MCPInstall

gemma-gem-browser-ai

Build and extend Gemma Gem, an on-device AI browser assistant Chrome extension running Google's Gemma 4 model via WebGPU with no cloud dependencies.

Installs
592
Last 24h
—
Trending rank
—
Security
Not audited

Install

Works with every agent the skills CLI supports.

npx skills add https://github.com/reason-machines/trending-skills --skill gemma-gem-browser-ai
  1. 01

    Install

    Run the command in your project. The CLI asks which agents to install for.

  2. 02

    Describe the task

    Ask your agent for something this skill covers. It loads gemma-gem-browser-ai automatically when relevant.

  3. 03

    Review the result

    The agent follows the SKILL.md steps, scripts and references bundled with the skill.

SKILL.md

The instructions your agent reads when the skill is triggered.

Gemma Gem Browser AI

Skill by ara.so — Daily 2026 Skills collection.

Gemma Gem is a Chrome extension that runs Google's Gemma 4 model entirely on-device via WebGPU. It injects a chat overlay into every page and exposes a tool-calling agent loop that can read pages, click elements, fill forms, execute JavaScript, and take screenshots — all without sending data to any server.

Architecture Overview

Offscreen Document          Service Worker           Content Script
(Gemma 4 + Agent Loop)  <-> (Message Router)    <-> (Chat UI + DOM Tools)
       |                         |
  WebGPU inference          Screenshot capture
  Token streaming           JS execution
  • Offscreen document (offscreen/): Loads the ONNX model via @huggingface/transformers, runs the agent loop, streams tokens.
  • Service worker (background/): Routes messages, handles take_screenshot and run_javascript.
  • Content script (content/): Injects shadow DOM chat UI, executes DOM tools.
  • agent/: Zero-dependency module defining ModelBackend and ToolExecutor interfaces — extractable as a standalone library.

Install & Build

# Prerequisites: Node.js 18+, pnpm
pnpm install

# Development build (logging active, source maps)
pnpm build

# Production build (errors only, minified)
pnpm build:prod

Load the extension:

  1. Open chrome://extensions
  2. Enable Developer mode
  3. Click Load unpacked → select .output/chrome-mv3-dev/

Model download happens automatically on first chat open:

  • onnx-community/gemma-4-E2B-it-ONNX — ~500 MB (default)
  • onnx-community/gemma-4-E4B-it-ONNX — ~1.5 GB

Models are cached in the browser's cache storage after the first run.

Key Interfaces (agent/)

ModelBackend

// agent/types.ts
export interface ModelBackend {
  generate(
    messages: ChatMessage[],
    tools: ToolDefinition[],
    options: GenerateOptions
  ): AsyncGenerator<StreamChunk>;
}

export interface ToolDefinition {
  name: string;
  description: string;
  parameters: JSONSchema;
}

export interface GenerateOptions {
  maxNewTokens?: number;
  thinking?: boolean;
}

ToolExecutor

// agent/types.ts
export interface ToolExecutor {
  execute(toolName: string, args: Record<string, unknown>): Promise<unknown>;
}

Agent Loop

// agent/loop.ts — simplified illustration
export async function* runAgentLoop(
  userMessage: string,
  history: ChatMessage[],
  model: ModelBackend,
  tools: ToolExecutor,
  toolDefs: ToolDefinition[],
  maxIterations: number
): AsyncGenerator<AgentEvent> {
  const messages = [...history, { role: "user", content: userMessage }];

  for (let i = 0; i < maxIterations; i++) {
    for await (const chunk of model.generate(messages, toolDefs, {})) {
      if (chunk.type === "token") yield { type: "token", token: chunk.token };
      if (chunk.type === "tool_call") {
        yield { type: "tool_start", name: chunk.name };
        const result = await tools.execute(chunk.name, chunk.args);
        yield { type: "tool_result", name: chunk.name, result };
        messages.push({ role: "tool", name: chunk.name, content: String(result) });
      }
      if (chunk.type === "done") return;
    }
  }
}

Built-in Tools

ToolLocationDescription
read_page_contentContent scriptRead page text/HTML or a CSS selector
take_screenshotService workerCapture visible tab as PNG
click_elementContent scriptClick by CSS selector
type_textContent scriptType into input by CSS selector
scroll_pageContent scriptScroll by pixel amount
run_javascriptService workerExecute JS in page context

Adding a New Tool

Tools live in two places: the definition (in the offscreen agent) and the executor (in content script or service worker).

