Does this extension collect user data?
Yes - The extension collects:
- Page URLs and titles (to provide context to AI)
- DOM element information (to enable automation)
- Screenshots (for AI vision analysis, if enabled)
- User-entered goals and preferences
This data is:
- Processed locally on the user's device, OR
- Sent to the user-configured AI provider (OpenAI/Anthropic/Google/Other) for processing
- NOT sold or transferred to third parties
- NOT used for advertising
Description
This extension automates browser tasks using AI. Users describe goals in natural language, and the AI agent navigates websites, clicks elements, fills forms, and completes multi-step workflows automatically.
Permission Justifications
activeTab
Required to interact with the currently active tab when the user initiates a browser automation task. The extension reads page content (DOM elements, text, URLs) and executes actions (clicks, typing) only on the tab the user is actively working with. This is triggered only by explicit user action (clicking "Start Agent" or using keyboard shortcuts).
tabs
Required to manage browser tabs during multi-step automation workflows. The extension may need to open new tabs, switch between tabs, or close tabs as part of completing user-requested tasks. For example, when searching for information that spans multiple pages or when the user's goal requires navigating to different websites.
scripting
Required to execute automation scripts on web pages. The extension injects scripts to identify interactive elements (buttons, links, input fields), read page content, and perform actions (clicking, typing text) as directed by the AI agent. Scripts are only executed on pages where the user has initiated an automation task.
debugger
Required for advanced element detection and interaction using Chrome DevTools Protocol (CDP). This enables precise identification of clickable elements, form fields, and interactive components on complex web pages. The debugger API provides reliable browser automation capabilities that are not possible through content scripts alone.
storage
Required to save user preferences such as preferred AI model, server connection settings, and session history. All data is stored locally on the user's device using Chrome's storage API. No user data is transmitted to external servers unless the user explicitly configures a connection to the PraisonAI bridge server.
sidePanel
Required to display the main user interface. The side panel provides the primary interaction point where users enter their goals, view automation progress, and see the history of completed actions. This panel remains accessible while browsing, allowing users to monitor and control automation tasks without switching windows.
contextMenus
Required to add right-click menu options for quick access to extension features. Users can right-click on page elements to quickly trigger automation actions, such as "Fill this form" or "Click this element". This provides a convenient alternative to opening the side panel for simple tasks.
notifications
Required to alert users when automation tasks complete, fail, or require attention. Notifications inform users of task completion when the browser is not in focus, ensuring they are aware of the automation status without constantly monitoring the side panel.
offscreen
Required for browser session recording functionality. The offscreen document hosts MediaRecorder API to capture tab video/audio streams, process recorded video data into WebM format, and handle Base64 encoding of media files. This is necessary because MediaRecorder and Canvas APIs are not available in service workers (Manifest V3 requirement). The offscreen document is created when user initiates recording via the side panel.tabCapture
Required to capture visual screenshots of web pages for AI vision analysis. The AI agent uses screenshots to understand page layout, identify visual elements, and make decisions about how to interact with websites. Screenshots are processed locally or sent to the configured AI provider for analysis.
alarms ()
Required to keep the service worker alive during long-running automation tasks. Manifest V3 service workers can be terminated after 30 seconds of inactivity. The alarms API provides periodic wake-up calls to prevent termination during multi-step automation workflows that may take several minutes to complete.host_permissions ()
This is a browser automation tool that must work on any website the user chooses.
Users describe goals like "search Google", "book flights", or "fill this form" -
requiring the ability to interact with arbitrary websites. The extension only
accesses pages when the user explicitly initiates an automation task via the
side panel or keyboard shortcut.Remote Code
This extension does not execute remote code. All JavaScript code is bundled within the extension package during build time. The extension may connect to:
1. Local PraisonAI bridge server (localhost) - for AI agent communication
2. AI provider APIs (OpenAI, Anthropic, Google) - for language model inference
These connections transmit data for AI processing but do not download or execute code. All automation scripts are pre-bundled in the extension.
Privacy Practices
| Field | Value |
|---|---|
| Single purpose | Browser automation with side panel, screenshots, and session recording |
| activeTab | See justification above |
| tabs | See justification above |
| scripting | See justification above |
| debugger | See justification above |
| storage | See justification above |
| sidePanel | See justification above |
| contextMenus | See justification above |
| notifications | See justification above |
| offscreen | See justification above |
| tabCapture | See justification above |
| alarms | See justification above |
| host_permissions | See justification above |
| Remote code | No remote code execution |
| Data certification | Certify compliance |
TESTER VERIFICATION GUIDE
This section provides step-by-step instructions for Chrome Web Store reviewers to verify each feature and permission.
