What does the Actionbook AI skill do?

This skill should be used when the user needs to automate multi-step website tasks. Activates for browser automation, web scraping, UI testing, or building AI agents. Provides complete action manuals with step-by-step instructions and verified selectors.

Why use Actionbook on TypingMind?

Because you install it once and use it with any model. Actionbook is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Actionbook in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/smallnest/goclaw/tree/master/internal/builtin_skills/actionbook. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Actionbook?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Actionbook?

As many as you like. As long as a model supports skills, you can use Actionbook with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Actionbook AI skill free?

Yes. It is published on GitHub by smallnest under the MIT license. You only pay your own AI provider for the tokens you use.

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

Actionbook by smallnest — AI Skill for Any LLM

When to Use This Skill

Activate this skill when the user:

Needs to complete a multi-step task ("Send a LinkedIn message", "Book an Airbnb")
Asks how to interact with a website ("How do I post a tweet?")
Builds browser-based AI agents or web scrapers
Writes E2E tests for external websites
Navigates to any new page during browser automation
Wants to control their existing Chrome browser (Extension mode)

Browser Modes

Actionbook supports two browser control modes:

Mode	Flag	Use Case
CDP (default)	(none)	Launches a dedicated browser instance via Chrome DevTools Protocol
Extension	`--extension`	Controls the user's existing Chrome browser via a Chrome Extension + WebSocket bridge

When to use Extension mode:

The user wants to operate on their already-open Chrome (with existing logins, cookies, tabs)
The task requires interacting with pages that need the user's real session state
The user explicitly mentions their Chrome browser, extension, or existing tabs

When to use CDP mode (default):

Clean browser environment is preferred
Headless automation or CI/CD
Profile-based session isolation is needed

All actionbook browser commands work identically in both modes. The only difference is adding --extension flag (or setting ACTIONBOOK_EXTENSION=1).

How to Use

Phase 1: Get Action Manual

bash
# Step 1: Search for action manuals
actionbook search "arxiv search papers"
# Returns: area IDs with descriptions

# Step 2: Get the full manual (use area_id from search results)
actionbook get "arxiv.org:/search/advanced:default"
# Returns: Page structure, UI Elements with CSS/XPath selectors

Phase 2: Execute with Browser (CDP mode — default)

bash
# Step 3: Open browser
actionbook browser open "https://arxiv.org/search/advanced"

# Step 4: Use CSS selectors from Action Manual directly
actionbook browser fill "#terms-0-term" "Neural Network"
actionbook browser select "#terms-0-field" "title"
actionbook browser click "#date-filter_by-2"
actionbook browser fill "#date-year" "2025"
actionbook browser click "form[action='/search/advanced'] button.is-link"

# Step 5: Wait for results
actionbook browser wait-nav

# Step 6: Extract data
actionbook browser text

# Step 7: Close browser
actionbook browser close

Phase 2 (alt): Execute with Extension mode

Extension mode uses identical browser commands — just add --extension. But you must follow the full lifecycle below.

bash
# Step 3: Open URL in user's Chrome
actionbook --extension browser open "https://arxiv.org/search/advanced"

# Step 4-7: Same commands, just add --extension
actionbook --extension browser fill "#terms-0-term" "Neural Network"
actionbook --extension browser select "#terms-0-field" "title"
actionbook --extension browser click "#date-filter_by-2"
actionbook --extension browser fill "#date-year" "2025"
actionbook --extension browser click "form[action='/search/advanced'] button.is-link"
actionbook --extension browser wait-nav
actionbook --extension browser text

# Step 8: Cleanup (CRITICAL — see Extension Mode Lifecycle below)
actionbook --extension browser close    # release debug connection FIRST
actionbook extension stop               # then stop bridge server

Action Manual Format

Action manuals return:

