility**, **redesign**,
  **seo-review**, **skill-creator**, and **video-creator**. They're maintained by Mythos and are
  read-only — you can preview them but not edit or delete them.
* **Your skills** are the ones you add. They're private to your account and you can edit, disable, or
  delete any of them.

<Note>You can't create a skill whose name collides with a built-in (for example, `accessibility`).
Pick a different name.</Note>

## How to add a skill

Open the **Skills** panel from your account (Account → Skills on the dashboard, or Services → Account
in the workspace), then use the **Add** menu. There are four ways to add one.

<Steps>
  <Step title="Build with Mythos">
    Starts a chat session with the built-in **skill-creator**, which interviews you and writes the
    `SKILL.md` for you. Good when you know the behavior you want but not how to phrase it as a skill.
  </Step>

  <Step title="Write manually">
    Opens a form: a **name** (lowercase letters, digits, and hyphens only), a **description** of when
    to use it, and the **body** of instructions. Save, and it joins your library.
  </Step>

  <Step title="Import from GitHub">
    Paste a GitHub URL to a `SKILL.md` (or a repository containing skills) and Mythos pulls it in.
    Imported skills are marked with their source.
  </Step>

  <Step title="Upload a ZIP">
    Upload a ZIP bundle that contains one or more `SKILL.md` files.
  </Step>
</Steps>

## How a skill gets used in a build

You don't have to do anything special for a skill to apply — but you can also invoke one directly.

<Steps>
  <Step title="Automatic (description match)">
    When you send a prompt, the agent sees the names and descriptions of your enabled skills. If your
    request matches a skill's description, the agent applies it on its own.
  </Step>

  <Step title="On demand (slash picker)">
    Type `/` at the start of a chat message to open a picker of your enabled skills plus the built-in
    ones. Choose one to attach it to the message, so the agent runs that skill for this request.
  </Step>

  <Step title="Toggle skills on and off">
    Each of your skills has an on/off switch. A disabled skill stays in your library but is ignored
    during builds — useful when a skill is too aggressive or only relevant sometimes.
  </Step>
</Steps>

## Examples

A skill that enforces a convention every time you add a component:

```
Name: tailwind-conventions
Description: Use when writing new components. Enforce our class-merging conventions.

Body:
- Merge classes with the cn() helper, never string concatenation.
- Mobile-first: base styles first, then sm:/md:/lg: overrides.
- No raw hex colors in JSX — use theme tokens.
```

Invoking a built-in skill directly from chat:

```
/seo-review check the landing page for missing meta tags and heading structure
```

Letting auto-invocation do the work — this prompt matches the accessibility skill's description, so
it applies without being named:

```
Audit the whole site for accessibility problems and fix them.
```

## Limitations & good to know

* **Skills don't run shell commands.** For safety, any shell-execution markers in a skill body are
  stripped, and shell execution from skills is disabled regardless. A skill is instructions for the
  agent, not a script.
* **Imported skills run with the agent's full permissions.** Treat a skill you import like code you're
  about to run — read it first. Mythos shows the source of imported skills so you can check where one
  came from.
* **You can have at most 100 of your own skills.** Built-in skills don't count toward this. Delete an
  unused one to make room.
* **Names are lowercase only.** A skill name must match lowercase letters, digits, and hyphens (for
  example `seo-review`). Names are fixed once created — to rename, create a new one.
* **Too many enabled skills add overhead.** Every enabled skill's description is sent to the agent on
  each build, which adds to the tokens you're billed for. Disable skills you're not actively using.
* **Changes apply per account, not per tab.** If you edit a skill in one browser tab, refresh another
  open tab to see it.

<Warning>Deleting a skill is permanent — there's no undo. Disable it instead if you might want it
back.</Warning>

## FAQ

<AccordionGroup>
  <Accordion title="Do skills cost credits?">No. Managing skills is free. Credits are spent only when a build or edit runs, the same as any other generation.</Accordion>
  <Accordion title="Are my skills shared across projects?">Yes. Skills are tied to your account, so they're available in every project you build.</Accordion>
  <Accordion title="Can I edit a built-in skill?">No. The five built-in skills are read-only. You can preview them, but to customize behavior, write your own skill.</Accordion>
  <Accordion title="How do I make the agent use a specific skill?">Type `/` at the start of a chat message and pick the skill, or rely on auto-invocation if the skill's description matches your request.</Accordion>
  <Accordion title="Why isn't my skill being applied?">Check that it's enabled (the toggle is on) and that its description clearly states when it should be used — auto-invocation depends on that description matching your request. Or invoke it directly with `/`.</Accordion>
</AccordionGroup>

## Related

* [Chat attachments](/features/chat-attachments) — another way to give the agent context, per message.
* [Plan mode](/features/plan-mode) — shape a build before code is written.
* [Credits & usage](/introduction/credits-and-usage) — what a build or edit costs when a skill runs.