Quick Verification Checklist
| Feature | How to Verify | Expected Result |
|---|---|---|
| Side Panel | Click toolbar icon | Side panel opens on right side |
| Screenshot | Press Ctrl+Shift+S | Notification appears “Screenshot Captured” |
| Context Menu | Right-click any page | “PraisonAI” menu with options appears |
| Keyboard Shortcuts | Press Ctrl+Shift+P | Side panel toggles open/closed |
| Storage | Open side panel, change settings | Settings persist after browser restart |
Detailed Verification Steps
1. SIDE PANEL (sidePanel permission)
Steps:
- Install the extension
- Click the PraisonAI icon in the Chrome toolbar
- The side panel should open on the right side of the browser
Expected: Side panel displays with PraisonAI interface showing input field and controls.
2. SCREENSHOT CAPTURE (activeTab, tabCapture, notifications permissions)
Steps:
- Navigate to any website (e.g., https://www.google.com)
- Press Ctrl+Shift+S (or Cmd+Shift+S on Mac)
- OR right-click on the page and select “PraisonAI” → “Capture Screenshot”
Expected: A notification appears confirming “Screenshot Captured” with message “Screenshot saved successfully”.
3. CONTEXT MENU (contextMenus permission)
Steps:
- Navigate to any website
- Right-click anywhere on the page
- Look for “PraisonAI” in the context menu
Expected: Context menu shows “PraisonAI” with sub-options including “Capture Screenshot”.
4. KEYBOARD SHORTCUTS (commands)
Steps:
- Press Ctrl+Shift+P (or Cmd+Shift+P on Mac) to toggle side panel
- Press Alt+A (or Option+A on Mac) to start agent
- Press Ctrl+Shift+S (or Cmd+Shift+S on Mac) to capture screenshot
Expected: Each shortcut performs its designated action.
5. STORAGE (storage permission)
Steps:
- Open the side panel
- If there are any settings/preferences, modify them
- Close and reopen the browser
- Open the side panel again
Expected: Settings are preserved between browser sessions.
6. OFFSCREEN DOCUMENT (offscreen permission)
Purpose: Used for session recording with MediaRecorder API (not available in service workers).
Steps to verify offscreen is used:
- Open Chrome DevTools (F12)
- Go to Application tab → Service Workers
- The extension registers an offscreen document for recording functionality
- The offscreen.html file is included in the extension package
Technical verification: The offscreen document contains MediaRecorder code for video capture, which requires DOM APIs unavailable in Manifest V3 service workers.
7. DEBUGGER (debugger permission)
Purpose: Used for Chrome DevTools Protocol (CDP) to reliably detect and interact with page elements.
Steps:
- Open side panel
- Navigate to a website with interactive elements
- The extension uses CDP to identify clickable elements, form fields, etc.
Expected: Extension can detect and list interactive elements on the page.
8. TABS (tabs permission)
Purpose: Manage browser tabs during multi-step automation.
Steps:
- Open side panel
- The extension can read tab URLs and titles to provide context
- During automation, it may open/switch tabs as needed
Expected: Extension displays current tab information in side panel.
9. SCRIPTING (scripting permission)
Purpose: Inject content scripts to interact with page elements.
Steps:
- Navigate to any website
- The content script (content.js) is automatically injected
- This enables the extension to read page content and execute actions
Expected: Extension can read and interact with page elements.
10. HOST PERMISSIONS ()
Purpose: Browser automation tool must work on any website the user chooses.
Justification: Users may want to automate tasks on any website (shopping, booking, forms, etc.). The extension only accesses pages when user explicitly initiates an action via side panel, keyboard shortcut, or context menu.
Files in Extension Package
| File | Purpose |
|---|---|
| manifest.json | Extension configuration and permissions |
| background.js | Service worker handling events and CDP |
| content.js | Content script for page interaction |
| sidepanel.html/js | Side panel user interface |
| offscreen.html/js | MediaRecorder for session recording |
| icons/ | Extension icons (16, 32, 48, 128px) |
No Remote Code Execution
All JavaScript code is bundled within the extension package at build time. The extension does not download or execute any remote code. Network connections are only used for:
- Optional local development server (localhost only)
- User-configured AI API endpoints (if user chooses to configure)
Privacy
Screenshots and recordings are saved locally only
All processing happens locally in the browser
No data is sent to external servers unless user explicitly configures an AI provider
No browsing history is stored or transmitted