Guide · Agent skill
How to use the Agentic Data Science Architecture Designer agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
This agent skill helps an AI agent design a multi-product Google Cloud solution for an agent-based data analytics or machine learning project. Instead of guessing your requirements, the agent asks clarifying questions, waits for your answers and confirmation, and then produces a solution architecture guide with a Mermaid diagram for your approval. This tutorial explains what the skill does, how to install it, how to trigger it, and what to check before you rely on its output.
Agentic Data Science Architecture Designergoogle/skills · Apache-2.0 Download the skill Quick start
- 1Download and unzip the skill. Download the skill ZIP from this page and unzip it so the folder contains SKILL.md.
- 2Place it in the skills folder. Move the folder to ~/.claude/skills/ for personal use or to .claude/skills/ inside a project. Alternatively, upload it under Settings → Capabilities → Skills in the Claude apps.
- 3Make sure tooling is available. Confirm the agent can search or fetch Google Cloud documentation, and that gcloud, REST API, or MCP tools are available if you want Google Cloud commands run.
- 4Send a design request. Describe your agent-based data analytics or ML workload on Google Cloud and ask for an architecture design.
- 5Answer the clarifying questions. Reply to the agent's questions about data, users, query types, and constraints. The agent waits for your answers before continuing.
- 6Confirm the decomposition. Review the technical decomposition and confirm it, or request changes until you explicitly approve it.
- 7Review the solution architecture guide. Check the solution-architecture-guide.md draft and the Mermaid diagram, then approve or request changes before any implementation plan is generated.
What is an agent skill and what does this one make the agent do?
An agent skill is a folder containing a SKILL.md file with instructions that an AI agent loads when a task matches the skill's description. This skill, named google-cloud-solution-agentic-ai-data-science-workflow, is for designing (and planning the implementation of) a tailored multi-product solution on Google Cloud for an agent-based data analytics or ML workload.
The workflow has four phases: requirements discovery and analysis, solution design, implementation plan, and solution validation. The agent proposes a coordinator design pattern as the primary option, maps components to Google Cloud products, drafts a Mermaid architecture diagram, and writes a solution-architecture-guide.md file based on the included template.
- Phase 1: asks clarifying questions about data sources, end users, expected queries, and constraints, then waits for your answers and confirmation.
- Phase 2: retrieves referenced Google Cloud documentation, selects a design pattern, maps products, drafts the diagram and design recommendations, and waits for your approval.
- Phase 3: generates an implementation plan, prerequisites, and Infrastructure as Code such as Terraform.
- Phase 4: validates that the deployment meets the workload requirements.
How do you install the skill?
Download the skill ZIP from this page and unzip it. For a personal install, place the unzipped folder in ~/.claude/skills/. For a single project, place it in that project's .claude/skills/ folder.
You can also upload the skill in the Claude apps: open Settings, go to Capabilities, then Skills, and upload it there.
- Personal install: ~/.claude/skills/
- Project install: .claude/skills/ inside your project
- Claude apps: Settings → Capabilities → Skills
What do you need before you start?
The skill needs access to Google Cloud documentation through search or fetch tools. It also expects gcloud, REST API, or MCP tooling if the agent runs Google Cloud commands. The skill does not run scripts, but it does use the network.
Before the agent executes any gcloud, REST API, or MCP command, the skill requires it to read references/attribution.md. That file asks the agent to tag commands with a usage attribution value that identifies only the skill and its version.
- Access to Google Cloud documentation via search or fetch tools
- gcloud, REST API, or MCP tooling if you want the agent to run Google Cloud commands
- Network access (the skill uses the network; it runs no scripts)
How do you trigger the skill with a natural request?
Describe a multi-agent data analytics or ML system on Google Cloud and say you want it designed or architected. The skill's description matches requests about architecting multi-product solutions for agent-based data analytics or ML workloads.
Expect the agent to ask questions first. Answer them fully, because the agent will not proceed to component identification until you respond. Then confirm the component decomposition, and review the architecture guide before approving Phase 3.
- Example request: "Design a Google Cloud architecture for a multi-agent system that answers analyst questions over our sales data."
- Example request: "Help me plan an agentic ML system on Google Cloud with a design document and diagram before any deployment."
- Example request: "Architect a multi-agent data science workflow on Google Cloud and ask me clarifying questions first."
What does a typical session look like?
Plan on several back-and-forth turns. The agent asks about data sources and types, end users and network access, expected query types, and performance, security, or governance constraints. After you confirm the technical decomposition, it drafts the solution architecture guide with requirements, product mapping, a Mermaid diagram, and design recommendations.
Design recommendations in the skill include least-privilege IAM, using the Cloud Data Loss Prevention API to discover and de-identify sensitive data, and securing frontend Cloud Run endpoints by disabling default run.app URLs and using a regional external Application Load Balancer with Google Cloud Armor.
What should you check and watch out for?
The output is a design draft, not verified production infrastructure. The skill's validation phase is only as reliable as the agent's execution of it, so review every product choice and command yourself.
Google Cloud product names change. The skill tells the agent to check current documentation, and notes that underlying APIs, Terraform resources, and IAM roles may keep legacy identifiers. For example, Vertex AI is listed as Gemini Enterprise Agent Platform, and Vertex AI Agent Engine as Agent Runtime.
The skill's own description says it should not be used for simple queries, non-agentic pipelines, general cloud reviews, or writing agent code.
- Review the generated gcloud commands before running them.
- Check that the product names match current Google Cloud documentation.
- Do not treat the Terraform or deployment steps as production-ready without your own testing.
What people use it for
- Planning a multi-agent analytics assistant that queries company data for analysts and needs a documented architecture before build work starts.
- Designing an agentic ML system on Google Cloud with a coordinator agent and specialist tools, with a diagram for stakeholder review.
- Producing a structured design document with requirements, product mapping, and security recommendations for an architecture review meeting.
- Drafting an implementation plan and Infrastructure as Code outline after the design has been approved.
Frequently asked questions
Is the Agentic Data Science Architecture Designer skill free?
The skill is published under the Apache-2.0 licence in the google/skills repository. Using it does not require a separate purchase, but any Google Cloud resources you deploy are billed under your own Google Cloud account.
What do I need to install this agent skill?
Download the ZIP from this page, unzip it, and place the folder in ~/.claude/skills/ or a project's .claude/skills/ folder. You can also upload it in the Claude apps under Settings → Capabilities → Skills.
Does this skill run scripts or use the network?
According to the listing, the skill does not run scripts but does use the network. It relies on search or fetch tools to read Google Cloud documentation.
Do I need gcloud installed to use it?
Not to get a design. The skill expects gcloud, REST API, or MCP tooling only if the agent runs Google Cloud commands. When it does, the skill requires the agent to tag those commands with an attribution value first.
How is this different from asking a chatbot to design a cloud system?
This skill imposes a fixed workflow: it asks questions, waits for your answers and confirmation, proposes a design pattern, and produces a structured guide with a diagram for approval. It also steers the agent toward current Google Cloud product names.
When should I not use this skill?
Skip it for simple one-off data questions, single queries, non-agentic pipelines, or general cloud reviews. The multi-phase workflow is heavy for those tasks, and the skill itself says it should not be used for them.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if you are an architect or engineer planning a multi-agent data science or ML system on Google Cloud and want a structured design with a diagram and sign-off steps before any deployment work. Pick something else if you need a quick answer to a data question, a general cloud review, or finished production infrastructure, because the output is a design draft that needs your own verification.
Download the skill