Read this in other languages: English, Español
⚡ The ultimate companion CLI for the viral
antigravity-awesome-skillslibrary.
A complete management system to structure your projects (workspaces) and dynamically inject skills using the famous antigravity-awesome-skills repository. While the original repository provides specialized knowledge (+250 skills), this manager brings the orchestration needed for your AI assistants (Antigravity, Claude Code, Cursor) to load exclusively the necessary context into each project.
- Dynamic Location: Intelligent detection of the base project path, allowing you to invoke the script from any subdirectory within your environment.
- Smart Wizard: A step-by-step terminal interface to easily create workspaces and auto-enable recommended skills tailored to your tech stack.
- Environment Isolation: Each project (workspace) maintains its own
skill-config.jsonconfiguration file and a symbolic link (symlink) environment routing only to the designated skills. - Transparent Synchronization: Integrated system to safely sync (clone/update) the global folder from the official GitHub repository, including automatic backups.
- Auto-Repair: Detection and cleanup of orphaned or broken skills in case they are removed from the parent repository.
Get the wsm alias and initialize everything in one command:
curl -sSL https://raw.githubusercontent.com/amartelr/antigravity-workspace-manager/main/install.sh | bashRestart your terminal after installation to enable the wsm command.
If you prefer to do it manually:
# 1. Clone the repository
git clone https://github.com/amartelr/antigravity-workspace-manager.git
cd antigravity-workspace-manager
# 2. Initialize
wsm init💡 Portability Tip: You can move the folder wherever you prefer (e.g.,
~/MyProjects). The script will auto-detect its new location.
The highly recommended workflow is to use the interactive wizard:
wsm wizardThe completely redesigned wizard will guide you through 5 steps to define your environment:
- The Name and Description of the workspace.
- The Project Type (14 detailed options):
- From classics like API Backend, Web Frontend, Mobile App, to new flows like Microservices, AI/ML, Data Engineering, DevOps/Infra, Game Dev, SEO/Marketing, Security/Pentesting and Blockchain/Web3.
- The Primary Language (15 options):
- Python, JavaScript/TypeScript, Go, Dart/Flutter, Rust, Java/Kotlin, C#/.NET, C/C++, Swift/SwiftUI, Ruby, PHP, Elixir, Scala, Julia or Haskell.
- The Database (11 options):
- PostgreSQL, Supabase, MongoDB/NoSQL, MySQL, SQLite, Redis, Firebase, Neon Postgres, Google Sheets, Elasticsearch or DynamoDB.
- Intelligent Skill Recommendation (Multi-select):
- Based on your previous selections, the wizard will suggest highly relevant categories to inject into your project.
- Examples of dynamic suggestions:
- 🏗️ Architecture & Quality:
architecture,microservices-patterns,clean-code... - 🧪 Testing & Debugging:
tdd-workflow,playwright-skill... - 🚀 DevOps & Deploy:
docker-expert,github-actions-templates... - 🔒 Security:
api-security-best-practices... - 🤖 AI/ML:
prompt-engineering,rag-implementation... - 📈 SEO & Marketing:
seo-fundamentals,analytics-tracking... - ⚡ Automation:
workflow-automation,n8n-mcp-tools-expert...
- 🏗️ Architecture & Quality:
| Action | Command |
|---|---|
| Initialize Structure | wsm init |
| Assisted Setup | wsm wizard |
| Manual Creation | wsm create project-name |
| View Active Workspaces | wsm list |
| View Full Skills Catalog | wsm list-skills |
| View Project Skills | wsm list-skills project-name |
| Enable Skill | wsm enable project-name skill-name |
| Disable Skill | wsm disable project-name skill-name |
| Recommend Skills | wsm reco-skills project-name |
| View Skill Detail | wsm show skill-name [--lang es] |
| Sync and Repair Skills | wsm sync --auto-fix |
After executing the init command, the script will automatically deploy a robust hierarchy for your orchestration:
/your-manager-base-path/
├── workspace-manager.py ← CLI Environment
├── .agent/
│ ├── skills/ ← The entire skills repository
│ │ ├── public/ ← Official skills cloned from the public remote
│ │ ├── private/ ← Your private skills or enterprise guidelines
│ │ └── user/ ← Locally developed skills
│ └── skills_backup/ ← Periodic manager backups
├── workspaces/ ← Container directory for your working folders
│ ├── my-project/
│ │ ├── .agent/
│ │ │ └── skills ← Static link (symlink) to the main library
│ │ ├── skill-config.json ← Explicit declaration of your necessary dependencies
│ │ └── README.md ← Auto-generated foundational document
├── skill-config-templates/ ← Pre-packaged default templates and collections
If you prefer to bypass the guided interface (Wizard), you can speed up scaffolding by relying on bundles:
# Will collectively inject all skills related to the frontend area
wsm create my-webapp -t frontend-bundleExamples of default available templates:
- frontend-bundle: UI/UX design components, react/tailwind patterns, frontend testing.
- backend-bundle: clean code, api guidelines, and transactional patterns.
- mobile-bundle: flutter best practices, mobile security.
Once your workspace is created, a base README.md file will automatically populate within the local folder. This file includes an excerpt specifically designed to pass context to your respective AI Assistant:
Workspace: [project-name]
Collect the described skills logic reading from the local context path ./skill-config.json
Confirm the exact libraries you now have under context.
Consider making it a habit to refresh the packages comprising your skills catalog by running periodic syncs.
# Clone, verify version diffs, delete outdated ones, and update references in one go
wsm sync --auto-fixThanks to the script auto-detecting paths independently of where it's invoked, you can automate cronjobs by passing the absolute path directly (no need for the classic preceding cd). For instance, automatic updates every Sunday at dawn:
0 2 * * 0 wsm sync --auto-fix- Symlink Issues (Especially on Windows):
Console often requires extended privileges to handle deep directory mappings.
Resolve it by enabling the compatibility Developer Mode, and opening your command terminal with Administrator Privileges. WSL (Windows Subsystem for Linux) also prevents this issue 100%.
- Rejections Executing Listed Commands or Command not found:
Make sure the
wsmalias is configured. Run the installer again with./install.shand restart your terminal. Alternatively, you can executepython3 workspace-manager.py ...directly.
Develop faster, and endow your AI with the exact universal context. 🚀
