Empower agents with professional capabilities in specific fields (such as full-stack development, complex logic planning, multimedia processing, etc.) through modular Skills definitions, allowing agents to solve complex problems systematically like human experts.
.
├── .claude-plugin/ # Claude plugin configuration files
├── skills/ # Antigravity Skills library
│ ├── skill-name/ # Individual skill directory
│ │ ├── SKILL.md # Core skill definition and Prompt (Required)
│ │ ├── scripts/ # Scripts relied upon by the skill (Optional)
│ │ ├── examples/ # Skill usage examples (Optional)
│ │ └── resources/ # Templates and resources relied upon by the skill (Optional)
├── docs/ # User manual and documentation guides
├── scripts/ # Maintenance scripts
├── skills_sources.json # Skills synchronization source config
├── skills_index.json # Skills metadata index
├── spec/ # Specification documents
├── template/ # New skill template
└── README.md
Antigravity Skills follow the universal SKILL.md format and can work seamlessly with any AI coding assistant that supports Agentic Skills:
| Tool Name (Agent) | Type | Compatibility | Project Path | Global Path |
|---|---|---|---|---|
| Antigravity | IDE | ✅ Full | .agent/skills/ |
~/.gemini/antigravity/skills/ |
| Claude Code | CLI | ✅ Full | .claude/skills/ |
~/.claude/skills/ |
| Gemini CLI | CLI | ✅ Full | .gemini/skills/ |
~/.gemini/skills/ |
| Codex | CLI | ✅ Full | .codex/skills/ |
~/.codex/skills/ |
| Cursor | IDE | ✅ Full | .cursor/skills/ |
~/.cursor/skills/ |
| GitHub Copilot | Extension | .github/skills/ |
~/.copilot/skills/ |
|
| OpenCode | CLI | ✅ Full | .opencode/skills/ |
~/.config/opencode/skills/ |
| Windsurf | IDE | ✅ Full | .windsurf/skills/ |
~/.codeium/windsurf/skills/ |
| Trae | IDE | ✅ Full | .trae/skills/ |
~/.trae/skills/ |
Tip
Most tools will automatically discover skills in .agent/skills/. For maximum compatibility, please clone/copy into this directory.
First, clone this repository locally (it is recommended to place it in a fixed location for global reference):
git clone https://github.com/guanyang/antigravity-skills.git ~/antigravity-skillsWe strongly recommend using Symbolic Links (Symlink) for installation, so that when you update this repository via git pull, all tools will automatically sync the latest features.
Enable skills only for the current project. Run in your project root:
mkdir -p .agent/skills
ln -s ~/antigravity-skills/skills/* .agent/skills/Enable skills by default in all projects. Run the corresponding command based on the tool; common examples:
| Tool Name | Global Installation Command (macOS/Linux) |
|---|---|
| General | mkdir -p ~/.agent/skills && ln -s ~/antigravity-skills/skills/* ~/.agent/skills/ |
| Claude Code | mkdir -p ~/.claude/skills && ln -s ~/antigravity-skills/skills/* ~/.claude/skills/ |
| Antigravity | mkdir -p ~/.gemini/antigravity/skills && ln -s ~/antigravity-skills/skills/* ~/.gemini/antigravity/skills/ |
| Gemini | mkdir -p ~/.gemini/skills && ln -s ~/antigravity-skills/skills/* ~/.gemini/skills/ |
| Codex | mkdir -p ~/.codex/skills && ln -s ~/antigravity-skills/skills/* ~/.codex/skills/ |
If you primarily use Claude Code, you can install with one click via the plugin marketplace (this method automatically handles skill loading):
# 1. Start Claude Code
# 2. Add the plugin marketplace
/plugin marketplace add guanyang/antigravity-skills
# 3. Install the plugin from the marketplace
/plugin install antigravity-skills@antigravity-skillsEnter @[skill-name] or /skill-name in the chat box to invoke them, for example:
/canvas-design Help me design a 16:9 blog cover about "Deep Learning"
- View Manual: For detailed usage, please refer to docs/Antigravity_Skills_Manual.en.md.
