[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6F6HRTMGSR58PBGP19EBJY7":3,"skill-same:01M6F6HRTMGSR58PBGP19EBJY7":74},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":26,"publishedAt":46,"linkOnly":47,"markdownContent":48,"markdownBody":49,"frontMatter":50,"refName":51,"commitSha":52,"skillDirectory":53,"forksCount":54,"repositoryHtmlUrl":55,"repositoryHomepageUrl":56,"ownerHtmlUrl":57,"ownerBlogUrl":6,"ownerDescription":58,"updatedAt":59,"install":60,"resources":73},"01M6F6HRTMGSR58PBGP19EBJY7","Google Cloud Agent SDK Master",null,"Automatic activation for ALL Google Cloud Agent Development Kit (ADK) and Agent Starter Pack operations - multi-agent systems, containerized deployment, RAG agents, and production orchestration.\n\n**TRIGGER PHRASES:**\n- \"adk\", \"agent development kit\", \"agent starter pack\", \"multi-agent\", \"build agent\"\n- \"cloud run agent\", \"gke deployment\", \"agent engine\", \"containerized agent\"\n- \"rag agent\", \"react agent\", \"agent orchestration\", \"agent templates\"\n\n**AUTO-INVOKES FOR:**\n- Agent creation and scaffolding\n- Multi-agent system design\n- Containerized agent deployment\n- RAG (Retrieval-Augmented Generation) implementation\n- CI\u002FCD pipeline setup for agents\n- Agent evaluation and monitoring","ComeOnOliver\u002Fskillshub","ComeOnOliver","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F43215937?v=4","01M6E5JNFECJAAVFE0W7SMWWD2","Haoran Shu","01M6E5JNNDQS7AJZDYFQ8TPZCK",67,"MIT","https:\u002F\u002Fgithub.com\u002FComeOnOliver\u002Fskillshub\u002Fblob\u002Fmain\u002Fskills\u002Fjeremylongshore\u002Fclaude-code-plugins-plus-skills\u002Fagent-sdk-master\u002FSKILL.md",[18,23],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000011","运维与部署","SCENARIO","使用场景",{"categoryNo":24,"categoryName":25,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000023","Skill与智能体开发",[27,28,29,30,31,5,32,33,34,35,36,37,38,39,40,41,42,43,44,45],"google","sdk","agent sdk","google cloud","cloud","master","build agent","design","creation","scaffolding","agent orchestration","system design","auto","agent development","pipeline","ci","orchestration","agent development kit","adk","2026-11-10 10:12:30",false,"---\nname: Google Cloud Agent SDK Master\ndescription: |\n  Automatic activation for ALL Google Cloud Agent Development Kit (ADK) and Agent Starter Pack operations - multi-agent systems, containerized deployment, RAG agents, and production orchestration.\n\n  **TRIGGER PHRASES:**\n  - \"adk\", \"agent development kit\", \"agent starter pack\", \"multi-agent\", \"build agent\"\n  - \"cloud run agent\", \"gke deployment\", \"agent engine\", \"containerized agent\"\n  - \"rag agent\", \"react agent\", \"agent orchestration\", \"agent templates\"\n\n  **AUTO-INVOKES FOR:**\n  - Agent creation and scaffolding\n  - Multi-agent system design\n  - Containerized agent deployment\n  - RAG (Retrieval-Augmented Generation) implementation\n  - CI\u002FCD pipeline setup for agents\n  - Agent evaluation and monitoring\n---\n\n# Google Cloud Agent SDK Master - Production-Ready Agent Systems\n\nThis Agent Skill provides comprehensive mastery of Google's Agent Development Kit (ADK) and Agent Starter Pack for building and deploying production-grade containerized agents.\n\n## Core Capabilities\n\n### ? Agent Development Kit (ADK)\n\n**Framework Overview:**\n- **Open-source Python framework** from Google\n- Same framework powering Google Agentspace and CES\n- Build production agents in \u003C100 lines of code\n- Model-agnostic (optimized for Gemini)\n- Deployment-agnostic (local, Cloud Run, GKE, Agent Engine)\n\n**Supported Agent Types:**\n1. **LLM Agents**: Dynamic routing with intelligence\n2. **Workflow Agents**:\n   - Sequential: Linear execution\n   - Loop: Iterative processing\n   - Parallel: Concurrent execution\n3. **Custom Agents**: User-defined implementations\n4. **Multi-agent Systems**: Hierarchical coordination\n\n**Key Features:**\n- Flexible orchestration (workflow & LLM-driven)\n- Tool ecosystem (search, code execution, custom functions)\n- Third-party integrations (LangChain, CrewAI)\n- Agents-as-tools capability\n- Built-in evaluation framework\n- Cloud Trace