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

# Skills

> Install and use core Actionbook skills

Skills give your AI agent reusable instructions for how to use Actionbook correctly.

## Install Actionbook skills

Use the Actionbook skills package:

```bash theme={null}
npx skills add actionbook/actionbook
```

In the interactive flow, choose the skill you want to install (for example `actionbook`, `active-research`, or `extract`). If you need multiple, run the same command again and select the next skill.

## What each skill does

### `actionbook/actionbook`

Core browser-operation workflow for agents:

* `actionbook search` to find relevant manuals
* `actionbook get` to fetch verified selectors
* execute browser actions with less selector guesswork

### `actionbook/active-research`

Research workflow for deep analysis and report generation:

* structured multi-source collection
* selector-aware browsing with Actionbook
* output-ready report workflow for long-form research

### `actionbook/extract`

Data extraction workflow that produces a reusable Playwright script and structured output:

* obtain selectors via Actionbook (`search/get` first)
* if selectors are sufficient, start from them and script directly
* run lightweight mechanism probes when needed (hydration/virtualized list/pagination)
* use snapshot/screenshot only as fallback when selectors are missing or unstable
* generate a standalone Playwright script that handles the detected mechanisms
* output extracted data as JSON (default) or CSV

Unlike `actionbook` (interactive operation) or `active-research` (report generation), `extract` is purpose-built for **pulling structured data** from a page and leaving behind a **re-runnable script**.

## Example prompt

```text theme={null}
Use Actionbook skill workflow: search first, then get the best manual, then execute browser actions.
```