- Environment Dependencies: Some skills rely on Python environments; please ensure your system has necessary libraries installed (e.g.,
pdf2docx,pandas, etc.).
Many skills in this project originate from excellent open-source communities. To keep in sync with upstream repositories, you can update them in the following ways:
-
Configuration: The
skills_sources.jsonfile in the root directory is pre-configured with the upstream repositories for major skills and usually does not need manual adjustment. -
Run Sync: You can choose to sync all skills or just a specific one:
# Sync all configured sources ./scripts/sync_skills.sh # Sync only a specific source (e.g., anthropics-skills) ./scripts/sync_skills.sh anthropics-skills
The script will automatically pull the latest code and update the corresponding skill directories.
Note: The
ui-ux-pro-maxskill has a special directory structure and does not support automatic synchronization via script for now. Please use its official installation commanduipro init --ai antigravityto install or update.
These skills focus on visual expression, UI/UX design, and artistic creation.
@[algorithmic-art]: Create algorithmic and generative art using p5.js code.@[canvas-design]: Create posters and artworks (PNG/PDF output) based on design philosophies.@[json-canvas]: Create and edit JSON Canvas files (.canvas) with nodes, edges, and groups (commonly used in Obsidian).@[frontend-design]: Create high-quality, production-grade frontend interfaces and Web components.@[ui-ux-pro-max]: Professional UI/UX design intelligence, providing full design schemes for colors, fonts, layouts, etc.@[web-artifacts-builder]: Build complex, modern Web apps (based on React, Tailwind, Shadcn/ui).@[theme-factory]: Generate matching themes for documents, slides, HTML, etc.@[brand-guidelines]: Apply Anthropic's official brand design specifications (colors, typography, etc.).@[remotion]: Best practices for Remotion - Video creation in React.@[web-design-guidelines]: Review UI code for Web Interface Guidelines compliance (accessibility, UX, design audit).@[slack-gif-creator]: Create high-quality animated GIFs optimized specifically for Slack.
These skills cover the full lifecycle of coding, testing, debugging, and code review.
@[composition-patterns]: React composition patterns for building scalable, flexible component libraries.@[react-best-practices]: Vercel's official React and Next.js performance optimization guidelines.@[react-native-skills]: React Native and Expo best practices for performant mobile apps.@[supabase-postgres-best-practices]: Postgres performance optimization and best practices from Supabase.@[test-driven-development]: Test-Driven Development (TDD) - write tests before implementation code.@[systematic-debugging]: Systematic debugging for resolving bugs, test failures, or abnormal behaviors.@[webapp-testing]: Use Playwright for interactive testing and verification of local web applications.@[receiving-code-review]: Handle code review feedback using technical verification rather than blind modification.@[requesting-code-review]: Proactively initiate code reviews to verify code quality before merging or completion.@[finishing-a-development-branch]: Guide the finalization of a development branch (merges, PRs, cleanups, etc.).@[subagent-driven-development]: Coordinate multiple sub-agents to perform independent development tasks in parallel.
These skills are used for handling professional documents and office needs in various formats.
@[doc-coauthoring]: Guide users through collaborative writing of structured documents (proposals, tech specs, etc.).@[obsidian-markdown]: Create and edit Obsidian Flavored Markdown with wikilinks, embeds, callouts, and properties.@[obsidian-bases]: Create and edit Obsidian Bases (.base) files with views, filters, formulas, and summaries.@[obsidian-cli]: Interact with Obsidian vaults using the Obsidian CLI to read, create, search, and manage notes from the command line.@[defuddle]: Extract clean markdown content from web pages using Defuddle CLI, removing clutter and navigation.@[docx]: Create, edit, and analyze Word documents.@[xlsx]: Create, edit, and analyze Excel spreadsheets (supporting formulas and charts).@[pptx]: Create and modify PowerPoint presentations.@[pdf]: Process PDF documents, including extracting text/tables, merging/splitting, and filling forms.@[internal-comms]: Draft various corporate internal communication documents (weekly reports, announcements, FAQs, etc.).@[notebooklm]: Query Google NotebookLM notebooks for definitive, document-grounded answers.