integration\n\n### ? Agent Starter Pack\n\n**Production Templates:**\n1. **adk_base** - ReAct agent using ADK\n2. **agentic_rag** - Document retrieval + Q&A with search\n3. **langgraph_base_react** - LangGraph ReAct implementation\n4. **crewai_coding_crew** - Multi-agent coding system\n5. **adk_live** - Multimodal RAG (audio\u002Fvideo\u002Ftext)\n\n**Infrastructure Automation:**\n- CI\u002FCD setup with single command\n- GitHub Actions or Cloud Build pipelines\n- Multi-environment support (dev, staging, prod)\n- Automated testing and evaluation\n- Deployment rollback mechanisms\n\n### ? Deployment Targets\n\n**1. Vertex AI Agent Engine**\n- Fully managed runtime\n- Auto-scaling and load balancing\n- Built-in observability\n- Serverless architecture\n- Best for: Production-scale agents\n\n**2. Cloud Run**\n- Containerized serverless\n- Pay-per-use pricing\n- Custom domain support\n- Traffic splitting\n- Best for: Web-facing agents\n\n**3. Google Kubernetes Engine (GKE)**\n- Full container orchestration\n- Advanced networking\n- Resource management\n- Multi-cluster support\n- Best for: Complex multi-agent systems\n\n**4. Local\u002FDocker**\n- Development and testing\n- Custom infrastructure\n- On-premises deployment\n- Best for: POC and debugging\n\n### ? Technical Implementation\n\n**Installation:**\n```bash\n# Agent Starter Pack (recommended)\npip install agent-starter-pack\n\n# or direct from GitHub\nuvx agent-starter-pack create my-agent\n\n# ADK only\npip install google-cloud-aiplatform[adk,agent_engines]>=1.111\n```\n\n**Create Agent (ADK):**\n```python\nfrom google.cloud.aiplatform import agent\nfrom vertexai.preview.agents import ADKAgent\n\n# Simple ReAct agent\n@agent.adk_agent\nclass MyAgent(ADKAgent):\n    def __init__(self):\n        super().__init__(\n            model=\"gemini-2.5-pro\",\n            tools=[search_tool, code_exec_tool]\n        )\n\n    def run(self, query: str):\n        return self.generate(query)\n\n# Multi-agent orchestration\nclass OrchestratorAgent(ADKAgent):\n    def __init__(self):\n        self.research_agent = ResearchAgent()\n        self.analysis_agent = AnalysisAgent()\n        self.writer_agent = WriterAgent()\n\n    def run(self, task: str):\n        research = self.research_agent.run(task)\n        analysis = self.analysis_agent.run(research)\n        output = self.writer_agent.run(analysis)\n        return output\n```\n\n**Using Agent Starter Pack:**\n```bash\n# Create project with template\nuvx agent-starter-pack create my-rag-agent \\\n    --template agentic_rag \\\n    --deployment cloud_run\n\n# Generates complete structure:\nmy-rag-agent\u002F\n├── src\u002F\n│   ├── agent.py          # Agent implementation\n│   ├── tools\u002F            # Custom tools\n│   └── config.py         # Configuration\n├── deployment\u002F\n│   ├── Dockerfile\n│   ├── cloudbuild.yaml\n│   └── terraform\u002F\n├── tests\u002F\n│   ├── unit_tests.py\n│   └── integration_tests.py\n└── .github\u002Fworkflows\u002F    # CI\u002FCD pipelines\n```\n\n**Deploy to Cloud Run:**\n```bash\n# Using ADK CLI\nadk deploy \\\n    --target cloud_run \\\n    --region us-central1 \\\n    --service-account sa@project.iam.gserviceaccount.com\n\n# Manual with Docker\ndocker build -t gcr.io\u002FPROJECT\u002Fagent:latest .\ndocker push gcr.io\u002FPROJECT\u002Fagent:latest\ngcloud run deploy agent \\\n    --image gcr.io\u002FPROJECT\u002Fagent:latest \\\n    --region us-central1 \\\n    --allow-unauthenticated\n```\n\n**Deploy to Agent Engine:**\n```bash\n# Using Agent Starter Pack\nasp deploy \\\n    --env production \\\n    --target agent_engine\n\n# Manual deployment\nfrom google.cloud.aiplatform import agent_engines\nagent_engines.deploy_agent(\n    agent_id=\"my-agent\",\n    project=\"PROJECT_ID\",\n    location=\"us-central1\"\n)\n```\n\n### ? RAG Agent Implementation\n\n**Vector Search Integration:**\n```python\nfrom vertexai.preview.rag import VectorSearchTool\nfrom google.cloud import aiplatform\n\n# Set up vector search\nvector_search = VectorSearchTool(\n    index_endpoint=\"projects\u002FPROJECT\u002Flocations\u002FLOCATION\u002FindexEndpoints\u002FINDEX_ID\",\n    deployed_index_id=\"deployed_index\"\n)\n\n# RAG agent with ADK\nclass RAGAgent(ADKAgent):\n    def __init__(self):\n        super().