🚀 Antigravity Supercharged Project Engine

Welcome to the ultimate starter kit for AI-assisted software development using the Antigravity architecture.

This repository bundles a highly customized set of AI Agent Skills and Model Context Protocol (MCP) server configurations designed to drastically improve large-scale project planning, architecture design, and codebase exploration.

🌟 What's inside?

1. Integrated jCodeMunch (O(1) Codebase Retrieval) Instead of the AI agents reading thousands of lines of files blindly and chewing through tokens, jcodemunch uses tree-sitter AST indexation. The agents query exactly the symbols (functions/classes) they need, cutting token costs by up to 99%.

2. Integrated NotebookLM Deep Research Agents can query your private Google NotebookLM notebooks to synthesize vast amounts of architecture documents, PRDs, and domain knowledge before writing a single line of code.

3. Curated OpenSpec Workflows Included are optimized openspec-explore and openspec-apply-change skills that natively trigger the jcodemunch indexer before proceeding.

4. Massive Suite of Specialized Agents & Workflows Included in the .agent folder are dozens of specialized skills:

   • Architecture & Design: expert-product-manager, architecture-c4-model, design-system-architect, Frontend/Backend Architects.
   • Quality & Testing: elite-qa-architect, process-systematic-debugging, webapp-testing.
   • Code Engine: openspec (Explore, Propose, Apply) for spec-driven development.
   • Research: notebooklm-researcher for massive context integration.
   • ...and many more tailored for complete software lifecycle management.

🛠️ Installation

To turbocharge your next Antigravity project, navigate to your target project directory and simply run:

npx antigravity-awesome-skills

What does the NPX script do?

1. Checks for and installs uv (the lightning-fast Python package installer).
2. Installs the notebooklm-mcp-cli globally.
3. Copies the AGENTS.md and the entire .agent/ suite into your current working directory.
4. Bootstraps CONTINUITY.md so the OpenSpec 5-Gate process begins working out of the box.
5. Safely injects the MCP server definitions (for notebooklm-mcp and jcodemunch) into your global ~/.gemini/antigravity/mcp_config.json.

⚠️ Post-Installation Setup

After running the script, you must authenticate the NotebookLM service:

nlm login

This will open a Chrome browser profile. Log in to your Google account that has access to your NotebookLM instances. Close the browser, restart your Antigravity session, and you are ready to build!

🔄 The 5-Gate OpenSpec Workflow

This starter kit enforces a disciplined, 5-Gate Spec-Driven development cycle outlined in your AGENTS.md:

1. GATE 1 — UNDERSTAND (Continuity + Memory + Research) The agent checks CONTINUITY.md, boots a session via antigravity-memory, and uses jcodemunch to index and search your codebase (saving up to 99% tokens).
2. GATE 2 — ANALYZE (Risks & Impact) The agent structurally evaluates which architectural layers are affected and scores the risk before writing code.
3. GATE 3 — PLAN (Spec-Driven via OpenSpec) Using /opsx:propose, the agent generates rigid change proposals, design docs, and atomic tasks to gain your approval.
4. GATE 4 — IMPLEMENT (RARV Loop) Using /opsx:apply, the agent surgically applies code using jcodemunch (O(1) code retrieval) in small, testable chunks.
5. GATE 5 — VERIFY (Testing & Memory) Tests are run, knowledge is stored permanently in the agent memory (memory_save_note), and the cycle completes.

🧠 The Skill Creator Workflow

Need a highly specialized agent for a niche framework? Use the built-in Skill Creator Workflow:

1. Just ask the agent: "Turn this workflow into a skill" or "Create a new skill for Next.js caching".
2. The elite-skill-architect is invoked automatically.
3. It drafts a structured SKILL.md matching our strict YAML-frontmatter standard.
4. It natively saves the file into your .agent/skills/ directory and auto-registers the triggers globally inside AGENTS.md.

🙏 Acknowledgments

This starter kit stands on the shoulders of giants. Special thanks to the creators of the core MCP servers that make this workflow possible:

• jacob-bd/notebooklm-mcp-cli: For reverse-engineering the CDP flow and bringing Google's NotebookLM natively into the Model Context Protocol.
• jgravelle/jcodemunch-mcp: For creating the ingenious O(1) AST-based codebase indexer that saves millions of tokens.

Build faster. Build smarter. Defy gravity.