Page URL - Target page address
Page Structure - DOM hierarchy and key sections
UI Elements - CSS/XPath selectors with element metadata

yaml
  ### button_advanced_search

  - ID: button_advanced_search
  - Description: Advanced search navigation button
  - Type: link
  - Allow Methods: click
  - Selectors:
    - role: getByRole('link', { name: 'Advanced Search' }) (confidence: 0.9)
    - css: button.button.is-small.is-cul-darker (confidence: 0.65)
    - xpath: //button[contains(@class, 'button')] (confidence: 0.55)

Action Search Commands

bash
actionbook search "<query>"                    # Basic search
actionbook search "<query>" --domain site.com  # Filter by domain
actionbook search "<query>" --url <url>        # Filter by URL
actionbook search "<query>" -p 2 -s 20         # Page 2, 20 results

actionbook get "<area_id>"                     # Full details with selectors
# area_id format: "site.com:/path:area_name"

actionbook sources list                        # List available sources
actionbook sources search "<query>"            # Search sources by keyword

Extension Setup & Management

Commands for managing the Chrome Extension bridge:

bash
actionbook extension install              # Install extension files to local config dir
actionbook extension path                 # Show extension directory (for Chrome "Load unpacked")
actionbook extension serve                # Start WebSocket bridge (keep running in background)
actionbook extension stop                 # Stop the running bridge server (sends SIGTERM)
actionbook extension status               # Check bridge and extension connection status
actionbook extension ping                 # Ping the extension to verify link is alive

Setup flow (one-time):

actionbook extension install — extract extension files and register native messaging host
Open chrome://extensions → enable Developer mode → Load unpacked → select the path from actionbook extension path
actionbook extension serve — start bridge (keep running)
Extension auto-connects via native messaging (no manual token needed in most cases). If auto-pairing fails: copy token from serve output → paste in extension popup → Save

Connection check before automation:

bash
actionbook extension status    # should show "running"
actionbook extension ping      # should show "responded"

Browser Commands

All browser commands below work in both CDP and Extension mode. For Extension mode, add --extension flag or set ACTIONBOOK_EXTENSION=1.

Navigation

bash
actionbook browser open <url>                  # Open URL in new tab
actionbook browser goto <url>                  # Navigate current page
actionbook browser back                        # Go back
actionbook browser forward                     # Go forward
actionbook browser reload                      # Reload page
actionbook browser pages                       # List open tabs
actionbook browser switch <page_id>            # Switch tab
actionbook browser close                       # Close browser
actionbook browser restart                     # Restart browser
actionbook browser connect <endpoint>          # Connect to existing browser (CDP port or URL)

Interactions (use CSS selectors from Action Manual)

bash
actionbook browser click "<selector>"                  # Click element
actionbook browser click "<selector>" --wait 1000      # Wait then click
actionbook browser fill "<selector>" "text"            # Clear and type
actionbook browser type "<selector>" "text"            # Append text
actionbook browser select "<selector>" "value"         # Select dropdown
actionbook browser hover "<selector>"                  # Hover
actionbook browser focus "<selector>"                  # Focus
actionbook browser press Enter                         # Press key

Get Information

bash
actionbook browser text                        # Full page text
actionbook browser text "<selector>"           # Element text
actionbook browser html                        # Full page HTML
actionbook browser html "<selector>"           # Element HTML
actionbook browser snapshot                    # Accessibility tree
actionbook browser viewport                    # Viewport dimensions
actionbook browser status                      # Browser detection info

Wait

bash
actionbook browser wait "<selector>"                   # Wait for element
actionbook browser wait "<selector>" --timeout 5000    # Custom timeout
actionbook browser wait-nav                            # Wait for navigation

Screenshots & Export

bash
# Ensure target directory exists before saving screenshots
actionbook browser screenshot                  # Save screenshot.png
actionbook browser screenshot output.png       # Custom path
actionbook browser screenshot --full-page      # Full page
actionbook browser pdf output.pdf              # Export as PDF