__init__(\n            model=\"gemini-2.5-pro\",\n            tools=[vector_search, web_search_tool]\n        )\n\n    def run(self, query: str):\n        # Retrieves relevant docs automatically\n        response = self.generate(\n            f\"Answer this using retrieved context: {query}\"\n        )\n        return response\n```\n\n**Vertex AI Search Integration:**\n```python\nfrom vertexai.preview.search import VertexAISearchTool\n\n# Enterprise search integration\nvertex_search = VertexAISearchTool(\n    data_store_id=\"DATA_STORE_ID\",\n    project=\"PROJECT_ID\"\n)\n\nagent = ADKAgent(\n    model=\"gemini-2.5-pro\",\n    tools=[vertex_search]\n)\n```\n\n### ? CI\u002FCD Automation\n\n**GitHub Actions (auto-generated):**\n```yaml\nname: Deploy Agent\non:\n  push:\n    branches: [main]\n\njobs:\n  deploy:\n    runs-on: ubuntu-latest\n    steps:\n      - uses: actions\u002Fcheckout@v3\n      - name: Test Agent\n        run: pytest tests\u002F\n      - name: Deploy to Cloud Run\n        run: |\n          gcloud run deploy agent \\\n            --source . \\\n            --region us-central1\n```\n\n**Cloud Build Pipeline:**\n```yaml\nsteps:\n  # Build container\n  - name: 'gcr.io\u002Fcloud-builders\u002Fdocker'\n    args: ['build', '-t', 'gcr.io\u002F$PROJECT_ID\u002Fagent', '.']\n\n  # Run tests\n  - name: 'gcr.io\u002F$PROJECT_ID\u002Fagent'\n    args: ['pytest', 'tests\u002F']\n\n  # Deploy to Cloud Run\n  - name: 'gcr.io\u002Fcloud-builders\u002Fgcloud'\n    args:\n      - 'run'\n      - 'deploy'\n      - 'agent'\n      - '--image=gcr.io\u002F$PROJECT_ID\u002Fagent'\n      - '--region=us-central1'\n```\n\n### ? Multi-Agent Orchestration\n\n**Hierarchical Agents:**\n```python\n# Coordinator agent with specialized sub-agents\nclass ProjectManagerAgent(ADKAgent):\n    def __init__(self):\n        self.researcher = ResearchAgent()\n        self.analyst = AnalysisAgent()\n        self.writer = WriterAgent()\n        self.reviewer = ReviewAgent()\n\n    def run(self, project_brief: str):\n        # Coordinate multiple agents\n        research = self.researcher.run(project_brief)\n        analysis = self.analyst.run(research)\n        draft = self.writer.run(analysis)\n        final = self.reviewer.run(draft)\n        return final\n```\n\n**Parallel Agent Execution:**\n```python\nimport asyncio\n\nclass ParallelResearchAgent(ADKAgent):\n    async def research_topic(self, topics: list[str]):\n        # Run multiple agents concurrently\n        tasks = [\n            self.specialized_agent(topic)\n            for topic in topics\n        ]\n        results = await asyncio.gather(*tasks)\n        return self.synthesize(results)\n```\n\n### ? Evaluation & Monitoring\n\n**Built-in Evaluation:**\n```python\nfrom google.cloud.aiplatform import agent_evaluation\n\n# Define evaluation metrics\neval_config = agent_evaluation.EvaluationConfig(\n    metrics=[\"accuracy\", \"relevance\", \"safety\"],\n    test_dataset=\"gs:\u002F\u002Fbucket\u002Feval_data.jsonl\"\n)\n\n# Run evaluation\nresults = agent.evaluate(eval_config)\nprint(f\"Accuracy: {results.accuracy}\")\nprint(f\"Relevance: {results.relevance}\")\n```\n\n**Cloud Trace Integration:**\n```python\nfrom google.cloud import trace_v1\n\n# Automatic tracing\n@traced_agent\nclass MonitoredAgent(ADKAgent):\n    def run(self, query: str):\n        # All calls automatically traced\n        with self.trace_span(\"retrieval\"):\n            docs = self.retrieve(query)\n\n        with self.trace_span(\"generation\"):\n            response = self.generate(query, docs)\n\n        return response\n```\n\n### ? Security & Best Practices\n\n**1. Service Account Management:**\n```bash\n# Create minimal-permission service account\ngcloud iam service-accounts create agent-sa \\\n    --display-name \"Agent Service Account\"\n\n# Grant only required permissions\ngcloud projects add-iam-policy-binding PROJECT_ID \\\n    --member=\"serviceAccount:agent-sa@PROJECT.iam.gserviceaccount.com\" \\\n    --role=\"roles\u002Faiplatform.user\"\n```\n\n**2. Secret Management:**\n```python\nfrom