antigravity-os

Antigravity-OS 🚀

The Production Runtime, Dual-Memory Architecture & 503-Skill Playbook for Google Antigravity & Universal AI Coding Agents.
Turn your AI coding assistant into a disciplined, autonomous software engineering team in under 2 minutes.

---

---

⚡ The 2-Minute Quickstart

Install the complete runtime, all 503+ skills and 71 specialist agents, global MCP servers, and the 4-Pillar Synergy rules with a single command:

🪟 Windows (PowerShell)

irm https://raw.githubusercontent.com/ratulbd/antigravity-os/main/scripts/install.ps1 | iex

(Or clone this repo and run ./scripts/install.ps1 in PowerShell as Administrator or regular user)

🍎 macOS (Terminal)

curl -fsSL https://raw.githubusercontent.com/ratulbd/antigravity-os/main/scripts/install.sh | bash

🐧 Linux (Bash)

curl -fsSL https://raw.githubusercontent.com/ratulbd/antigravity-os/main/scripts/install.sh | bash

---

📊 Graphical Architecture

flowchart TD
    subgraph P1["Pillar 1: Prescriptive Specs"]
        Specs["/docs/01_PRD.md → 06_Plan.md<br><b>WHAT & WHY</b>: Design tokens, CSS contracts, API schemas"]
    end

    subgraph P2["Pillar 2: Structural AST Graph"]
        AST["codebase-memory-mcp<br><b>HOW code is wired</b>: 158-Language Tree-sitter + SQLite AST"]
    end

    subgraph P3["Pillar 3: Episodic Session Memory"]
        Memory[".trail/learnings/ + ai-memory<br><b>WHAT was tried & ruled out</b>: Prevents repeating failed bugs"]
    end

    subgraph P4["Pillar 4: Specialist Execution & Runtime Gates"]
        Runtime["Gravitas + Superpowers + Ponytail<br><b>HOW code is mutated</b>: Write-scope locks & fresh validator gates"]
    end

    P1 --> P2
    P2 --> P3
    P3 --> P4

---

💡 Why Antigravity-OS?

Modern AI agents (even frontier models like Gemini 3.8 and Claude 3.7) suffer from three fatal production issues:

1. Context Amnesia & Hallucination: Agents forget past debugging dead-ends and repeatedly re-attempt failed fixes.
2. Unconstrained Scope Creep: An agent instructed to tweak a button might edit 8 files, refactor configs, and bump package versions.
3. Premature "Victory" Declarations: Saying "I have fixed the bug!" without actually executing a fresh test runner after the latest file mutation.

Antigravity-OS fixes this mechanically by pairing:

• The 4-Pillar Synergy Protocol: Strict non-overlapping layers for specs, AST graphs, episodic memory, and execution.
• The Gravitas Runtime (6 Invariants): Hard write-scope locks (allowed_write_scope) and physical test-evidence gates.
• 182 Battle-Tested Skills: Production-grade capabilities from Anthropic, Supabase, Vercel, Expo, Superpowers, and more.

---

📈 Empirical Proofs: Measured Token Savings

Workflow Scenario	Raw LLM Baseline	Antigravity-OS Runtime	Verified Reduction
Symbol Call-Chain Discovery	~32,000 tokens (Reading 5 full files)	~350 tokens (Tree-sitter AST Graph)	▼ 98.9%
Investigating Known Bug	~28,500 tokens (3–4 trial-and-error loops)	~2,400 tokens (.trail + Gravitas gate)	▼ 91.5%
New Feature Implementation	~24,000 tokens (Unconstrained scans)	~4,100 tokens (Pillar 1 /docs spec)	▼ 82.9%
Boilerplate Minimization	~1,200 LOC output (Over-engineered)	~560 LOC output (Ponytail YAGNI filter)	▼ 53.3% LOC▼ 45% Tokens

---

🗂️ The 23 Curated Repositories (503+ Skills & 71 Specialist Agents)

Antigravity-OS automatically aggregates, normalizes, and stages skills from the most respected agent repositories:

