As the landscape of AI tools evolves, developers are increasingly leveraging agentic AI to enhance their work with Google Cloud products, including Firebase, the Gemini API, BigQuery, and GKE. A key challenge is ensuring that these AI agents have access to accurate and current information about these technologies.
One effective solution is connecting AI agents to a reliable, real-time information source. Google provides a Model Context Protocol (MCP) server for developer documentation; however, excessive reliance on MCP servers can lead to 'context bloat.' This occurs when too much context is loaded, which can confuse the model and increase token costs.
To address this issue, Google has introduced Agent Skills, a streamlined format designed to equip agents with new capabilities and expertise. Each skill serves as concise, agent-focused documentation for specific technologies or tasks. Written in Markdown, these skills can include reference files, code snippets, and other resources. Agents can load skill information as needed, which helps minimize the risk of context bloat.
On the first day of Google Cloud Next 2026, Google announced the launch of its official Agent Skills repository, available at github.com/google/skills.
The repository currently features thirteen skills centered on Google Cloud technologies, including:
- Product Skills: AlloyDB, BigQuery, Cloud Run, Cloud SQL, Firebase, Gemini API, and Google Kubernetes Engine (GKE).
- Well-Architected Pillar Skills: Security, Reliability, and Cost Optimization.
- Recipe Skills: Google Cloud Onboarding, Authentication, and Network Observability.
To install these skills, users can execute npx skills install github.com/google/skills for their preferred agents, such as Antigravity, Gemini CLI, and various third-party agents.
More skills are expected to be added to the repository in the upcoming weeks and months, encouraging developers to start building with these new resources.