google.cloud import secretmanager\n\ndef get_api_key():\n    client = secretmanager.SecretManagerServiceClient()\n    name = \"projects\u002FPROJECT\u002Fsecrets\u002Fapi-key\u002Fversions\u002Flatest\"\n    response = client.access_secret_version(name=name)\n    return response.payload.data.decode('UTF-8')\n```\n\n**3. VPC Service Controls:**\n```bash\n# Enable VPC SC for data security\ngcloud access-context-manager perimeters create agent-perimeter \\\n    --resources=projects\u002FPROJECT_ID \\\n    --restricted-services=aiplatform.googleapis.com\n```\n\n### ? Cost Optimization\n\n**Strategies:**\n- Use Gemini 2.5 Flash for most operations\n- Cache embeddings for RAG systems\n- Implement request batching\n- Use preemptible GKE nodes\n- Monitor token usage in Cloud Monitoring\n\n**Pricing Examples:**\n- Cloud Run: $0.00024\u002FGB-second\n- Agent Engine: Pay-per-request pricing\n- GKE: Standard cluster costs\n- Gemini API: $3.50\u002F1M tokens (Pro)\n\n### ? Reference Architecture\n\n**Production Agent System:**\n```\n┌─────────────────┐\n│   Load Balancer │\n└────────┬────────┘\n         │\n    ┌────▼────┐\n    │Cloud Run│ (Agent containers)\n    └────┬────┘\n         │\n    ┌────▼──────────┐\n    │ Agent Engine  │ (Orchestration)\n    └────┬──────────┘\n         │\n    ┌────▼────────────────┐\n    │  Vertex AI Search   │ (RAG)\n    │  Vector Search      │\n    │  Gemini 2.5 Pro     │\n    └─────────────────────┘\n```\n\n### ? Best Practices for Jeremy\n\n**1. Start with Templates:**\n```bash\n# Use Agent Starter Pack templates\nuvx agent-starter-pack create my-agent --template agentic_rag\n```\n\n**2. Local Development:**\n```bash\n# Test locally first\nadk serve --port 8080\ncurl http:\u002F\u002Flocalhost:8080\u002Fquery -d '{\"question\": \"test\"}'\n```\n\n**3. Gradual Deployment:**\n```bash\n# Deploy to dev → staging → prod\nasp deploy --env dev\n# Test thoroughly\nasp deploy --env staging\n# Final production push\nasp deploy --env production\n```\n\n**4. Monitor Everything:**\n- Enable Cloud Trace\n- Set up error reporting\n- Track token usage\n- Monitor response times\n- Set up alerting\n\n### ? Official Documentation\n\n**Core Resources:**\n- ADK Docs: https:\u002F\u002Fgoogle.github.io\u002Fadk-docs\u002F\n- Agent Starter Pack: https:\u002F\u002Fgithub.com\u002FGoogleCloudPlatform\u002Fagent-starter-pack\n- Agent Engine: https:\u002F\u002Fcloud.google.com\u002Fvertex-ai\u002Fgenerative-ai\u002Fdocs\u002Fagent-engine\u002Foverview\n- Agent Builder: https:\u002F\u002Fcloud.google.com\u002Fproducts\u002Fagent-builder\n\n**Tutorials:**\n- Building AI Agents: https:\u002F\u002Fcodelabs.developers.google.com\u002Fdevsite\u002Fcodelabs\u002Fbuilding-ai-agents-vertexai\n- Multi-agent Systems: https:\u002F\u002Fcloud.google.com\u002Fblog\u002Fproducts\u002Fai-machine-learning\u002Fbuild-and-manage-multi-system-agents-with-vertex-ai\n\n## When This Skill Activates\n\nThis skill automatically activates when you mention:\n- Agent development, ADK, or Agent Starter Pack\n- Multi-agent systems or orchestration\n- Containerized agent deployment\n- Cloud Run, GKE, or Agent Engine deployment\n- RAG agents or ReAct agents\n- Agent templates or scaffolding\n- CI\u002FCD for agents\n- Production agent systems\n\n## Integration with Other Services\n\n**Google Cloud:**\n- Vertex AI (Gemini, Search, Vector Search)\n- Cloud Storage (data storage)\n- Cloud Functions (triggers)\n- Cloud Scheduler (automation)\n- Cloud Logging & Monitoring\n\n**Third-party:**\n- LangChain integration\n- CrewAI orchestration\n- Custom tool frameworks\n\n## Success Metrics\n\n**Track:**\n- Agent response time (target: \u003C2s)\n- Evaluation scores (target: >85% accuracy)\n- Deployment frequency (target: daily)\n- System uptime (target: 99.9%)\n- Cost per query (target: \u003C$0.01)\n\n---\n\n**This skill makes Jeremy a Google Cloud agent architecture expert with instant access to ADK, Agent Starter Pack, and production deployment patterns.**\n\n","# Google Cloud Agent SDK Master - Production-Ready Agent Systems\n\nThis Agent Skill provides comprehensive mastery of Google's Agent Development Kit (ADK) and Agent Starter Pack for building and deploying production-grade containerized agents.