#	Source Repository	Skills Included	Purpose & Domain
1	pbakaus/impeccable	impeccable	24 UI/UX design commands (craft, audit, polish, harden) and anti-slop rules.
2	nextlevelbuilder/ui-ux-pro-max-skill	ui-ux-pro-max, design-system, ui-styling, design, brand, banner-design, slides	Design tokens, color palettes, typography, and HTML/Chart.js presentation slide decks.
3	Nutlope/hallmark	hallmark	Anti-template visual design, intentional layout variety, and non-generic web UI.
4	dietrichgebert/ponytail	ponytail, ponytail-audit, ponytail-debt, ponytail-gain, ponytail-help, ponytail-review	"Lazy Senior Dev" mode: YAGNI minimal code (▼45% output tokens, ▼53% LOC), stdlib first.
5	cathrynlavery/diagram-design	diagram-design	Editorial HTML & SVG system diagrams and interactive flowcharts.
6	blader/humanizer	humanizer	Removes AI writing tropes, filler, and inflated phrasing from UI copy and documentation.
7	d4vinci/Scrapling	scrapling	Adaptive Python web scraping, stealth browser automation, and concurrent spiders.
8	sgunin/ReverseEngineerAnything	reverse-engineer-anything	Reverse-engineering binaries, protocols, JavaScript bundles, and runtime behavior.
9	tonydzi/second-brain-starter-kit	last30days	Nightly signal synthesis of what changed on any technical topic in the last 30 days.
10	Onyx-Systems/onyx-chat-template	onyx-chat-template	Express.js + nlp.js + TypeORM + OpenAI hybrid conversational backend blueprint.
11	obra/superpowers	brainstorming, writing-plans, test-driven-development, systematic-debugging, verification-before-completion, subagent-driven-development, etc. (15 total)	Strict TDD, root-cause debugging, parallel subagent dispatch, and verification workflows.
12	proyecto26/system-design-skills	system-design, requirements-scoping, back-of-the-envelope, api-design, data-storage, caching, blob-store, sharded-counters, etc. (22 total)	Distributed system design, QPS/storage estimation, caching, queues, and resilience.
13	codewithowais/senior-backend-dev	senior-backend-dev	Production backend engineering (Node/TS, Python, Go, Rust), Clean Architecture.
14	supabase/agent-skills	supabase, supabase-postgres-best-practices	PostgreSQL indexing, RLS, query performance optimization, and lock-free migrations.
15	vercel-labs/agent-skills	react-best-practices, composition-patterns, web-design-guidelines, writing-guidelines, vercel-optimize, deploy-to-vercel, etc. (9 total)	React/Next.js waterfall elimination, bundle optimization, and React 19 patterns.
16	expo/skills	expo-native-ui, expo-router, expo-data-fetching, expo-animation, expo-design-system, eas-workflows, eas-update, etc. (23 total)	Complete React Native and Expo engineering, routing, EAS deployments, and native UI controls.
17	anthropics/skills	frontend-design, mcp-builder, webapp-testing, web-artifacts-builder, skill-creator, pdf, docx, xlsx, pptx, canvas-design, etc. (14 total)	Presentation decks (pptx), brochures/posters, Playwright testing, doc coauthoring, MCP building.
18	skillcreatorai/Ai-Agent-Skills	database-design, backend-development, llm-application-dev, verification-agent, code-documentation, changelog-generator, etc. (10 total)	RAG patterns, LLM application architecture, changelogs, and technical documentation.
19	ComposioHQ/awesome-claude-skills	invoice-organizer, competitive-ads-extractor, lead-research-assistant, meeting-insights-analyzer, file-organizer, etc. (7 total)	Document automation, competitor intelligence scraping, and meeting transcript analysis.
20	coreyhaines31/marketingskills	cro, sales-enablement, copywriting, copy-editing, product-marketing, lead-magnets, ad-creative, pricing, seo-audit, etc. (52 total)	Full-funnel growth engineering, conversion rate optimization, sales decks, and SEO audits.
21	udaysharmadev/Gravitas	gravitas, gravitas-highstakes, gravitas-perf, gravitas-python, gravitas-security, gravitas-typescript (6 total)	Engineering reliability runtime: 6 core invariants, allowed_write_scope lock, and validator gates.