JavaScript & Inspection

bash
actionbook browser eval "document.title"               # Execute JS
actionbook browser inspect 100 200                     # Inspect at coordinates
actionbook browser inspect 100 200 --desc "login btn"  # With description

Cookies

bash
actionbook browser cookies list                # List all cookies
actionbook browser cookies get "name"          # Get cookie
actionbook browser cookies set "name" "value"  # Set cookie
actionbook browser cookies set "name" "value" --domain ".example.com"
actionbook browser cookies delete "name"       # Delete cookie
actionbook browser cookies clear               # Clear all

Global Flags

bash
actionbook --json <command>              # JSON output
actionbook --headless <command>          # Headless mode (CDP only)
actionbook --verbose <command>           # Verbose logging
actionbook -P <profile> <command>        # Use specific profile (CDP only)
actionbook --cdp <port|url> <command>    # CDP connection
actionbook --extension <command>         # Use Chrome Extension mode
# or: ACTIONBOOK_EXTENSION=1 actionbook <command>

Guidelines

Search by task description, not element name ("arxiv search papers" not "search button")
Use Action Manual selectors first - they are pre-verified and don't require snapshot
Prefer CSS ID selectors (#id) over XPath when both are provided
Fallback to snapshot when selectors fail - use actionbook browser snapshot then CSS selectors from the output
Re-snapshot after navigation - DOM changes invalidate previous state
Extension mode: follow the full lifecycle — pre-flight → connect → execute → cleanup (see Extension Mode Lifecycle)
Extension mode: verify extension is installed before starting bridge; prefer auto-pair over manual token
Extension mode: always run browser close before stopping the bridge to release the debug connection
Extension mode: the user's real browser is being controlled — avoid destructive actions (clearing all cookies, closing all tabs) without confirmation
Extension mode: L3 operations (some cookie/storage modifications) may require manual approval in the extension popup

Fallback Strategy

When Fallback is Needed

Actionbook stores pre-computed page data captured at indexing time. This data may become outdated as websites evolve:

Selector execution failure - The returned CSS/XPath selector does not match any element
Element mismatch - The selector matches an element with unexpected type or behavior
Multiple selector failures - Several selectors from the same action fail consecutively

Fallback Approaches

When Action Manual selectors don't work:

Snapshot the page - actionbook browser snapshot to get the current accessibility tree
Inspect visually - actionbook browser screenshot to see the current state
Inspect by coordinates - actionbook browser inspect <x> <y> to find elements
Execute JS - actionbook browser eval "document.querySelector(...)" for dynamic queries

When to Exit

If actionbook search returns no results or action fails unexpectedly, use other available tools to continue the task.

Examples

End-to-end with Action Manual

bash
# 1. Find selectors
actionbook search "airbnb search" --domain airbnb.com

# 2. Get detailed selectors (area_id from search results)
actionbook get "airbnb.com:/:default"

# 3. Automate using pre-verified selectors
actionbook browser open "https://www.airbnb.com"
actionbook browser fill "input[data-testid='structured-search-input-field-query']" "Tokyo"
actionbook browser click "button[data-testid='structured-search-input-search-button']"
actionbook browser wait-nav
actionbook browser text
actionbook browser close

Extension mode: Operate on user's Chrome

bash
# Verify bridge is running
actionbook extension status

# Use the user's existing logged-in session
actionbook --extension browser open "https://github.com/notifications"
actionbook --extension browser wait-nav
actionbook --extension browser text ".notifications-list"
actionbook --extension browser screenshot notifications.png

Extension Mode Lifecycle (CRITICAL)

When using Extension mode, always follow this complete lifecycle: pre-flight → connect → execute → cleanup.

1. Pre-flight: Ask user about extension installation

Before any technical checks, ask the user whether they have the Actionbook Chrome Extension installed.