\n\n## Core Capabilities\n\n### ? Agent Development Kit (ADK)\n\n**Framework Overview:**\n- **Open-source Python framework** from Google\n- Same framework powering Google Agentspace and CES\n- Build production agents in \u003C100 lines of code\n- Model-agnostic (optimized for Gemini)\n- Deployment-agnostic (local, Cloud Run, GKE, Agent Engine)\n\n**Supported Agent Types:**\n1. **LLM Agents**: Dynamic routing with intelligence\n2. **Workflow Agents**:\n   - Sequential: Linear execution\n   - Loop: Iterative processing\n   - Parallel: Concurrent execution\n3. **Custom Agents**: User-defined implementations\n4. **Multi-agent Systems**: Hierarchical coordination\n\n**Key Features:**\n- Flexible orchestration (workflow & LLM-driven)\n- Tool ecosystem (search, code execution, custom functions)\n- Third-party integrations (LangChain, CrewAI)\n- Agents-as-tools capability\n- Built-in evaluation framework\n- Cloud Trace integration\n\n### ? Agent Starter Pack\n\n**Production Templates:**\n1. **adk_base** - ReAct agent using ADK\n2. **agentic_rag** - Document retrieval + Q&A with search\n3. **langgraph_base_react** - LangGraph ReAct implementation\n4. **crewai_coding_crew** - Multi-agent coding system\n5. **adk_live** - Multimodal RAG (audio\u002Fvideo\u002Ftext)\n\n**Infrastructure Automation:**\n- CI\u002FCD setup with single command\n- GitHub Actions or Cloud Build pipelines\n- Multi-environment support (dev, staging, prod)\n- Automated testing and evaluation\n- Deployment rollback mechanisms\n\n### ? Deployment Targets\n\n**1. Vertex AI Agent Engine**\n- Fully managed runtime\n- Auto-scaling and load balancing\n- Built-in observability\n- Serverless architecture\n- Best for: Production-scale agents\n\n**2. Cloud Run**\n- Containerized serverless\n- Pay-per-use pricing\n- Custom domain support\n- Traffic splitting\n- Best for: Web-facing agents\n\n**3. Google Kubernetes Engine (GKE)**\n- Full container orchestration\n- Advanced networking\n- Resource management\n- Multi-cluster support\n- Best for: Complex multi-agent systems\n\n**4. Local\u002FDocker**\n- Development and testing\n- Custom infrastructure\n- On-premises deployment\n- Best for: POC and debugging\n\n### ? Technical Implementation\n\n**Installation:**\n```bash\n# Agent Starter Pack (recommended)\npip install agent-starter-pack\n\n# or direct from GitHub\nuvx agent-starter-pack create my-agent\n\n# ADK only\npip install google-cloud-aiplatform[adk,agent_engines]>=1.111\n```\n\n**Create Agent (ADK):**\n```python\nfrom google.cloud.aiplatform import agent\nfrom vertexai.preview.agents import ADKAgent\n\n# Simple ReAct agent\n@agent.adk_agent\nclass MyAgent(ADKAgent):\n    def __init__(self):\n        super().__init__(\n            model=\"gemini-2.5-pro\",\n            tools=[search_tool, code_exec_tool]\n        )\n\n    def run(self, query: str):\n        return self.generate(query)\n\n# Multi-agent orchestration\nclass OrchestratorAgent(ADKAgent):\n    def __init__(self):\n        self.research_agent = ResearchAgent()\n        self.analysis_agent = AnalysisAgent()\n        self.writer_agent = WriterAgent()\n\n    def run(self, task: str):\n        research = self.research_agent.run(task)\n        analysis = self.analysis_agent.run(research)\n        output = self.writer_agent.run(analysis)\n        return output\n```\n\n**Using Agent Starter Pack:**\n```bash\n# Create project with template\nuvx agent-starter-pack create my-rag-agent \\\n    --template agentic_rag \\\n    --deployment cloud_run\n\n# Generates complete structure:\nmy-rag-agent\u002F\n├── src\u002F\n│   ├── agent.py          # Agent implementation\n│   ├── tools\u002F            # Custom tools\n│   └── config.py         # Configuration\n├── deployment\u002F\n│   ├── Dockerfile\n│   ├── cloudbuild.yaml\n│   └── terraform\u002F\n├── tests\u002F\n│   ├── unit_tests.py\n│   └── integration_tests.py\n└── .github\u002Fworkflows\u002F    # CI\u002FCD pipelines\n```\n\n**Deploy to Cloud Run:**\n```bash\n# Using ADK CLI\nadk deploy \\\n    --target cloud_run \\\n    --region us-central1 \\\n    --service-account sa@project.iam.gserviceaccount.com\n\n# Manual with Docker\ndocker build -t gcr.io\u002FPROJECT\u002Fagent:latest .