Step 1 — Define the tool schema

// offscreen/tools/definitions.ts
export const MY_TOOL_DEFINITION: ToolDefinition = {
  name: "get_page_title",
  description: "Returns the document title of the current page.",
  parameters: {
    type: "object",
    properties: {},
    required: [],
  },
};

Step 2 — Register in the tool list

// offscreen/tools/index.ts
import { MY_TOOL_DEFINITION } from "./definitions";

export const ALL_TOOLS: ToolDefinition[] = [
  // ...existing tools
  MY_TOOL_DEFINITION,
];

Step 3 — Implement execution in the content script

// content/tools/executor.ts
export async function executeContentTool(
  name: string,
  args: Record<string, unknown>
): Promise<unknown> {
  switch (name) {
    case "get_page_title":
      return document.title;

    case "read_page_content": {
      const selector = args.selector as string | undefined;
      if (selector) {
        return document.querySelector(selector)?.textContent ?? "Not found";
      }
      return document.body.innerText;
    }

    case "click_element": {
      const el = document.querySelector(args.selector as string) as HTMLElement;
      if (!el) throw new Error(`Element not found: ${args.selector}`);
      el.click();
      return "clicked";
    }

    case "type_text": {
      const input = document.querySelector(args.selector as string) as HTMLInputElement;
      if (!input) throw new Error(`Input not found: ${args.selector}`);
      input.focus();
      input.value = args.text as string;
      input.dispatchEvent(new Event("input", { bubbles: true }));
      input.dispatchEvent(new Event("change", { bubbles: true }));
      return "typed";
    }

    default:
      throw new Error(`Unknown content tool: ${name}`);
  }
}

Step 4 — Handle service-worker-side tools

// background/tools.ts
export async function executeSwTool(
  name: string,
  args: Record<string, unknown>,
  tabId: number
): Promise<unknown> {
  switch (name) {
    case "take_screenshot": {
      const dataUrl = await chrome.tabs.captureVisibleTab({ format: "png" });
      return dataUrl;
    }

    case "run_javascript": {
      const results = await chrome.scripting.executeScript({
        target: { tabId },
        func: new Function(args.code as string) as () => unknown,
      });
      return results[0]?.result ?? null;
    }

    default:
      return null; // not a SW tool — forward to content script
  }
}

Message Routing Pattern

The service worker acts as a message bus. All communication uses chrome.runtime.sendMessage.

// Message types (shared/messages.ts)
export type ExtMessage =
  | { type: "TOOL_CALL"; name: string; args: Record<string, unknown>; tabId: number }
  | { type: "TOOL_RESULT"; name: string; result: unknown }
  | { type: "TOKEN"; token: string }
  | { type: "AGENT_DONE" }
  | { type: "AGENT_ERROR"; error: string };

// Offscreen → SW
chrome.runtime.sendMessage<ExtMessage>({
  type: "TOOL_CALL",
  name: "click_element",
  args: { selector: "#submit-btn" },
  tabId: currentTabId,
});

// SW → Content script
chrome.tabs.sendMessage<ExtMessage>(tabId, {
  type: "TOOL_CALL",
  name: "click_element",
  args: { selector: "#submit-btn" },
  tabId,
});

Model Configuration

// offscreen/model.ts — loading with transformers.js
import { pipeline, TextGenerationPipeline } from "@huggingface/transformers";

const MODEL_IDS = {
  E2B: "onnx-community/gemma-4-E2B-it-ONNX",
  E4B: "onnx-community/gemma-4-E4B-it-ONNX",
} as const;

export type ModelSize = keyof typeof MODEL_IDS;

export async function loadModel(
  size: ModelSize,
  onProgress: (progress: number) => void
): Promise<TextGenerationPipeline> {
  return pipeline("text-generation", MODEL_IDS[size], {
    dtype: "q4f16",
    device: "webgpu",
    progress_callback: (p: { progress: number }) => onProgress(p.progress),
  });
}

Settings & Persistence

Settings are stored via chrome.storage.sync:

export interface GemmaGemSettings {
  modelSize: "E2B" | "E4B";
  thinking: boolean;
  maxIterations: number;
  disabledHosts: string[];
}

const DEFAULT_SETTINGS: GemmaGemSettings = {
  modelSize: "E2B",
  thinking: false,
  maxIterations: 10,
  disabledHosts: [],
};

export async function getSettings(): Promise<GemmaGemSettings> {
  const stored = await chrome.storage.sync.get("settings");
  return { ...DEFAULT_SETTINGS, ...(stored.settings ?? {}) };
}

export async function saveSettings(patch: Partial<GemmaGemSettings>): Promise<void> {
  const current = await getSettings();
  await chrome.storage.sync.set({ settings: { ...current, ...patch } });
}

