Search skills/

Google Antigravity SDK skill Claude skill

A skill that teaches a coding agent to build, configure and debug agents with Google's Antigravity Python SDK, with reference docs and starter examples.

google-antigravity/antigravity-sdk-python 1 skill Skill & agent tooling
Our verdict Recommended

A vendor-maintained, well-referenced skill that is useful if and only if you build on the Antigravity SDK.

3/5 Depth expertise shipped Note Security Review before installing 1d Maintenance Updated today 3.9K Adoption stars OK License Apache-2.0, commercial OK

What Google Antigravity SDK skill does

This repository is Google's Python SDK for building agents on Antigravity and Gemini, and it ships one agent skill, google-antigravity-sdk, in the skills folder. With the skill installed, a coding agent that is asked to create or orchestrate an Antigravity agent first checks the project for the google-antigravity dependency, installs it if needed, and walks you through authentication. The default path is a GEMINI_API_KEY, with a pointer to Google AI Studio for getting one. The skill also covers Vertex-style enterprise setups (express mode with an API key, or standard mode using a project, location and Application Default Credentials) and local model configs that need no key.

The skill file is about 8.5 KB. It sits next to eight reference documents (agent configuration, architecture, built-in tools, error handling, local models, MCP integration, observability, safety policies) and a folder of getting-started examples. The main SDK relies on a compiled runtime binary that is only included in the platform wheels on PyPI, so cloning the repo is not enough; the README says to run pip install google-antigravity. The skill is documentation for the agent, not a set of scripts: it has no helper scripts of its own. Its value is limited to people who are actually building on this particular SDK.

Good for

  • Building agents on the Google Antigravity Python SDK
  • Setting up Gemini or Vertex authentication for an SDK project
  • Looking up SDK configuration, tools and safety policy options from inside the agent

Not for

  • Projects that do not use the Antigravity SDK
  • General agent-building advice for other frameworks

What's inside depth 3/5

One 8.5 KB skill with 8 reference docs and 19 example files, no helper scripts or evals.

Skills
1
Reference docs
8
Scripts
0
Tests / evals
0
Templates
19
Avg SKILL.md
8.3 KB

Counted from the repository file tree. How we score depth.

Dependencies and costs

  • google-antigravity package from PyPI (includes a compiled runtime binary)
  • GEMINI_API_KEY, or Vertex credentials via Application Default Credentials
  • Python

Security review

Review before installing

Google's Python SDK for building Antigravity agents plus a skill that guides an agent through using it. The SDK launches a closed-source local harness binary that talks to Gemini with your API key, passes basic client information (SDK, Python and OS versions) to the harness, and can be configured to allow unrestricted shell access (not the default). Only a small part of the code was read directly.

Worth knowing

  • Relies on a CLI or package we could not reviewThe SDK spawns a compiled localharness binary (found in the pip wheel's google/antigravity/bin, via ANTIGRAVITY harness-path env var, or on PATH) with the merged environment, talks to it over stdio and a local WebSocket, and lets it run the agent loop and tools; the binary's code is not in this repo.google/antigravity/connections/local/local_connection.py:1491
  • Needs an API key for its own serviceThe skill has the agent check for GEMINI_API_KEY (environment or .env) and help you obtain one; the SDK sends it as x-goog-api-key to Google's Gemini / Vertex endpoints.skills/google-antigravity-sdk/SKILL.md:22
  • Makes network requestsAgent prompts, tool results and files you give the agent are sent to hosted Gemini models (or Vertex) by default; LiteRT and local OpenAI-compatible connections keep traffic on your machine.google/antigravity/connections/local/local_connection.py:1477
  • Installs packages or toolsThe skill tells the agent to make sure the google-antigravity Python package (which bundles the harness binary) is installed in the project.skills/google-antigravity-sdk/SKILL.md:18

What we checked

The repository (183 files) on Oct 29, 2026. Not a full review: we read 1 of 115 code files; it relies on code outside the repository: The compiled localharness binary shipped in the google-antigravity wheel, and the Gemini / Vertex services; Not read in full: the SDK package (about 40 non-test modules), 70 test files, 40 example scripts and .kokoro release scripts; inspected by grep and targeted reading of local_connection.py's harness-spawn code.

NVIDIA SkillSpector static scan: 105 pattern hits (Excessive Agency 26, Tool Misuse 24, Dangerous Code Execution 19, Supply Chain 18, MCP Rug Pull 4, Privilege Escalation 3). The 'rm -rf /' hits are test inputs for the policy and hook engines, exec() appears in a workflow test, '{**os.environ, **env_map}' passes the environment to the harness subprocess, and 'send messages to' is a trigger docstring. The MCP metadata poisoning match is a pirate-themed example MCP server. Kokoro scripts are Google's internal release automation.

A review, not a certification. Skills change after we check them — read the files before installing.

License

Apache-2.0 — commercial use allowed.

Can it run unattended?

Not really. It is a reference and setup guide for building agents, not a loop that runs on its own.

Alternatives to Google Antigravity SDK skill

Google agents-cli

google/agents-cli

Google's CLI and seven skills that teach a coding agent to scaffold, build, evaluate, deploy and observe ADK agents on Google Cloud.

With caveats Check first Depth 4/5

Google Agent Skills

google/skills

Google's official collection of about 180 skills for Google Cloud, Gemini APIs, Analytics, Data Manager, sign-in and other Google products.

Recommended Broad access Depth 3/5

FAQ

What is the Google Antigravity SDK skill?

It is a single agent skill, google-antigravity-sdk, bundled in Google's antigravity-sdk-python repo. It tells a coding agent how to install the SDK, authenticate and write or debug Antigravity agents, backed by eight reference docs.

How do I install the Antigravity SDK and its skill?

Install the SDK with pip install google-antigravity, because the compiled runtime is only in the PyPI wheels. The skill lives in the repo's skills folder and can be copied into your agent's skills directory.

Do I need an API key?

For hosted Gemini models you need a GEMINI_API_KEY or Vertex credentials. The docs say local model configs can be used without a key or cloud login.

Sources: GitHub repository (stars, license, file tree, last push Oct 28, 2026). Review based on the README and SKILL.md files as published.