\ndocker push gcr.io\u002FPROJECT\u002Fagent:latest\ngcloud run deploy agent \\\n    --image gcr.io\u002FPROJECT\u002Fagent:latest \\\n    --region us-central1 \\\n    --allow-unauthenticated\n```\n\n**Deploy to Agent Engine:**\n```bash\n# Using Agent Starter Pack\nasp deploy \\\n    --env production \\\n    --target agent_engine\n\n# Manual deployment\nfrom google.cloud.aiplatform import agent_engines\nagent_engines.deploy_agent(\n    agent_id=\"my-agent\",\n    project=\"PROJECT_ID\",\n    location=\"us-central1\"\n)\n```\n\n### ? RAG Agent Implementation\n\n**Vector Search Integration:**\n```python\nfrom vertexai.preview.rag import VectorSearchTool\nfrom google.cloud import aiplatform\n\n# Set up vector search\nvector_search = VectorSearchTool(\n    index_endpoint=\"projects\u002FPROJECT\u002Flocations\u002FLOCATION\u002FindexEndpoints\u002FINDEX_ID\",\n    deployed_index_id=\"deployed_index\"\n)\n\n# RAG agent with ADK\nclass RAGAgent(ADKAgent):\n    def __init__(self):\n        super().__init__(\n            model=\"gemini-2.5-pro\",\n            tools=[vector_search, web_search_tool]\n        )\n\n    def run(self, query: str):\n        # Retrieves relevant docs automatically\n        response = self.generate(\n            f\"Answer this using retrieved context: {query}\"\n        )\n        return response\n```\n\n**Vertex AI Search Integration:**\n```python\nfrom vertexai.preview.search import VertexAISearchTool\n\n# Enterprise search integration\nvertex_search = VertexAISearchTool(\n    data_store_id=\"DATA_STORE_ID\",\n    project=\"PROJECT_ID\"\n)\n\nagent = ADKAgent(\n    model=\"gemini-2.5-pro\",\n    tools=[vertex_search]\n)\n```\n\n### ? CI\u002FCD Automation\n\n**GitHub Actions (auto-generated):**\n```yaml\nname: Deploy Agent\non:\n  push:\n    branches: [main]\n\njobs:\n  deploy:\n    runs-on: ubuntu-latest\n    steps:\n      - uses: actions\u002Fcheckout@v3\n      - name: Test Agent\n        run: pytest tests\u002F\n      - name: Deploy to Cloud Run\n        run: |\n          gcloud run deploy agent \\\n            --source . \\\n            --region us-central1\n```\n\n**Cloud Build Pipeline:**\n```yaml\nsteps:\n  # Build container\n  - name: 'gcr.io\u002Fcloud-builders\u002Fdocker'\n    args: ['build', '-t', 'gcr.io\u002F$PROJECT_ID\u002Fagent', '.']\n\n  # Run tests\n  - name: 'gcr.io\u002F$PROJECT_ID\u002Fagent'\n    args: ['pytest', 'tests\u002F']\n\n  # Deploy to Cloud Run\n  - name: 'gcr.io\u002Fcloud-builders\u002Fgcloud'\n    args:\n      - 'run'\n      - 'deploy'\n      - 'agent'\n      - '--image=gcr.io\u002F$PROJECT_ID\u002Fagent'\n      - '--region=us-central1'\n```\n\n### ? Multi-Agent Orchestration\n\n**Hierarchical Agents:**\n```python\n# Coordinator agent with specialized sub-agents\nclass ProjectManagerAgent(ADKAgent):\n    def __init__(self):\n        self.researcher = ResearchAgent()\n        self.analyst = AnalysisAgent()\n        self.writer = WriterAgent()\n        self.reviewer = ReviewAgent()\n\n    def run(self, project_brief: str):\n        # Coordinate multiple agents\n        research = self.researcher.run(project_brief)\n        analysis = self.analyst.run(research)\n        draft = self.writer.run(analysis)\n        final = self.reviewer.run(draft)\n        return final\n```\n\n**Parallel Agent Execution:**\n```python\nimport asyncio\n\nclass ParallelResearchAgent(ADKAgent):\n    async def research_topic(self, topics: list[str]):\n        # Run multiple agents concurrently\n        tasks = [\n            self.specialized_agent(topic)\n            for topic in topics\n        ]\n        results = await asyncio.gather(*tasks)\n        return self.synthesize(results)\n```\n\n### ? Evaluation & Monitoring\n\n**Built-in Evaluation:**\n```python\nfrom google.cloud.aiplatform import agent_evaluation\n\n# Define evaluation metrics\neval_config = agent_evaluation.EvaluationConfig(\n    metrics=[\"accuracy\", \"relevance\", \"safety\"],\n    test_dataset=\"gs:\u002F\u002Fbucket\u002Feval_data.jsonl\"\n)\n\n# Run evaluation\nresults = agent.evaluate(eval_config)\nprint(f\"Accuracy: {results.accuracy}\")\nprint(f\"Relevance: {results.relevance}\")\n```\n\n**Cloud Trace Integration:**\n```python\nfrom google.cloud import trace_v1\n\n# Automatic tracing\n@traced_agent\nclass MonitoredAgent(ADKAgent):\n    def run(self, query: str):\n        # All calls automatically traced\n        with self.trace_span(\"retrieval\"):\n            docs = self.retrieve(query)\n\n        with self.trace_span(\"generation\"):\n            response = self.generate(query, docs)\n\n        return response\n```\n\n### ? Security & Best Practices\n\n**1. Service Account Management:**\n```bash\n# Create minimal-permission service account\ngcloud iam service-accounts create agent-sa \\\n    --display-name \"Agent Service Account\"\n\n# Grant only required permissions\ngcloud projects add-iam-policy-binding PROJECT_ID \\\n    --member=\"serviceAccount:agent-sa@PROJECT.iam.gserviceaccount.com\" \\\n    --role=\"roles\u002Faiplatform.user\"\n```\n\n**2. Secret Management:**\n```python\nfrom google.cloud import secretmanager\n\ndef get_api_key():\n    client = secretmanager.SecretManagerServiceClient()\n    name = \"projects\u002FPROJECT\u002Fsecrets\u002Fapi-key\u002Fversions\u002Flatest\"\n    response = client.access_secret_version(name=name)\n    return response.payload.data.decode('UTF-8')\n```\n\n**3. VPC Service Controls:**\n```bash\n# Enable VPC SC for data security\ngcloud access-context-manager perimeters create agent-perimeter \\\n    --resources=projects\u002FPROJECT_ID \\\n    --restricted-services=aiplatform.googleapis.com\n```\n\n### ? Cost Optimization\n\n**Strategies:**\n- Use Gemini 2.5 Flash for most operations\n- Cache embeddings for RAG systems\n- Implement request batching\n- Use preemptible GKE nodes\n- Monitor token usage in Cloud Monitoring\n\n**Pricing Examples:**\n- Cloud Run: $0.00024\u002FGB-second\n- Agent Engine: Pay-per-request pricing\n- GKE: Standard cluster costs\n- Gemini API: $3.50\u002F1M tokens (Pro)\n\n### ? Reference Architecture\n\n**Production Agent System:**\n```\n┌─────────────────┐\n│   Load Balancer │\n└────────┬────────┘\n         │\n    ┌────▼────┐\n    │Cloud Run│ (Agent containers)\n    └────┬────┘\n         │\n    ┌────▼──────────┐\n    │ Agent Engine  │ (Orchestration)\n    └────┬──────────┘\n         │\n    ┌────▼────────────────┐\n    │  Vertex AI Search   │ (RAG)\n    │  Vector Search      │\n    │  Gemini 2.5 Pro     │\n    └─────────────────────┘\n```\n\n### ? Best Practices for Jeremy\n\n**1. Start with Templates:**\n```bash\n# Use Agent Starter Pack templates\nuvx agent-starter-pack create my-agent --template agentic_rag\n```\n\n**2. Local Development:**\n```bash\n# Test locally first\nadk serve --port 8080\ncurl http:\u002F\u002Flocalhost:8080\u002Fquery -d '{\"question\": \"test\"}'\n```\n\n**3. Gradual Deployment:**\n```bash\n# Deploy to dev → staging → prod\nasp deploy --env dev\n# Test thoroughly\nasp deploy --env staging\n# Final production push\nasp deploy --env production\n```\n\n**4. Monitor Everything:**\n- Enable Cloud Trace\n- Set up error reporting\n- Track token usage\n- Monitor response times\n- Set up alerting\n\n### ? Official Documentation\n\n**Core Resources:**\n- ADK Docs: https:\u002F\u002Fgoogle.github.io\u002Fadk-docs\u002F\n- Agent Starter Pack: https:\u002F\u002Fgithub.com\u002FGoogleCloudPlatform\u002Fagent-starter-pack\n- Agent Engine: https:\u002F\u002Fcloud.google.com\u002Fvertex-ai\u002Fgenerative-ai\u002Fdocs\u002Fagent-engine\u002Foverview\n- Agent Builder: https:\u002F\u002Fcloud.google.com\u002Fproducts\u002Fagent-builder\n\n**Tutorials:**\n- Building AI Agents: https:\u002F\u002Fcodelabs.developers.google.com\u002Fdevsite\u002Fcodelabs\u002Fbuilding-ai-agents-vertexai\n- Multi-agent Systems: https:\u002F\u002Fcloud.google.com\u002Fblog\u002Fproducts\u002Fai-machine-learning\u002Fbuild-and-manage-multi-system-agents-with-vertex-ai\n\n## When This Skill Activates\n\nThis skill automatically activates when you mention:\n- Agent development, ADK, or Agent Starter Pack\n- Multi-agent systems or orchestration\n- Containerized agent deployment\n- Cloud Run, GKE, or Agent Engine deployment\n- RAG agents or ReAct agents\n- Agent templates or scaffolding\n- CI\u002FCD for