Google A2A Blueprints
30by e-vicius
System prompts and architectural patterns for building scalable AI agents with the Google A2A Protocol and AI SDK.
Getting Started
README
Google Agent Blueprints
System instructions and architectural rules for building scalable AI Agents with the Google A2A Protocol and Google AI Agent SDK.
📖 About This Repository
Building production-grade AI agents requires more than just good prompts—it requires strict architectural enforcement.
Google Agent Blueprints is a curated collection of System Prompts and IDE Rulesets designed to force your AI coding assistant (Cursor, Windsurf, Antigravity, etc.) to adhere to best practices defined by Google's Agent-to-Agent (A2A) protocol and the Google AI Agent SDK.
Instead of manually reminding your AI to "use a shared database" or "follow the A2A handshake," you can drop these files into your project to enforce those patterns automatically.
🚀 The Tech Stack
This repository focuses specifically on:
- Google AI Agent SDK: Best practices for scaffolding agents, tool definitions, and standard implementations.
- Google A2A (Agent-to-Agent) Protocol: Rules for defining how independent agents discover, handshake, and communicate with each other securely.
- Context Systems: Architectural patterns for shared memory and state management across multi-agent fleets.
📂 Repository Contents
| Blueprint | Description | Use Case |
|---|---|---|
rules/context-systems.md |
Enforces the "Shared Memory" architecture. Prevents siloed agent state by mandating a central memory microservice. | Multi-Agent Fleets |
rules/a2a-discovery.md |
Enforces the "Agent Card" pattern for discovery and interoperability. | Multi-Agent Fleets |
rules/multi-agent-orchestration.md |
Enforces the "Coordinator Pattern" for orchestrating multi-agent workflows. | Multi-Agent Fleets |
rules/agent-tools-and-safety.md |
Enforces the "Pydantic Mandate" for tool definition and safety. | Agent Development |
rules/agent-testing-standards.md |
Enforces the "Golden Dataset" pattern for testing. | Agent Development |
other/agent-observability.md |
Enforces the "Correlation ID" mandate for observability. | Agent Development |
🛠 How to Use
For Cursor / Windsurf / Antigravity
- Navigate to the pattern you want to implement (e.g.,
context-systems.md). - Copy the raw content.
- Paste it into your project's
.cursorrulesor.windsurfrulesfile (or your specific IDE's system instruction settings). - Result: Your AI assistant now "knows" how to architect systems according to that specific Google protocol and will refuse to generate code that violates it.
For Custom Agents
You can also inject these markdown files directly into your Agent's System Prompt to ensure they understand their own architectural constraints at runtime.
🤝 Contributing
We welcome contributions! If you have optimized a set of rules for the Google Agent ecosystem or found a better way to instruct AI IDEs on A2A patterns, please submit a PR.