Electron Integration in XSky

This skill provides knowledge for Electron integration in XSky.

Package Location

packages/ai-agent-electron/ - Electron adapter package

Key Files

ai-agent-electron/src/
├── browser.ts      # BrowserAgent implementation
├── file.ts         # FileAgent for Electron
├── index.ts        # Package exports
├── preload.ts      # Preload script template
└── mcp/
    └── index.ts    # MCP client for Electron

BrowserAgent Constructor

import { BrowserAgent } from "@xsky/ai-agent-electron";

const browserAgent = new BrowserAgent(
  detailView,           // WebContentsView instance
  mcpClient,            // Optional MCP client
  customPrompt,         // Optional system prompt
  {
    useContextIsolation: true,
    preloadPath: path.join(__dirname, 'preload.js')
  }
);

Security Options

interface BrowserAgentSecurityOptions {
  useContextIsolation?: boolean;  // Recommended for production
  preloadPath?: string;           // Required when contextIsolation true
}

Preload Script Setup

Create preload.js:

const { contextBridge, ipcRenderer } = require('electron');

contextBridge.exposeInMainWorld('xskyAgent', {
  executeScript: async (fnString, args) => {
    // Safe script execution via IPC
    return await ipcRenderer.invoke('xsky:execute-script', fnString, args);
  }
});

WebContentsView Setup

import { WebContentsView } from 'electron';

const view = new WebContentsView({
  webPreferences: {
    contextIsolation: true,
    nodeIntegration: false,
    sandbox: true,
    preload: path.join(__dirname, 'preload.js')
  }
});

mainWindow.contentView.addChildView(view);
view.setBounds({ x: 0, y: 0, width: 800, height: 600 });

Script Execution

With contextIsolation (secure):

// Uses preload API
await window.xskyAgent.executeScript(fn.toString(), args);

Without contextIsolation (legacy):

// Direct executeJavaScript
await webContents.executeJavaScript(code, true);

Screenshot Capture

protected async screenshot(ctx: AgentContext) {
  const image = await this.detailView.webContents.capturePage();
  return {
    imageBase64: image.toDataURL(),
    imageType: "image/jpeg"
  };
}

PDF Support

BrowserAgent includes PDF extraction:

// Automatically detects PDF pages
// Uses pdf.js for content extraction
const content = await agent.extract_page_content(ctx);
// Returns: { title, page_url, page_content, total_pages, content_type: 'pdf' }

Key Source Files

File	Purpose
packages/ai-agent-electron/src/browser.ts	BrowserAgent
packages/ai-agent-electron/src/file.ts	FileAgent
packages/ai-agent-electron/src/preload.ts	Preload template
packages/ai-agent-electron/src/mcp/stdio.ts	MCP STDIO client