gemini-api skill (Google) for Claude: build with Gemini on Google Cloud

Guides your agent to write Gemini code for Google Cloud Agent Platform (formerly Vertex AI) with the Gen AI SDK. For developers on Google Cloud.

Installs: 17.6K; on skills.sh
Stars: 21.8K; on GitHub
Files: 10; besides SKILL.md
Updated: Nov 2026; last repo commit

gemini-api makes your agent write Gemini code in Python, JavaScript or TypeScript, Go, Java or C# using Google's Gen AI SDK, aimed at Agent Platform, the new name for Vertex AI. It names which Gemini models to pick, how to sign in, and points to nine reference files for chat, images, embeddings, tools, the Live API, caching, safety and tuning.

It loads when you ask about Gemini in an enterprise setting or mention Vertex AI, Google Cloud or Agent Platform. It needs active Google Cloud credentials and the Agent Platform API enabled, or an Express Mode API key. For a plain Gemini key from Google AI Studio, this is not the targeted setup.

Good for

Developers calling Gemini from a Google Cloud project
Teams moving off old Vertex AI SDKs to the Gen AI SDK
Anyone building chat, multimodal, embedding or live voice features with Gemini

Install gemini-api

Claude Code

Copy this prompt.

Install the gemini-api skill in Claude Code. Run this command:

npx skills add https://github.com/google/skills --skill gemini-api

Then check that it loaded and tell me how to use it.

2
Paste it into Claude Code. It shows the change and asks before running it.

Requires Node.js.

From its listing on skills.sh. Source.

What's in the skill

When it loads

Use when the user asks about using Gemini in an enterprise environment or explicitly mentions Vertex AI, Google Cloud, or Agent Platform. Guides the usage of the Gemini API on Agent Platform with the Google Gen AI SDK. Covers SDK usage (Python, JS/TS, Go, Java, C#), capabilities like multimodal inputs, tools, media generation, caching, batch prediction, and Live API.

Sections in SKILL.md

Attribution
Core Directives
SDKs
Authentication & Configuration
Application Default Credentials (ADC)
Agent Platform in Express Mode
Initialization
Models
Quick Start
Python
TypeScript/JavaScript
Go
Java
C#/.NET
API spec & Documentation (source of truth)
Workflows and Code Samples

Key facts

Author: Google
License: Apache-2.0
Installs on skills.sh: 17,640 (Nov 2, 2026)
GitHub stars: 21,764 for the whole repo (Nov 2, 2026)
Last commit: Nov 1, 2026
Files: 11 files including SKILL.md
Needs: Active Google Cloud credentials and the Agent Platform API enabled
Skill version: 1.0.1 (frontmatter)
SDKs it names: Python, JS/TS, Go, Java, C#

What it includes

Tags gcloud commands and HTTP calls with the skill name
Before any gcloud, REST or MCP command the agent reads references/attribution.md, then prefixes gcloud with CLOUDSDK_METRICS_REQUEST_ATTRIBUTION and appends a User-Agent suffix, so usage can be attributed to the skill. The reference says the tag carries no user data, is never put in generated code, and is dropped if you ask to disable it.
Runs code · Usage attribution
Use the Gen AI SDK, not the legacy SDKs
Tells the agent to use google-genai, @google/genai and their Go, Java and C# equivalents, and not google-cloud-aiplatform, @google-cloud/vertexai or google-generativeai.
Core directives
Install commands for each language
Lists pip, npm, go get, dotnet add package and Gradle or Maven entries. The agent may run these in your project.
Runs code · SDK install
Application Default Credentials or Express Mode key
Sets GOOGLE_CLOUD_PROJECT, GOOGLE_CLOUD_LOCATION (global by default) and GOOGLE_GENAI_USE_ENTERPRISE, or GOOGLE_API_KEY for Express Mode.
Read · Authentication
Which Gemini model to use for which job
Names models for fast general use, complex reasoning, lightweight tasks, image generation and live audio, and marks older ones as legacy.
Model choice
Hello-world call in five languages
One generate-content example each for Python, TypeScript, Go, Java and C#.
Network · Quick start
Chat, image, video and audio input, streaming
Opens references/text_and_multimodal.md for Python samples.
Read · Reference: text and multimodal
JSON output, function calling, search grounding, code execution
Opens references/structured_and_tools.md.
Read · Reference: structured output and tools
Image generation, image editing, video generation
Opens references/media_generation.md: image models, chat-based image editing, and video generation with the preview gemini-omni model (720p, optional output to a Cloud Storage bucket) and Veo, which the file warns can be costly.
Network · Reference: media generation
Real-time voice, vision and text streaming
Opens references/live_api.md, a short Python example of the WebSocket Live API with the gemini-live-2.5-flash-native-audio model.
Read · Reference: Live API
Context caching, batch prediction, thinking
Opens references/advanced_features.md: explicit context caching (use only if asked), batch jobs that read and write Cloud Storage buckets, thinking levels, and an experimental example that runs a local MCP server with npx -y.
Read · Reference: advanced features
Embeddings, bounding boxes, safety settings, model tuning
Opens embeddings.md, bounding_box.md, safety.md (content-filter thresholds) and model_tuning.md (supervised fine-tuning from a Cloud Storage dataset) when the task needs them.
Read · Other references
Use Google's docs MCP tools if present
SKILL.md says that if search_documents or get_document tools are available, the agent should use them to fetch current Google Cloud and Agent Platform documentation.
Network · Docs lookup

