# google-gemini/gemini-skills: Installing Gemini API Context Skills for AI Coding Assistants > gemini-skills is a GitHub repository maintained by the google-gemini organization that packages Gemini API documentation and best practices as installable context files for AI coding assistants. It addresses the problem that language models generate outdated or incorrect Gemini SDK code because they were trained before the API changed. **google-gemini/gemini-skills** — Skills for the Gemini API, SDK and model/agent interactions - Repository: https://github.com/google-gemini/gemini-skills - Website: https://ai.google.dev/gemini-api/docs - Stars: 4,233 · Forks: 448 - Language: Python - License: Apache-2.0 - Published: 2026-09-23 · Updated: 2026-09-23 · Language: en - Canonical page: https://hysenlabs.com/projects/google-gemini-gemini-skills ## The Knowledge Gap gemini-skills Addresses Language models are trained on data up to a fixed cutoff date. The Gemini API changes faster than a training cycle. New capabilities are introduced, SDK method signatures shift, and best practices evolve with each release. The result is that an AI coding assistant answering a question about the Gemini API may produce code that looks correct but uses deprecated patterns or ignores recent additions. gemini-skills solves this by packaging current Gemini API documentation and recommended usage patterns into small text files, called skills, that a developer installs into their coding assistant. The assistant reads those files as additional context before responding to Gemini-related questions. The README describes skills as a lightweight technique for adding relevant context to agents and points to agentskills.io as the underlying specification. The repository is maintained by the google-gemini organization. The README includes an explicit disclaimer that this is not an officially supported Google product and is not eligible for the Google Open Source Software Vulnerability Rewards Program. This is a meaningful caveat for teams that depend on support guarantees. ## The Three Skills in the Repository The repository provides three skills, each covering a different part of the Gemini API surface. The gemini-api-dev skill targets the Gemini Interactions API. The README lists its coverage as text generation, multi-turn chat, streaming, function calling, structured output, image generation, speech generation (TTS), Voice Design, Voice Replication, managed agents, Deep Research, deprecated model guardrails, and both Python and TypeScript SDKs. The gemini-live-api-dev skill covers the Gemini Live API, which is the real-time bidirectional streaming interface. Its documented coverage includes WebSocket-based audio, video, and text streaming; voice activity detection; native audio features; function calling; and session management. The gemini-omni-flash-api skill is specialized. It focuses on the Gemini Omni 1.1 Flash model (model identifier: gemini-omni-1.1-flash) and covers video editing, image and video referenced generation, first and last frame transitions, and video extensions through the Interactions API. The README also notes that a fourth skill, vertex-ai-api-dev, previously lived in this repository but has moved to the google/skills repository under skills/cloud/gemini-api. ## Installing gemini-skills in Claude Code, Cursor, and Other Assistants The README documents six installation paths. The correct path depends on which AI coding assistant you use. For tools supporting the Vercel skills CLI: ```sh npx skills add google-gemini/gemini-skills --list npx skills add google-gemini/gemini-skills --skill gemini-api-dev ``` The first command opens an interactive browser to select a skill; the second installs a specific one directly. For tools supporting Context7: ```sh npx ctx7 skills install /google-gemini/gemini-skills npx ctx7 skills install /google-gemini/gemini-skills gemini-api-dev ``` For Claude Code, the README gives two commands: first register the marketplace source, then install the plugin: ```sh /plugin marketplace add google-gemini/gemini-skills /plugin install gemini-skills@gemini-skills ``` For Cursor, install from the plugin directory or run: ```text /add-plugin google-gemini/gemini-skills ``` For OpenAI Codex: ```sh codex plugin marketplace add https://github.com/google-gemini/gemini-skills codex plugin add gemini-skills@gemini-skills ``` Antigravity bundles the Gemini skills by default. The README directs Antigravity users to the Customizations panel under Build with Google Plugins, or they can install using the AGY CLI: ```sh agy plugin install https://github.com/google-gemini/gemini-skills ``` ## What the README Claims About Accuracy Improvement The README includes a performance claim: it states that adding the gemini-skills skill improved an agent's ability to generate correct API code following best practices to 87% with Gemini 3 Flash and 96% with Gemini 3.1 Pro. The README links to a blog post titled "Closing the knowledge gap with agent skills" on developers.googleblog.com for methodology details. These numbers are from the repository maintainers' own evaluation. The README does not document the test set, the baseline, the task definition, or whether the evaluation covered Python, TypeScript, or both. A developer reading this as an objective benchmark would be