home / skills / jeremylongshore / claude-code-plugins-plus-skills / google-cloud-agent-sdk-master

This skill helps design production-ready ADK architectures for Google Cloud agents, outlining scaffolds, validation steps, and deployment plans.

npx playbooks add skill jeremylongshore/claude-code-plugins-plus-skills --skill google-cloud-agent-sdk-master

Review the files below or copy the command above to add this skill to your agents.

Files (5)
SKILL.md
2.2 KB
---
name: google-cloud-agent-sdk-master
description: |
  Execute automatic activation for all google cloud agent development kit (adk) Use when appropriate context detected. Trigger with relevant phrases based on skill purpose.
allowed-tools: Read, WebFetch, WebSearch, Grep
version: 1.0.0
author: Jeremy Longshore <[email protected]>
license: MIT
---
# 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

1. Clarify requirements and choose an ADK architecture (single vs multi-agent; orchestration pattern).
2. Define tool interfaces (inputs, outputs, and error contracts) and how secrets are managed.
3. Provide an implementation plan with a minimal scaffold and incremental milestones.
4. 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): `{baseDir}/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`

Overview

This skill automates activation and guidance for Google Cloud Agent Development Kit (ADK) workflows, producing production-ready plans instead of ad-hoc code snippets. It helps choose architectures, define tool contracts, scaffold minimal implementations, and specify validation and deployment checks tailored to Vertex AI or local environments. Use it to get repeatable, verifiable ADK patterns with operational guardrails.

How this skill works

When triggered, the skill asks clarifying questions about environment (local vs Vertex AI Agent Engine), core agent responsibilities, inputs/outputs, and constraints like latency or compliance. It then recommends a single-agent or multi-agent architecture, defines tool interfaces and secret handling, and emits a scaffolded implementation plan with incremental milestones, smoke tests, and CI/validation commands. It also suggests deployment steps, runtime health checks, and alternatives when features or regions are unavailable.

When to use it

  • Designing a new agent or migrating an ad-hoc script into a production ADK pipeline
  • Deciding between single-agent vs multi-agent orchestration patterns
  • Preparing CI/CD and validation plans for Vertex AI Agent Engine or local deployments
  • Defining tool contracts, error handling, and secret management for agents
  • Creating minimal scaffolds and stepwise milestones for incremental delivery
  • Troubleshooting region/version incompatibilities and proposing fallbacks

Best practices

  • Start by clarifying environment, expected inputs/outputs, and non-functional constraints
  • Define explicit tool interfaces: inputs, outputs, error contracts, and retry semantics
  • Keep a minimal scaffold for fast iteration, with clear incremental milestones
  • Automate smoke prompts, regression tests, and CI validation before deployment
  • Define operational guardrails: health checks, observability, secrets rotation, and rollback steps

Example use cases

  • Recommend architecture and scaffold when asked: “Should this be a single agent or multi-agent orchestrator?”
  • Produce a checklist of local and CI commands to validate an ADK scaffold before Vertex deployment
  • Define tool contracts and secret management for an agent that calls multiple external APIs
  • Create a stepwise rollout plan with smoke prompts and post-deploy health checks
  • Propose compatible alternatives when an API or feature is unavailable in a target region

FAQ

What inputs do you need to produce a reliable plan?

Provide environment target (local or Vertex Agent Engine), the agent’s core job, expected inputs/outputs, required tools, and constraints like latency, cost, or compliance.

How do you handle unavailable features or regional limits?

I propose compatible alternatives and fallback patterns, preferring the latest canonical standards and documented API substitutes when a feature or region is unsupported.