mcpField
C
core-actionbook

core-actionbook

github.com/actionbookAgent workflows1.6k1.9k installsUpdated 9w ago
Repository

About core-actionbook

core-actionbook provides precomputed browser action manuals for agents, replacing full-page HTML parsing with structured page details and selectors. It suits browser-automation workflows that need to find and execute page actions.

Key features

  1. 01Search manuals by keyword
  2. 02Retrieve page structure and selectors
  3. 03Return allowed element methods
  4. 04Filter by relevance and source

Use cases

  1. 01Find steps for a website task
  2. 02Locate selectors before browser actions

MCP support

It exposes `search_actions` and `get_action_by_id` as MCP tools for finding manuals and retrieving page details, selectors, and supported methods.

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/actionbook/rust-skills --skill core-actionbook

After installing, type/core-actionbookin Claude Code to confirm it loaded.

Bundled filesagents/openai.yaml

Details

From the project README

Actionbook

Pre-computed action manuals for browser automation. Agents receive structured page information instead of parsing entire HTML.

Workflow

  1. search_actions - Search by keyword, returns URL-based action IDs with content previews
  2. get_action_by_id - Get full action manual with page details, DOM structure, and element selectors
  3. Execute - Use returned selectors with your browser automation tool

MCP Tools

  • search_actions - Search by keyword. Returns: URL-based action IDs, content previews, relevance scores
  • get_action_by_id - Get full action details. Returns: action content, page element selectors (CSS/XPath), element types, allowed methods (click, type, extract), document metadata
Parameters

search_actions:

  • query (required): Search keyword (e.g., "airbnb search", "google login")
  • type: vector | fulltext | hybrid (default)
  • limit: Max results (default: 5)
  • sourceIds: Filter by source IDs (comma-separated)
  • minScore: Minimum relevance score (0-1)

get_action_by_id:

  • id (required): URL-based action ID (e.g., example.com/page)

Example Response

{
  "title": "Airbnb Search",
  "url": "www.airbnb.com/search",
  "elements": [
    {
      "name": "location_input",
      "selector": "input[data-testid='structured-search-input-field-query']",
      "type": "textbox",
      "methods": ["type", "fill"]
    }
  ]
}

Comments

Markdown supported