Google Launches Gemini API Skills to Fix Outdated AI Coding Knowledge

Google

Mar 31, 2026 · Updated Apr 25, 2026

Google introduced Gemini API Skills, a lightweight framework that provides AI agents with real-time SDK knowledge and documentation. By using the new Gemini API developer skill, the Gemini 3.1 Pro Preview model increased its coding task success rate from 28.2% to 96.6%. This shift demonstrates that providing agents with structured, up-to-date primitives is more effective than relying on static model weights for rapidly evolving software tools.

Google released the Gemini API developer skill, a package of instructions providing agents with feature explanations and verified code samples. It uses a system instruction and tools like activate_skill to retrieve live documentation, ensuring agents access fresh information instead of relying on static training data.

This update solves the problem of models hallucinating outdated code patterns for fast-moving APIs. In evaluations, the gemini-3.1-pro-preview model improved its success rate from 28.2% to 96.6% with the skill. The results show that models with strong reasoning capabilities see the most significant reliability gains when using these structured primitives.

You can integrate the gemini-api-dev skill into your workflows by installing it from GitHub. It guides agents toward the latest best practices and ensures they use correct SDK primitives. While Google is also exploring MCP, these skills offer a lightweight way to make any API agent-ready today.

View the full update on developers.googleblog.com

Google AI Developers

@googleaidevsMar 30

Give your coding agents up-to-date SDK knowledge with Gemini API skills. Our evaluation saw 96.6% success with the gemini-3.1-pro-preview model across 117 coding prompts, measuring model performance across agentic, chat, and SDK use cases. https://t.co/Zric8xoJfw

31445

View on X

Every HeadsUpAI update is written based on its original source and reviewed before it's published. Read our editorial standards →

See all AI news & updates from Google →

Keep reading

Google Launches Official MCP Server to Connect Coding Agents to Gemini Docs

Google released an official MCP server and developer skills that link coding agents directly to the latest Gemini API documentation. This allows autonomous agents to retrieve precise technical context on demand, enabling them to reach their highest performance potential while using significantly fewer tokens.

Google Gemma releases gemma-skills to accelerate agentic workflows with multi-token prediction

Google GemmaJun 2

Google Gemma releases gemma-skills to accelerate agentic workflows with multi-token prediction

Google launched the first iteration of gemma-skills, an open-source library of reusable capabilities for AI agents. By standardizing how agents select model sizes and use performance optimizations like MTP, Google is making it easier to build efficient, autonomous workflows on top of the Gemma ecosystem.

Google Labs Releases Agent Skills Library for Stitch MCP Server

Tom DörrMar 25

Google Labs Releases Agent Skills Library for Stitch MCP Server

Google Labs open-sourced stitch-skills, a library of agent skills for the Stitch MCP server following the Agent Skills open standard. Each skill installs with a single CLI command and runs across coding agents including Gemini CLI, Claude Code, and Cursor.

Google ADK Adds Six Agent Skills for Coding Agents to Its Docs

Richard SeroterMar 15

Google ADK Adds Six Agent Skills for Coding Agents to Its Docs

Google's Agent Development Kit now ships six agent skills in its docs repo, loadable into any coding agent via npx skills. The skills cover development lifecycle, evaluation, deployment, observability, and project scaffolding — structured ADK expertise for your coding agent.