These skills help optimize workflows, task planning, and execution efficiency.
@[brainstorming]: Brainstorm before starting any work to clarify requirements and design.@[writing-plans]: Write detailed execution plans (Specs) for complex multi-step tasks.@[planning-with-files]: A file-based planning system (Manus-style) suitable for complex tasks.@[executing-plans]: Execute existing implementation plans with checkpoints and review mechanisms.@[using-git-worktrees]: Create isolated Git worktrees for parallel development or task switching.@[verification-before-completion]: Run verification commands to ensure concrete evidence before declaring task completion.@[using-superpowers]: Guide users to discover and use these advanced skills.
These skills build the agent's mental models, memory systems, and context management capabilities.
@[bdi-mental-states]: Simulate Agent's Belief-Desire-Intention (BDI) models.@[memory-systems]: Build long-term memory and entity tracking systems based on knowledge graphs or vectors.@[context-fundamentals]: Understand and debug fundamental issues like context windows and attention mechanisms.@[context-optimization]: Optimize context efficiency to reduce Token costs via KV-cache or partitioning.@[context-compression]: Implement context compression and summarization to handle long window limits.@[context-degradation]: Diagnose and fix context degradation issues like "lost in the middle".@[filesystem-context]: Utilize the filesystem for dynamic context offloading and management.
These skills focus on architectural design, tool building, and quality assessment of AI systems.
@[project-development]: Full lifecycle design of LLM projects, including task-model matching and pipeline architecture.@[tool-design]: Design efficient and clear agent tool interfaces and MCP protocols.@[evaluation]: Establish multi-dimensional agent performance evaluation systems and quality gates.@[advanced-evaluation]: Implement advanced evaluation methods like LLM-as-a-Judge and pairwise comparison.
These skills allow me to extend my own capability boundaries.
@[mcp-builder]: Build MCP (Model Context Protocol) servers to connect external tools and data.@[skill-creator]: Create new skills or update existing ones to expand my knowledge base and workflows.@[writing-skills]: A subset of tools to assist in writing, editing, and verifying skill files.@[dispatching-parallel-agents]: Dispatch parallel tasks to multiple agents for processing.@[multi-agent-patterns]: Design advanced multi-agent collaboration patterns like Supervisor or Swarm.@[hosted-agents]: Build and deploy sandboxed, persistently running background agents.
This project integrates core ideas or skill implementations from the following excellent open-source projects. Respect to the original authors:
- Anthropic Skills: Official API usage paradigms and skill definition references provided by Anthropic.
- UI/UX Pro Max Skills: Top-tier UI/UX design intelligence, providing full design schemes for colors, layouts, etc.
- Superpowers: A toolkit and workflow inspiration aimed at giving LLMs "superpowers."
- Planning with Files: Implements a Manus-style file-based task planning system to enhance persistent memory for complex tasks.
- NotebookLM: Knowledge retrieval and Q&A skill implementation based on Google NotebookLM.
- Agent-Skills-for-Context-Engineering: In-depth Context Engineering skills covering compression, optimization, and degradation handling.
- Obsidian Skills: Professional Obsidian integration skills, including JSON Canvas and enhanced Markdown support.
- Remotion Skills: Official Remotion skills for AI agents to create videos programmatically.
- Vercel Agent Skills: Official Vercel skills for React best practices, composition patterns, and web design guidelines.
- Supabase Agent Skills: Official Supabase skills for Postgres performance optimization and best practices.
We take security seriously. Please refer to our Security Policy for information on supported versions and how to report vulnerabilities safely.
We welcome contributions! Please refer to our CONTRIBUTING.md for detailed guidelines on how to add new skills, improve documentation, and report issues.
This project is open-sourced under the MIT License.