---

🛡️ The Gravitas Reliability Contract (6 Invariants)

Antigravity-OS incorporates the Gravitas Runtime Engine to govern how code is inspected and modified:

1. Ground before mutating: Read target files, tests, and callers before editing. Never assume file state.
2. Stay in declared scope: Writes outside allowed_write_scope are rejected at the tool-call level.
3. Completion needs evidence: A task is only complete when a validator (test/build/lint) passes after the latest file edit.
4. Never exact-retry a failure: Byte-identical tool retries on failures are blocked, forcing a root-cause pivot.
5. Satisfy the contract: Acceptance criteria in contract.json mechanically gate completion.
6. Say what you can't verify: Downgrade confidence on unverified assumptions rather than fabricating certainty.

---

🔍 Health Check & Diagnostic Tool

Verify that your system is correctly configured at any time:

On Windows (PowerShell):

./scripts/verify.ps1

On macOS / Linux (Bash):

./scripts/verify.sh

Expected Diagnostic Output:

================================================================================
   ANTIGRAVITY-OS SYSTEM DIAGNOSTIC & VERIFICATION
================================================================================
1. SKILL INVENTORY AUDIT:
  [✓] Gemini Skills Path (~/.gemini/config/skills): 602 skills detected (Target: 503+)
  [✓] Universal Agent Skills Path (~/.agents/skills): 601 skills detected (Target: 503+)
  [✓] Critical Skill [gravitas]: Installed and resolvable
  [✓] Critical Skill [using-superpowers]: Installed and resolvable
  [✓] Critical Skill [ponytail]: Installed and resolvable
  [✓] Critical Skill [continuous-learning-v2]: Installed and resolvable
  [✓] Critical Skill [agent-security-hardening]: Installed and resolvable
  [✓] Critical Skill [design-system-audit]: Installed and resolvable

2. SPECIALIST AGENT PERSONAS AUDIT:
  [✓] Specialist Agent Personas (~/.gemini/config/agents): 71 personas detected

3. MCP SERVERS CONFIGURATION:
  [✓] Context7 MCP Server: Configured with @upstash/context7-mcp
  [✓] Playwright MCP Server: Configured with @playwright/mcp

4. MANDATORY ATTRIBUTION & DUAL-MEMORY RULES:
  [✓] Rule .gemini/GEMINI.md: Enforcing 4-Pillars & Token Ledger
  [✓] Rule .gemini/AGENTS.md: Enforcing 4-Pillars & Token Ledger
  [✓] Rule .agents/AGENTS.md: Enforcing 4-Pillars & Token Ledger
  [✓] Rule .claude/CLAUDE.md: Enforcing 4-Pillars & Token Ledger

Diagnostic Summary: All checks passed.
================================================================================

---

🌐 Universal IDE & Model Interoperability

Antigravity-OS installs into universal skill discovery roots, making your 503+ skills, 71 specialist agents, and dual-memory rules instantly available across:

Platform / Assistant	Supported Paths	Status
Google Antigravity	%USERPROFILE%\.gemini\config\skills\	🟢 Native
Gemini CLI	~/.gemini/config/skills/	🟢 Native
Claude Code	~/.claude/CLAUDE.md + ~/.agents/skills/	🟢 Native
Cursor IDE	.cursorrules + ~/.agents/skills/	🟢 Native
Windsurf IDE	.windsurfrules + ~/.agents/skills/	🟢 Native
Cline / Roo Code	.clinerules + ~/.agents/skills/	🟢 Native
OpenCode	~/.agents/skills/	🟢 Native

---

🤝 Contributing & Skill Attribution

Antigravity-OS stands on the shoulders of the open-source community. All skills maintain their upstream licenses and attributions. If you would like to contribute a new skill repository or improve runtime hooks:

1. Fork the repository.
2. Create your feature branch (git checkout -b feature/awesome-skill).
3. Commit your changes and open a Pull Request.

---

📄 License

Distributed under the MIT License. See LICENSE for full details.