A curated collection of Agent Skills for Google Antigravity, ported from the Claude Code Agents repository.

This vault transforms the extensive Claude Code ecosystem into Antigravity Skills, providing your agent with repeatable workflows, domain expertise, and specialized tools.

This repository contains 300+ specialized skills across software development, operations, security, and business domains. Each skill is a directory-based package that teaches Antigravity's agent how to perform specific tasks.

The skills are derived from three types of Claude Code components, all unified into the Antigravity Skill format:

1. Domain Skills (e.g., k8s-manifest-generator, async-python-patterns): Specialized knowledge packages.
2. Specialist Agents (e.g., backend-architect, security-auditor): Persona-based instruction sets for complex reasoning.
3. Commands & Workflows (e.g., full-stack-orchestration-full-stack-feature, conductor-implement): Structured, multi-step procedures.

Skills are flattened in the skills/ directory, but cover these broad categories:

• Python: python-pro, fastapi-pro, async-python-patterns, uv-package-manager
• JavaScript/TypeScript: typescript-pro, react-modernization, nextjs-app-router-patterns
• Systems: rust-pro, golang-pro, memory-safety-patterns
• Mobile: frontend-mobile-development-component-scaffold, react-native-architecture

• Kubernetes: kubernetes-architect, helm-chart-scaffolding, gitops-workflow
• Cloud: cloud-architect, terraform-module-library, cost-optimization
• CI/CD: cicd-automation-workflow-automate, github-actions-templates, gitlab-ci-patterns

• Security: security-auditor, sast-configuration, security-scanning-security-hardening
• Code Quality: code-review-ai-ai-review, code-refactoring-refactor-clean, codebase-cleanup-tech-debt
• Testing: unit-testing-test-generate, tdd-workflows-tdd-cycle, e2e-testing-patterns

• Conductor: conductor-implement, context-driven-development (Context-Driven Development)
• Architecture: c4-architecture-c4-architecture, microservices-patterns, api-design-principles
• Orchestration: full-stack-orchestration-full-stack-feature, incident-response-incident-response

• Data Engineering: data-engineer, spark-optimization, dbt-transformation-patterns
• AI/ML: ml-pipeline-workflow, prompt-engineering-patterns, rag-implementation

Antigravity loads metadata (name + description) from every installed skill at session start. More skills means more token usage and a higher chance of irrelevant auto-activation. Prefer targeted installs via search, tags, or bundles. install --all is advanced and not recommended for most projects.

This repo ships a generated catalog for discovery:

• CATALOG.md (human-readable index)
• catalog.json (machine-readable index used by the CLI)
• bundles.json (curated bundles)
• aliases.json (short names that map to long skill IDs)

Regenerate after adding/editing skills:

npm run build:catalog

When a conversation starts, Antigravity loads the metadata (name & description) from all skills. ANTIGRAVITY automatically activates a skill when your request matches its description.

Examples:

• "Help me design a REST API for a user service" → Activates api-design-principles and backend-architect.
• "Scaffold a new FastAPI project" → Activates python-development-python-scaffold.
• "Review this PR for security issues" → Activates security-scanning-security-hardening or security-auditor.
• "Start a new feature track for login" → Activates conductor-new-track.

You can install skills in two scopes:

• Workspace scope (project-specific): <workspace-root>/.agent/skills/
• Global scope (available in all projects): ~/.gemini/antigravity/skills/

You can easily install skills directly from the repository without cloning it manually.

1. Search skills (recommended first):

npx @rmyndharis/antigravity-skills search <query>
# Example:
npx @rmyndharis/antigravity-skills search kubernetes

2. List available skills:

npx @rmyndharis/antigravity-skills list

3. Install a specific skill to your current project:

npx @rmyndharis/antigravity-skills install <skill-name>
# Example:
npx @rmyndharis/antigravity-skills install bash-pro

4. Install by tag or bundle (targeted sets):

# By tag
npx @rmyndharis/antigravity-skills install --tag kubernetes

# By bundle
npx @rmyndharis/antigravity-skills install --bundle core-dev

Available bundles: core-dev, security-core, k8s-core, data-core, ops-core.

5. Install a skill globally:

npx @rmyndharis/antigravity-skills install <skill-name> --global
# Example:
npx @rmyndharis/antigravity-skills install bash-pro --global

6. Check installed skills:

# List local specific skills
npx @rmyndharis/antigravity-skills installed

# List globally installed skills
npx @rmyndharis/antigravity-skills installed --global

7. Update installed skills:

# Update a specific skill
npx @rmyndharis/antigravity-skills update <skill-name>

# Update ALL installed skills
npx @rmyndharis/antigravity-skills update

# Update global skills
npx @rmyndharis/antigravity-skills update --global

8. Doctor / stats:

npx @rmyndharis/antigravity-skills doctor
npx @rmyndharis/antigravity-skills stats

9. Install ALL skills (advanced, not recommended):

# To your current workspace
npx @rmyndharis/antigravity-skills install --all

# Globally
npx @rmyndharis/antigravity-skills install --all --global

Note: Installing all skills increases token usage and can trigger unrelated skills. Prefer targeted installs.

Aliases are supported via aliases.json (for long skill names).

Example:

npx @rmyndharis/antigravity-skills install full-stack-feature

If you prefer to clone the repository:

Option A — Install to a workspace

mkdir -p .agent/skills
cp -R /path/to/antigravity-skills/skills/<skill-name> .agent/skills/

Option B — Install globally

mkdir -p ~/.gemini/antigravity/skills
cp -R /path/to/antigravity-skills/skills/<skill-name> ~/.gemini/antigravity/skills/

Note: After copying skills, restart your agent session so Antigravity re-detects them.

1. Create a folder: skills/<skill-name>/
2. Add SKILL.md (required)
3. (Optional) Add helpers: scripts/, references/, assets/

• Keep SKILL.md concise; move long examples to resources/ or examples/.
• Use narrow descriptions (for example: "Use when you need to ...") and include "Do not use" to reduce over-activation.
• Add a "Safety" section for skills that propose terminal or infrastructure changes.
• Regenerate catalog files with npm run build:catalog, then run the local gate: npm test, STRICT=1 npm run validate:skills, and npm run check:catalog (CI runs all three).
• CI runs strict validation using validation-baseline.json for existing gaps; new skills must include the required sections, and any backticked resources//references//assets//scripts/ file referenced in SKILL.md must exist.
• The baseline is intentionally empty. Refreshing it (node scripts/validate-skills.js --write-baseline --yes) grandfathers — and thus silences — current soft violations, so avoid it unless you truly need to; without --yes the command only previews what it would grandfather.

---
name: <skill-name>
description: <one sentence describing when to use this skill>
---

# <Skill Title>

## Use this skill when
- ...

## Do not use this skill when
- ...

## Instructions
1. ...
2. ...

See SECURITY.md for safety expectations when writing skills that touch terminals or infrastructure.

MIT License. See LICENSE file for details.

Original content © Claude Code Agents. Ported to Antigravity Skills.