agents\n- Production agent systems\n\n## Integration with Other Services\n\n**Google Cloud:**\n- Vertex AI (Gemini, Search, Vector Search)\n- Cloud Storage (data storage)\n- Cloud Functions (triggers)\n- Cloud Scheduler (automation)\n- Cloud Logging & Monitoring\n\n**Third-party:**\n- LangChain integration\n- CrewAI orchestration\n- Custom tool frameworks\n\n## Success Metrics\n\n**Track:**\n- Agent response time (target: \u003C2s)\n- Evaluation scores (target: >85% accuracy)\n- Deployment frequency (target: daily)\n- System uptime (target: 99.9%)\n- Cost per query (target: \u003C$0.01)\n\n---\n\n**This skill makes Jeremy a Google Cloud agent architecture expert with instant access to ADK, Agent Starter Pack, and production deployment patterns.**\n\n",{"name":5,"description":7},"main","def8531e65114c0fca8fb8551c1871ee0eed705c","skills\u002Fjeremylongshore\u002Fclaude-code-plugins-plus-skills\u002Fagent-sdk-master",24,"https:\u002F\u002Fgithub.com\u002FComeOnOliver\u002Fskillshub","https:\u002F\u002Fskillshub.wtf","https:\u002F\u002Fgithub.com\u002FComeOnOliver","AI & Blockchain Engineer\r\n","2026-06-24 22:07:19",{"source":8,"skillName":5,"commands":61,"prompt":71,"zipAvailable":72},[62,65,68],{"tool":63,"command":64},"npx","npx skills add ComeOnOliver\u002Fskillshub --skill \"Google Cloud Agent SDK Master\"",{"tool":66,"command":67},"bunx","bunx skills add ComeOnOliver\u002Fskillshub --skill \"Google Cloud Agent SDK Master\"",{"tool":69,"command":70},"pnpm","pnpm dlx skills add ComeOnOliver\u002Fskillshub --skill \"Google Cloud Agent SDK Master\"","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FComeOnOliver\u002Fskillshub\u002Fblob\u002Fmain\u002Fskills\u002Fjeremylongshore\u002Fclaude-code-plugins-plus-skills\u002Fagent-sdk-master\u002FSKILL.md",true,[],[75,85,95,103,109,116],{"skillNo":76,"name":77,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":78,"categories":79,"keywords":81,"publishedAt":46},"01M6FKJN13GMG4KY57J3QD8GY5","model-hierarchy-skill","https:\u002F\u002Fgithub.com\u002FComeOnOliver\u002Fskillshub\u002Fblob\u002Fmain\u002Fskills\u002Fzscole\u002Fmodel-hierarchy-skill\u002Fmodel-hierarchy-skill\u002FSKILL.md",[80],{"categoryNo":24,"categoryName":25,"taxonomyType":21,"taxonomyTypeDesc":22},[82,83,84],"model","model hierarchy","model hierarchy skill",{"skillNo":86,"name":87,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":88,"categories":89,"keywords":91,"publishedAt":46},"01M6FKJJJ6GCAMR4PPPRD12M5X","zoom-mcp","https:\u002F\u002Fgithub.com\u002FComeOnOliver\u002Fskillshub\u002Fblob\u002Fmain\u002Fskills\u002Fzoom\u002Fskills\u002Fzoom-mcp\u002FSKILL.md",[90],{"categoryNo":24,"categoryName":25,"taxonomyType":21,"taxonomyTypeDesc":22},[92,93,94],"MCP","zoom","zoom mcp",{"skillNo":96,"name":97,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":98,"categories":99,"keywords":100,"publishedAt":46},"01M6FKJE3KYT3KPEZMEEHYKA6K","zoom-apps-sdk","https:\u002F\u002Fgithub.com\u002FComeOnOliver\u002Fskillshub\u002Fblob\u002Fmain\u002Fskills\u002Fzoom\u002Fskills\u002Fzoom-apps-sdk\u002FSKILL.md",[],[28,101,102,93],"zoom apps sdk","apps",{"skillNo":104,"name":105,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":106,"categories":107,"keywords":108,"publishedAt":46},"01M6FKJ9EDH9Q5VWDF8HAJRSP0","windows","https:\u002F\u002Fgithub.com\u002FComeOnOliver\u002Fskillshub\u002Fblob\u002Fmain\u002Fskills\u002Fzoom\u002Fskills\u002Fwindows\u002FSKILL.md",[],[105],{"skillNo":110,"name":111,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":112,"categories":113,"keywords":115,"publishedAt":46},"01M6FKJ666BEEFC3AE9GJ72Q93","whiteboard","https:\u002F\u002Fgithub.com\u002FComeOnOliver\u002Fskillshub\u002Fblob\u002Fmain\u002Fskills\u002Fzoom\u002Fskills\u002Fwhiteboard\u002FSKILL.md",[114],{"categoryNo":24,"categoryName":25,"taxonomyType":21,"taxonomyTypeDesc":22},[111],{"skillNo":117,"name":118,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":119,"categories":120,"keywords":121,"publishedAt":46},"01M6FKJ21WFCDGTJQ2Q91T9DVW","websockets","https:\u002F\u002Fgithub.com\u002FComeOnOliver\u002Fskillshub\u002Fblob\u002Fmain\u002Fskills\u002Fzoom\u002Fskills\u002Fwebsockets\u002FSKILL.md",[],[118]]