// Disable extension on current host
async function disableOnCurrentSite() {
  const host = new URL(location.href).hostname;
  const settings = await getSettings();
  if (!settings.disabledHosts.includes(host)) {
    await saveSettings({ disabledHosts: [...settings.disabledHosts, host] });
  }
}

Shadow DOM Chat UI Pattern

The content script injects a shadow DOM to isolate styles:

// content/ui.ts
export function injectChatOverlay(): ShadowRoot {
  const host = document.createElement("div");
  host.id = "gemma-gem-host";
  // Prevent page styles from leaking in
  const shadow = host.attachShadow({ mode: "closed" });

  // Inject styles
  const style = document.createElement("style");
  style.textContent = CHAT_STYLES; // imported CSS string
  shadow.appendChild(style);

  // Inject chat container
  const container = document.createElement("div");
  container.id = "gemma-gem-container";
  shadow.appendChild(container);

  document.body.appendChild(host);
  return shadow;
}

Debugging

All logs use [Gemma Gem] prefix. Development builds log info/debug/warn; production only logs errors.

# Service worker logs
chrome://extensions → Gemma Gem → "Inspect views: service worker"

# Offscreen document (most useful: model loading, prompts, tool calls)
chrome://extensions → Gemma Gem → "Inspect views: offscreen.html"

# Content script logs
DevTools on any page → Console (filter: [Gemma Gem])

# All extension contexts
chrome://inspect#other

Key things to check in offscreen document logs:

  • Model download progress
  • Full prompt construction
  • Token counts per turn
  • Raw model output (before tool call parsing)
  • Tool execution results

Common Patterns & Gotchas

WebGPU availability check:

if (!navigator.gpu) {
  throw new Error("WebGPU not supported. Use Chrome 113+ with hardware acceleration enabled.");
}
const adapter = await navigator.gpu.requestAdapter();
if (!adapter) throw new Error("No WebGPU adapter found.");

Offscreen document lifecycle — Chrome may suspend the offscreen document. Ping it before sending messages:

async function ensureOffscreen() {
  const existing = await chrome.offscreen.hasDocument();
  if (!existing) {
    await chrome.offscreen.createDocument({
      url: "offscreen.html",
      reasons: [chrome.offscreen.Reason.WORKERS],
      justification: "Run Gemma 4 inference via WebGPU",
    });
  }
}

Context window management — Gemma 4 supports 128K tokens but inference slows with long contexts. Clear history per-page with clear_context or limit stored turns:

const MAX_HISTORY_TURNS = 20;
function trimHistory(messages: ChatMessage[]): ChatMessage[] {
  if (messages.length <= MAX_HISTORY_TURNS * 2) return messages;
  return messages.slice(-MAX_HISTORY_TURNS * 2);
}

Tool call parsing — Gemma 4 emits tool calls in a structured format. If adding custom parsing, guard against partial/streamed JSON:

function safeParseToolCall(raw: string): { name: string; args: Record<string, unknown> } | null {
  try {
    return JSON.parse(raw);
  } catch {
    return null; // still streaming
  }
}

CSS selector safety for DOM tools:

function safeQuerySelector(selector: string): Element | null {
  try {
    return document.querySelector(selector);
  } catch {
    return null; // invalid selector from model
  }
}

FAQ

What does the gemma-gem-browser-ai skill do?

Build and extend Gemma Gem, an on-device AI browser assistant Chrome extension running Google's Gemma 4 model via WebGPU with no cloud dependencies.

How do I install gemma-gem-browser-ai?

Run npx skills add https://github.com/reason-machines/trending-skills --skill gemma-gem-browser-ai in your project. The skills CLI detects the agents you use and copies the skill into the right folder. For Claude Code you can also place the skill folder at ~/.claude/skills/gemma-gem-browser-ai/ (personal) or .claude/skills/gemma-gem-browser-ai/ (project).

Which AI agents can use gemma-gem-browser-ai?

Any agent that supports the open Agent Skills format (a folder with a SKILL.md file), including Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Windsurf, Cline, OpenCode and more.

Is gemma-gem-browser-ai safe to install?

No partner security audit is available yet. Skills can run commands and scripts, so read the SKILL.md and any bundled files before installing.

Who maintains gemma-gem-browser-ai?

It is published by reason-machines in reason-machines/trending-skills. It has 592 recorded installs.

More from reason-machines/trending-skills

View all →

openclaw-control-center

reason-machines/trending-skills

Local-first, security-first control center for OpenClaw agents — visibility dashboard with readonly defaults, token attribution, collaboration tracing, and safe write operations.