User confirms installed → proceed to Step 2 (Connect).
User says not installed → run the installation flow:

bash
# Install extension files locally
actionbook extension install
actionbook extension path
# → On macOS, copy to visible dir if needed:
#    cp -r "$(actionbook extension path)" ~/Document/actionbook-extension

Then guide the user to load it in Chrome:

Open chrome://extensions → enable Developer mode
Click "Load unpacked" → select the extension directory
After user confirms loaded → proceed to Step 2

Limitation: The CLI can only verify that extension files exist locally. There is no way to detect whether Chrome has actually loaded the extension until a connection is attempted in Step 2.

2. Connect: Start bridge, auto-pair with retry

Start the bridge server and attempt auto-pairing. Retry up to 3 times before considering manual fallback.

bash
# Start bridge (run in background)
actionbook extension serve

# Attempt 1: Wait for auto-pairing via Native Messaging
sleep 3
actionbook extension ping
# → If ping succeeds → proceed to Step 3 (Execute)

# Attempt 2: If ping fails, wait longer and retry
sleep 5
actionbook extension ping
# → If ping succeeds → proceed to Step 3 (Execute)

# Attempt 3: Final retry
sleep 5
actionbook extension ping
# → If ping succeeds → proceed to Step 3 (Execute)

Only after all 3 auto-pair attempts fail, escalate based on the error:

"Extension not connected" → Ask user to verify the extension is enabled in chrome://extensions and retry
"Invalid token" → Only now provide the token from serve output for manual paste in the extension popup
Other errors → Check actionbook extension status for diagnostics

IMPORTANT: Do NOT expose the session token prematurely. The token is a last-resort fallback — most users will connect successfully via auto-pair within 3 attempts.

3. Execute: Browser automation

bash
actionbook --extension browser open "https://example.com"
# ... perform browser operations ...

4. Cleanup: Release debug connection, THEN stop bridge

bash
# Step 1: Release the debugging connection (MUST come first)
actionbook --extension browser close

# Step 2: Stop the bridge server
actionbook extension stop

# Step 3: Verify Chrome no longer shows "debugging" banner
actionbook extension status    # should show "not running"

WARNING: Skipping Step 1 and directly killing the bridge process will leave Chrome showing "Actionbook is debugging this browser". Always release the debug connection before stopping the bridge.

Extension mode: Troubleshooting

bash
# Bridge not running?
actionbook extension serve              # Start it

# Extension not responding?
actionbook extension ping               # Check connectivity

# Token expired? (idle > 30 min)
# Restart serve and re-pair in extension popup
actionbook extension serve              # Prints new token

Deep-Dive Documentation

For detailed patterns and best practices:

Reference	Description
references/command-reference.md	Complete command reference with all features
references/authentication.md	Login flows, OAuth, 2FA handling, state reuse

Publisher	smallnest
Repository	`goclaw`
Skill name	`actionbook`
Stars	599
Forks	119
Bundled files	3
License	MIT
Links	Skill source Repository

Overview

Installation

Install in TypingMind

Install in another agent

Use it in TypingMind

Works with any AI model

Skill instructions

When to Use This Skill

Browser Modes

How to Use

Phase 1: Get Action Manual

Phase 2: Execute with Browser (CDP mode — default)

Phase 2 (alt): Execute with Extension mode

Action Manual Format

Action Search Commands

Extension Setup & Management

Browser Commands

Navigation

Interactions (use CSS selectors from Action Manual)

Get Information

Wait

Screenshots & Export

JavaScript & Inspection

Cookies

Global Flags

Guidelines

Fallback Strategy

When Fallback is Needed

Fallback Approaches

When to Exit

Examples

End-to-end with Action Manual

Extension mode: Operate on user's Chrome

Extension Mode Lifecycle (CRITICAL)

1. Pre-flight: Ask user about extension installation

2. Connect: Start bridge, auto-pair with retry

3. Execute: Browser automation

4. Cleanup: Release debug connection, THEN stop bridge

Extension mode: Troubleshooting

Deep-Dive Documentation

Bundled files

Frequently asked questions

Related AI Skills

Set up your own AI workspace now