Google AI Agent Skills

Agent Skills for Google AI services. These skills teach AI agents (Claude Code, Gemini CLI, Cursor, OpenCode, etc.) how to use Google's Gemini API effectively.

Built following the Agent Skills open standard.

Available Skills

Skill	Description
gemini-text	Text generation with Gemini models
gemini-image	Image generation with Gemini/Imagen
gemini-tts	Text-to-speech synthesis
gemini-batch	Batch processing for large workloads
gemini-embeddings	Text embeddings for semantic search
gemini-files	File upload to Gemini Files API

Setup

1. Get API Key

Get your API key from Google AI Studio.

2. Set Environment Variable

export GEMINI_API_KEY="your-api-key"

Or create a .env file:

GEMINI_API_KEY=your-api-key

3. Install SDK

npm install @google/genai@latest dotenv@latest

Usage with AI Agents

1. Skills CLI (npx skills)

The easiest way to use these skills is via the Skills CLI. It automatically installs skills into your .agent/skills directory for discovery by agents like Claude Code or Gemini CLI.

# Install a specific skill
npx skills add https://github.com/akrindev/google-studio-skills/tree/main/skills/gemini-text

# Install the entire collection
npx skills add akrindev/google-studio-skills

2. ClawdHub & Moltbot

Moltbot (formerly Clawdbot) is a local-first AI assistant. You can manage and run these skills using the clawdhub CLI.

Installation & Setup:

# Install Moltbot and ClawdHub CLI
npm install -g moltbot clawdhub

# Initialize Moltbot
moltbot onboard

Manage Skills:

# Search for Gemini skills on ClawdHub
clawdhub search gemini

# Install a skill to your workspace (~/molt/skills/)
clawdhub install gemini-text --workdir ~/molt

Running Skills: Once installed, interact with Moltbot via the terminal or connected chat apps (WhatsApp/Telegram). Simply ask:

"Molt, use the gemini-text skill to summarize this document."

3. Manual Installation (Claude Code / Gemini CLI / OpenCode)

These agents automatically discover skills in skills/ directories. You can manually point your agent to this repository:

# Clone the repository
git clone https://github.com/akrindev/google-studio-skills.git

# Navigate to your project and the agent will discover skills in the adjacent folder
# or add the path to your agent's configuration

4. Discovery on Skills.sh

Browse the full directory of available skills and see community rankings at skills.sh.

Manual Usage

Each skill contains standalone JavaScript scripts that can be run directly:

# Text generation
node skills/gemini-text/scripts/generate.js "Explain quantum computing"

# Image generation
node skills/gemini-image/scripts/generate_image.js "A sunset over mountains"

# Text-to-speech
node skills/gemini-tts/scripts/tts.js "Hello world" --voice Kore

# Embeddings
node skills/gemini-embeddings/scripts/embed.js "semantic search query"

Default Models (January 2026)

Capability	Default Model	Notes
Text	gemini-3-flash-preview	Fast, agentic, multimodal
Image	gemini-3.1-flash-image-preview	Nano Banana 2 default (up to 4K)
TTS	gemini-2.5-flash-preview-tts	Multiple voices
Embeddings	gemini-embedding-001	3072 dimensions

Skill Structure

Each skill follows the Agent Skills specification:

skills/gemini-text/
├── SKILL.md           # Instructions for AI agents
├── scripts/           # Executable JavaScript scripts
│   └── generate.js
└── references/        # Additional documentation
    └── models.md

Development

# Create virtual environment
python3 -m venv .venv
source .venv/bin/activate

# Install dependencies for a skill script
cd skills/gemini-text/scripts && npm install

# Format code
ruff format skills/

# Lint code
ruff check skills/

License

MIT