Google ADK (Python) Skill
This skill provides comprehensive documentation and Python examples for the Google Agent Development Kit (ADK). It maps documentation topics to their corresponding Python code snippets.
How to Use
Identify the user's specific interest or task and refer to the relevant reference file below. Each reference file contains links to the official documentation (Markdown) and the corresponding Python examples (raw code).
Topics
1. Getting Started
For installation, quickstarts, and basic agent setup.
- Reference: references/get-started.md
2. Agents & Models
For creating different types of agents (LLM, Workflow, Loop, Parallel, Sequential) and configuring specific models (Gemini, Anthropic, etc.).
- Agents Reference: references/agents.md
- Models Reference: references/models.md
3. Tools (Basic & Advanced)
For integrating tools like Google Search, Code Execution, BigQuery, third-party services (GitHub, Jira, etc.), MCP, and Grounding.
- Basic Tools Reference: references/tools.md
- Advanced Tools Reference: references/advanced-tools.md
4. Streaming
For building real-time, low-latency streaming agents (audio/video).
- Reference: references/streaming.md
5. Callbacks
For hooking into agent lifecycle events (before/after agent, model, tool execution).
- Reference: references/callbacks.md
6. Runtime & Architecture
For deep dives into the Runtime, Sessions, Memory, Context, Events, Artifacts, and Plugins.
- Reference: references/runtime-arch.md
7. Deployment & Operations
For deploying agents (Cloud Run, GKE) and observability (Logging, Tracing, Evaluation).
- Reference: references/deploy-ops.md
8. Tutorials & Samples
For end-to-end tutorials and complete agent samples (e.g., YouTube Shorts Assistant, Weather Agent).
- Reference: references/tutorials.md
9. API Reference
For REST API details.
- Reference: references/api.md
10. General Information
For project info, community, release notes, and limitations.
- Reference: references/general.md
Instructions
- When a user asks about a specific topic, load the corresponding reference file to get the URLs for the documentation and code.
- You can read the content of the linked files using
web_fetchorrun_shell_commandwithcurlif you need to provide the actual content to the user. - Always prefer providing the Python example code when explaining a concept.