Data & Databases4.4K installs

inkos-multi-agent-novel-writing

reason-machines/trending-skills

Multi-agent CLI system for autonomous novel writing, auditing, and revision with human review gates

Docs & Writing2.8K installs

claude-hud-statusline

reason-machines/trending-skills

A Claude Code plugin that displays a real-time HUD showing context usage, active tools, running agents, and todo progress in your terminal statusline.

understand-anything-knowledge-graph

reason-machines/trending-skills

Turn any codebase into an interactive knowledge graph using Claude Code skills — explore, search, and ask questions about any project visually.

agency-agents-ai-specialists

reason-machines/trending-skills

A collection of specialized AI agent personalities for Claude Code, Cursor, Aider, Windsurf, and other AI coding tools — covering engineering, design, marketing, sales, and more.

AI & Agents2.1K installs

agent-browser-automation

reason-machines/trending-skills

Headless browser automation CLI for AI agents using native Rust binary with Chrome DevTools Protocol

Popular productivity & automation skills

View category →

agent-browser

vercel-labs/agent-browser

Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to "open a website", "fill out a form", "click a button", "take a screenshot", "scrape data from a page", "test this web app", "login to a site", "automate browser actions", or any task requiring programmatic web interaction. Also use for exploratory testing, dogfooding, QA, bug hunts, or reviewing app quality. Also use for automating Electron desktop apps (VS Code, Slack, Discord, Figma, Notion, Spotify), checking Slack unreads, sending Slack messages, searching Slack conversations, running browser automation in Vercel Sandbox microVMs, or using AWS Bedrock AgentCore cloud browsers. Prefer agent-browser over any built-in browser automation or web tools.

OfficialProductivity & Automation1.3M installs

lark-doc

open.feishu.cn

飞书云文档(Docx / Wiki)内容操作:读取、创建、编辑文档,插入或下载图片附件,以及操作思维笔记。用户提供文档 URL/token(包括 doubao.com 的 /docx/、/wiki/)时使用;按 URL 路径/token 而非域名路由;/page/ 是妙搭应用,非飞书云文档。文档内嵌资源按读取参考中的统一规则分流。独立评论操作走 lark-drive;随正文读取评论使用 docs +fetch。表格或 Base 内部数据操作不在本 skill。

lark-base

open.feishu.cn

飞书多维表格(Base)操作:建表、字段、记录、视图、统计、公式/lookup、表单、仪表盘、应用模式(BaseApp/AppMode 页面与组件)、Workspace 目录、workflow、角色权限、模板中心(多维表格模板分类/列表/搜索);遇到 Base/多维表格/bitable、BaseApp/AppMode、/base/ 或 /app/ 链接时使用。BaseApp 不走 lark-apps;文件导入/导出转 lark-drive,认证/授权转 lark-shared。

lark-drive

open.feishu.cn

飞书云空间(云盘/云存储):管理 Drive 文件和文件夹,包含上传/下载、创建文件夹、复制/移动/删除、查看元数据、查询权限设置、评论/权限/订阅、标题、版本、飞书文档密级标签(secure labels)和本地文件导入。用户需要整理云盘目录、处理云空间资源 URL/token、判断链接类型/真实 token/标题,或导入 Word/Markdown/Excel/CSV/PPTX/.base 为 docx/sheet/bitable/slides 时使用;doubao.com 云空间 URL/token 也按资源路径和 token 路由,不回退 WebFetch。不负责:文档内容编辑(走 lark-doc)、表格/Base 表内数据操作(走 lark-sheets/lark-base)、知识空间节点/成员管理(走 lark-wiki)、原生 Markdown 文件读写/patch/diff(走 lark-markdown)。

lark-shared

open.feishu.cn

Use for lark-cli setup/auth tasks: auth login/status/logout, user vs bot identity, business-domain permissions (--domain, including all/docs/drive), missing scopes, revoking authorization, or handling notice JSON.

lark-im

open.feishu.cn

飞书即时通讯:收发消息和管理群聊。发送和回复消息、搜索聊天记录、管理群聊成员、上传下载图片和文件、管理表情回复、发送应用内/短信/电话加急、发送和处理交互卡片(Interactive Card)、监听卡片按钮回调(card.action.trigger)。当用户需要发消息、查看或搜索聊天记录、下载聊天中的文件、查看群成员、搜索群、创建群聊或话题群、管理标记数据、管理 Feed 置顶(添加/移除/查询置顶会话)、管理标签数据、处理卡片回调时使用。