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

# Skills

> Skills are Markdown capabilities the agent picks up automatically based on the task. fal ships a set, and you can add your own.

A skill is a Markdown file that tells the agent how to do a kind of work well. It encodes the guidelines a good result needs: which models to prefer, how to structure prompts, what to check before delivering. Skills stack. A look, a format, and a delivery spec can all apply to one run.

Open **Skills** in the sidebar to browse them. Tabs show **All**, **Yours**, and **By fal**.

## Built-in skills

fal's creative engineers maintain these skills. They are on by default and appear under **By fal**.

| Skill                     | Use it for                                                                                                                                                                              |
| :------------------------ | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **fal-video-generation**  | The general router for video work. Make a clip from a description, animate a still, interpolate between two frames, restyle footage, extend or upscale a clip, reframe it, or add sound |
| **fal-video-production**  | A finished short video ad from a brief. Script, shot list, one clip per shot, a pinned voice, a ducked music bed, and burned-in text, with a clip review before audio is paid for       |
| **fal-script-writer**     | A timed script and shot table before generation. Story beats or ad structure, duration and voiceover limits per shot, and continuity notes                                              |
| **fal-shot-craft**        | Wording the visual prompt of one video clip. Framing, camera, one settled state per clip, and specific gesture over named emotion                                                       |
| **fal-motion-graphics**   | Animated on-screen graphics for a finished video. Titles, lower thirds, badges, end cards, and logo cards with entrance and exit motion                                                 |
| **video-compositing**     | Captions, titles, animated logos, kinetic type, montages, and slideshows, rendered in the [sandbox](/docs/documentation/agent/tools/sandbox)                                                 |
| **fal-audio-generation**  | The router for audio. Voiceover and text to speech, voice replacement, dubbing, sound effects, music, and transcription                                                                 |
| **h3-cinematic-prompter** | Finished prompts for the MiniMax Hailuo H3 video model. Produces prompt text only                                                                                                       |
| **cinematography**        | Shot language, camera movement, lighting, lens choice, color grade, film texture, and scene blocking                                                                                    |
| **character-design**      | Original characters, reference sheets, expression sheets, outfit variations, identity-preserving edits, and character-to-video                                                          |
| **character-sheet**       | One character design sheet, turnaround, or model sheet as a single image                                                                                                                |
| **ugc**                   | Creator-style ads and social-proof videos. Talking-head scripts, product demos, testimonials, unboxing, and short vertical formats                                                      |
| **photo-editing**         | Pixel-exact edits in the sandbox. Background removal, exposure, crop, straighten, format conversion, and compositing. Semantic edits go to a generative model                           |
| **fal-gamedev**           | 2D pixel-art characters, sprite sheets, walk cycles, background removal, and parallax or isometric backgrounds                                                                          |
| **wrap-it-up**            | End-of-session handoff. Packages everything from the chat into one ZIP with a written progress document                                                                                 |
| **training**              | Fine-tune a LoRA on your own images. Available to accounts with [training access](/docs/documentation/agent/tools/training)                                                                  |

<Frame>
  <img src="https://mintcdn.com/fal-d8505a2e/DmEWlE6WV9vaz0Me/images/agent/skills-page.png?fit=max&auto=format&n=DmEWlE6WV9vaz0Me&q=85&s=d79b9dc437ae57a04970f2509e78b1a6" alt="The Skills page showing the fal skills with their descriptions and enable toggles" width="1400" height="863" data-path="images/agent/skills-page.png" />
</Frame>

## How a skill activates

* **Automatically.** The agent sees the name and description of every enabled skill. When a request matches, it loads the skill's full body.
* **On request.** Type `@` in the composer, open **Skills**, and pick one. It is loaded for that turn.

Once a skill is active in a chat, it stays active for the rest of the chat.

## Turn skills on and off

* Each skill card has a toggle. A disabled skill is invisible to the agent.
* **Settings → Skills → Enable skills** is the master switch. When it is off, the agent ignores all skills, including fal's and yours.

## Your own skills

You can write a skill in the editor or install one from a GitHub repository. See [Custom skills](/docs/documentation/agent/skills/custom-skills).

<Card title="Custom skills" href="/docs/documentation/agent/skills/custom-skills">
  Format, editor, GitHub install, updates, and limits
</Card>
