Google Cloud Agent SDK Master
Master Google’s Agent Development Kit (ADK) patterns for building and deploying production-grade agents with clear tool contracts, validation, and operational guardrails.
Overview
Use this skill to quickly answer “how do I do X with Google ADK?” and to produce a safe, production-oriented plan (structure, patterns, deployment, verification) rather than ad-hoc snippets.
Examples
Example: Pick the right ADK pattern
- Request: “Should this be a single agent or a multi-agent orchestrator?”
- Output: an architecture recommendation with tradeoffs, plus a minimal scaffold plan.
Prerequisites
- The target environment (local-only vs Vertex AI Agent Engine)
- The agent’s core job, expected inputs/outputs, and required tools
- Any constraints (latency, cost, compliance/security)
Instructions
- Clarify requirements and choose an ADK architecture (single vs multi-agent; orchestration pattern).
- Define tool interfaces (inputs, outputs, and error contracts) and how secrets are managed.
- Provide an implementation plan with a minimal scaffold and incremental milestones.
- Add validation: smoke prompts, regression tests, and deployment verification steps.
Output
- A recommended ADK architecture and scaffold layout
- A checklist of commands to validate locally and in CI
- Optional: deployment steps and post-deploy health checks
Error Handling
- If documentation conflicts, prefer the latest canonical standards in
000-docs/6767-*. - If an API feature is unavailable in a region/version, propose a compatible alternative.
Resources
- Full detailed guide (kept for reference):
${CLAUDE_SKILL_DIR}/references/SKILL.full.md - ADK / Agent Engine docs: https://cloud.google.com/vertex-ai/docs/agent-engine
- Canonical repo standards:
000-docs/6767-a-SPEC-DR-STND-claude-code-plugins-standard.md