overstating what the README supports. The claim is better read as an indication that the skill authors found the context files helpful for their own tests. ## What a Skills File Is Not: Context Windows and Manual Updates A skills file is a text document, not a code library. Installing it does not add a Python package, a JavaScript module, or a browser extension. It adds text to a coding assistant's context window before the assistant generates a response. This creates a practical constraint. A very long conversation or a context window already filled with code can push the skills text out of the available window. The README does not document how individual coding assistants handle context length when skills are installed, so behavior may vary. The repository also has no mechanism for automatically updating an installed skill. If a developer installs gemini-api-dev today and the Gemini API changes in three months, the skill will contain outdated information until the developer explicitly reinstalls or updates it. Finally, the repository covers only the Gemini API. It does not include skills for other Google Cloud APIs or for the Vertex AI platform; as noted, the Vertex AI skill moved to a different repository. ## When a Custom Instruction File Is the Better Choice The alternative to installing a published skill is maintaining a custom instruction file in your coding assistant. Cursor uses .cursorrules, Claude Code uses CLAUDE.md. Similar mechanisms exist in other tools. These files serve the same purpose as skills: they inject text into the assistant's context before answering questions. The difference in approach is curation versus customization. gemini-skills provides a maintained, Google-authored context document that tracks Gemini API changes. A custom instruction file is entirely under the developer's control: it can combine Gemini API documentation with project-specific conventions, internal API specifications, or team-specific constraints, but the developer must write and update the Gemini-specific content manually. For a team that works exclusively with the Gemini API and wants up-to-date context without maintaining it themselves, gemini-skills reduces that overhead. For a team that needs to mix Gemini documentation with proprietary context, or that uses a coding assistant not listed in the README, a custom instruction file is more practical. ## License, Official Support Status, and Last Push The repository is licensed under Apache-2.0. The last push was on 2026-09-23. The repository has no GitHub releases. Its structure follows the skills format: a plugin.json file at the root, skill content directories under skills/, and assistant-specific metadata directories (.claude-plugin/, .cursor-plugin/, .codex-plugin/). The plugin.json and the skill directories define what each coding assistant reads when the skill is installed. The README's explicit statement that this project is not officially supported by Google distinguishes it from the official Gemini SDK and documentation. Teams that need a contractual support path or a bug report process backed by Google should treat this repository as community-maintained context rather than an official resource. ## Conclusion gemini-skills is appropriate for a developer building Gemini API applications with an AI coding assistant who finds the assistant generating outdated SDK patterns. It is not a code library, a client SDK, or a replacement for the official Gemini documentation at ai.google.dev. Before installing, verify that your coding assistant is one of those documented in the README: Claude Code, Cursor, OpenAI Codex, Antigravity, and tools supporting the Vercel or Context7 CLIs. The repository is not an officially supported Google product and is not eligible for the Google Open Source Software Vulnerability Rewards Program. ## FAQ ### What are Gemini skills? In this repository, Gemini skills are context files installed into an AI coding assistant to provide current Gemini API documentation and best practices. They allow the assistant to generate more accurate Gemini SDK code than it would from its training data alone. ### How do I install gemini-skills? Installation depends on your coding assistant. For Claude Code, run /plugin marketplace add google-gemini/gemini-skills, then /plugin install gemini-skills@gemini-skills. For Cursor, run /add-plugin google-gemini/gemini-skills. For tools supporting the Vercel CLI, run npx skills add google-gemini/gemini-skills followed by --skill gemini-api-dev for a specific skill. ### Are Gemini skills the same as agents? No. The README describes skills as a lightweight technique for adding relevant context to agents. A skill is a context document that an agent reads. An agent is a system that acts on instructions; it reads skills as part of its context, but the two are separate concepts. ## Sources - [google-gemini/gemini-skills on GitHub](https://github.com/google-gemini/gemini-skills) - [Issues](https://github.com/google-gemini/gemini-skills/issues) - [License: Apache-2.0](https://github.com/google-gemini/gemini-skills/blob/main/LICENSE) - [Project website](https://ai.google.dev/gemini-api/docs) - [README](https://github.com/google-gemini/gemini-skills/blob/main/README.md) --- Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/google-gemini-gemini-skills