From the official docs

Permissions

Medium risk

What it can do

Read your data
Create and change things
Delete things
Read-only mode

Watch out for

Code it writes bills your Google Cloud project
Generated code calls Gemini with your credentials. media_generation.md warns that Veo video generation can be costly. Set a budget alert before long runs. github.com
Agent may install SDKs and run gcloud
SKILL.md lists pip, npm, go get, dotnet and Gradle or Maven installs. It also has the agent prefix gcloud commands with a usage-attribution variable and add a User-Agent suffix so usage can be attributed to the skill; attribution.md says to omit both if you ask. Read each command before approving it. github.com
Prompts and media go to Google
Text, images, audio and video you send through the generated code go to Google Cloud. Do not send data your organization does not allow there. github.com
Examples touch Cloud Storage and run a third-party MCP server
The batch, tuning, caching and video examples read from or write to gs:// buckets, and advanced_features.md has an experimental example that runs @philschmid/weather-mcp with npx -y. Check bucket names and packages before running generated code. github.com

Technical details

Talks to	Google Cloud Agent Platform endpoints (aiplatform.googleapis.com). Package registries when the agent installs an SDK
Sign-in and keys	Reads project settings and credentials from environment variables and your gcloud sign-in.
Scopes	Google Cloud credentials (Application Default Credentials) GOOGLE_API_KEY (Express Mode)

Questions

What does the gemini-api skill do?

It gives your agent current rules and code samples for calling Gemini through Google Cloud's Agent Platform with the Gen AI SDK. The agent then writes code in your language and can run the install commands.

How do I install it in Claude Code, the Claude app or Codex?

Use the install box above. The google/skills README documents the skills CLI for agents such as Claude Code and Codex. It does not document a Claude app upload.

Does it need an API key?

It needs Google Cloud credentials, or an Express Mode API key in GOOGLE_API_KEY. The skill's frontmatter says it requires active Google Cloud credentials and the Agent Platform API enabled.

Is it safe to install?

It bundles no scripts, but the agent may run package installs and gcloud commands, and the code it writes calls Google Cloud with your credentials. media_generation.md warns that Veo video generation can be costly. Approve commands one at a time and use a project with a budget alert.

Is this for the Gemini API key from Google AI Studio?

Not directly. SKILL.md covers Gemini on Agent Platform, formerly Vertex AI, and its auth steps are Google Cloud credentials or an Express Mode key. It does not describe a Google AI Studio setup. Its description says to use it when you mention Vertex AI, Google Cloud or Agent Platform.

What is the usage tag on its gcloud commands?

The skill tells the agent to prefix gcloud with an attribution variable that names the skill. The reference file says the tag identifies only the skill and its version and carries no user data, and that it is omitted if you ask to disable it.

Which models does it tell the agent to use?

It names gemini-3.8-flash for balanced work, gemini-3.1-pro-preview for complex reasoning and gemini-3.5-flash-lite for light tasks. It also lists image models and a live audio model, and calls gemini-2.0, 1.5 and 1.0 models legacy.

Information

updated Nov 2, 2026

Repository: google/skills
Folder: skills/cloud/gemini-api

google/skills skills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 8

gemini-api SKILL.md · github.com
attribution.md · github.com
google/skills README · github.com
advanced_features.md · github.com
media_generation.md · github.com
model_tuning.md · github.com
safety.md · github.com
live_api.md · github.com