Google Antigravity 2.0 & Slash Commands

Adarsh — Fri, 22 May 2026 13:33:27 +0000

TL;DR Summary:
Google has revolutionized AI-assisted development with Antigravity 2.0, an agent-first architecture replacing traditional code editors. Key updates include powerful slash commands for autonomous workflow management, the new Android CLI for direct IDE integration ("Vibe Coding"), Firebase Agent Skills for automated backend logic, and WebMCP/AppFunctions for seamless app-to-app data manipulation powered by Gemini.

Google pulled a massive surprise with Antigravity 2.0 by completely overhauling the app into an agent-first architecture. They stripped away the traditional code editor and replaced it with an advanced agent manager designed to run multiple AI agents across multiple projects concurrently.

Antigravity Slash Commands for Workflow Automation

Antigravity introduced powerful slash commands to assist and automate your developer workflow:

/goal – Instructs an AI agent to run autonomously until a specified complex task is complete, removing the need for intermediate human approval.
/schedule – Handles cron-based, recurring agent execution for automated maintenance.
/browser – Hands explicit browser navigation and automated UI testing control directly to the agent.
/grill-me – Forces the agent to aggressively cross-examine you with clarifying questions, ensuring complete alignment before writing a single line of code.

The Android CLI & "Vibe Coding"

Google just handed developers a massive productivity upgrade with the announcement of the Android CLI. This new Command Line Interface grants AI coding agents like Claude Code, Codex, or Antigravity direct access to your local development environment and Android Studio. The live demo showcased AI agents autonomously writing code, making edits, and executing tests directly on the Android emulator.

Furthermore, Google AI Studio is expanding into native prompt-based Android development—a paradigm shift we call Vibe Coding. Developers can design, prototype, and build completely functional native Android applications directly inside AI Studio and immediately export them directly to their physical devices.

Agent Skills for Firebase & Migration Assistant

Google also introduced Agent Skills for Firebase and a powerful new Migration Assistant. Agent Skills empower developers to write complex backend logic, cloud functions, and database rules for Firebase and Firestore directly inside the IDE via natural language. As a developer, this is a massive time-saver; you can hand off writing security rules to Antigravity, and the AI agent handles the rest securely.

Tackling cross-platform development headaches, Google's new Migration Assistant is specifically designed to transition legacy iOS, React Native, or web framework codebases over to modern, native Android seamlessly.

Re-Engineering the Web & Android 17 Intelligence Layer

Google is reshaping how agents interact with the web and operating systems:

WebMCP (Model Context Protocol): An open standard designed to let developers expose HTML forms and JavaScript functions directly to AI models, bridging the gap between web interfaces and LLMs.
Chrome DevTools for Agents: Allows AI agents to autonomously debug, profile, and resolve complex issues within web applications.
AppFunctions API: Turns installed Android apps into on-device Model Context Protocol (MCP) servers. This groundbreaking API allows Gemini to autonomously pull data from one application, manipulate it, and inject it into another with absolutely zero user intervention.

TIL: Building Agentic Workflows in Chrome with Gemini Skills

Adarsh — Sat, 16 May 2026 04:53:20 +0000

Evolution of the Prompt

Anthropic recently popularized “skills” in coding agents and cool tools like Claude Code and Gemini CLI. These allow agents to understand how to perform various tasks efficiently without needing to re-prompt every time based on the respective task. Skills fixed the traditional problem with LLM interactions, which was the constant need to re-prompt. Skills act as saved prompts that you can trigger instantly, which is a massive time saver if you manage multiple tasks simultaneously.

Before Skills was widely rolled out, I would manually save prompts in Google Keep or Google Docs and pin them to my browser for various prompts. For code, I stored context in .md files, which is exactly how skills operate today. Having these skills directly in Gemini in Chrome is a game-changer because the AI can see, reason, and act on all your open tabs at once. So it’s able to synthesize data from across your entire browser window and act simultaneously.

Safety and Privacy

If you're worried about the AI booking things or emailing people without your permission, don't be. Gemini requires a Human-in-the-Loop (HITL) confirmation before sending or submitting anything, and trust me, it's not easy to bypass. I've tried. And for those concerned about privacy in general, Google added a local 4GB model on your computer called Gemini Nano, which is able to perform many tasks locally without sending any of your data to the cloud.

Engineering Workflow Skills

1. Convert to Notes

Prompt: You are an expert student with superb comprehension skills. Your task is to summarize the active tab, providing high-impact bullet points for the [Main Ideas], [Steps/Process], and [Key Vocabulary]. If the text contains data, present a summary table of the most critical figures.

Why I love it: I use this to quickly synthesize long articles or documentation into short, actionable bullet points to write down or just summarize.

2. Live Mock Interviewer

Prompt: Act as a Senior Software Engineer at a FAANG company. I am looking at this LeetCode problem. Do not give me the answer. Instead, ask me clarifying questions about the constraints, challenge my initial assumptions, and push for details on time and space complexity. Give me a 'nudge' or hint only if I am completely stuck for more than two minutes.

Why I love it: It’s perfect for practicing technical interview patterns and learning new algorithms through live feedback and guidance.

3. Competitor Intelligence

Prompt: Analyze the landing page of the startup in this tab. Identify their unique value proposition, pricing structure, and target audience. Compare this to the industry leaders in this space and find three gaps in their product that my own project could solve.

Why I love it: When researching for my own apps or companies, I use this to break down a competitor’s strategy and find gaps in their product without reading their entire site.

4. Legal/Privacy Decoder

Prompt: Scan this Privacy Policy for critical risk signals. Specifically, audit the text for: 1. Data ownership (does the user retain rights?), 2. Third-party sharing (who gets my data?), and 3. Automatic renewals. Flag these as [Critical], [High Risk], or [Warning] and explain the plain-English implication of each.

Why I love it: Most people skip the fine print, but I use this to audit Terms of Service for red flags so I know my data stays safe.

Step-by-Step Guide to Creating Skills

Open the Gemini Sidebar: Click the Gemini icon in the top-right corner of your browser. Then, click the plus icon at the bottom left of the panel.
Access the Skills Manager: Click the Skills button. From there, select Manage Skills, which will open the settings in a new window.
Explore or Create: On the left sidebar, you’ll see Browse Skills and Your Skills. Use 'Browse Skills' to see Google’s premade options for inspiration. To make your own, click Your Skills.
Build Your Skill: Click the Add button. Fill in a name and your specific instructions.

Forem: Adarsh