How to use the Agentic Analytics Solution Designer agent skill

What is an agent skill and what does this one do?

An agent skill is a folder of instructions and supporting files that an AI agent loads when your request matches the skill's description. This skill is built around a SKILL.md file plus reference documents and a report template.

When installed, the agent follows a structured four-phase workflow: requirements discovery, solution architecture, solution validation, and solution packing and presentation. It asks questions about your data sources, security and compliance needs, and dependencies, then proposes an architecture, a validation plan with test scripts, and a consolidated solution guide.

• Phase 1: Requirements discovery, asking one question at a time.
• Phase 2: Solution architecture, built from the gathered requirements.
• Phase 3: Solution validation, with a plan, instructions and scripts, then a run.
• Phase 4: Solution packing and presentation, compiling the solution content.

Install the skill from the downloaded ZIP

Download the ZIP of the skill from this page and unzip it. Place the resulting skill folder in one of the following locations, depending on how you want to use it.

For a personal install available across your projects, use ~/.claude/skills/. For a single project, use the .claude/skills/ folder inside that project.

If you use the Claude apps instead, you can upload the skill under Settings → Capabilities → Skills.

• Personal: ~/.claude/skills/
• Project only: .claude/skills/ inside your project folder
• Claude apps: Settings → Capabilities → Skills, then upload

Check the requirements before you start

The skill is for design and validation, so some parts need a real Google Cloud environment. Validation commands need a Google Cloud account and either the gcloud CLI or REST API access. You also need access to Knowledge Catalog and Managed Service for Apache Spark.

If you want the agent grounded in Google Cloud Data Agent Kit, you need an agentic IDE such as VS Code or Antigravity.

• Google Cloud account, with gcloud CLI or REST API access for validation commands
• Access to Knowledge Catalog and Managed Service for Apache Spark
• An agentic IDE such as VS Code or Antigravity, if you use the Data Agent Kit

Trigger the skill with a natural request

Ask for help designing a governed analytics solution in plain language. The skill's description covers designing data science and analytics workflows across structured and unstructured data, including data in AWS S3, Azure Blob, AlloyDB and Iceberg, and establishing metadata governance with Knowledge Catalog aspect types.

Expect the agent to ask questions before it recommends anything. This is intentional: during requirements discovery it does not propose architectures, services or component mappings until ambiguities are resolved.

• "Help me design a governed analytics solution that joins Iceberg tables in S3 with AlloyDB data."
• "Plan the metadata governance for our analytics data using Knowledge Catalog aspect types."
• "Walk me through designing a Spark-based forecasting pipeline with requirements first."

Work through the requirement questions

The agent asks about your primary data sources, where they are hosted, how metadata is managed across clouds, your analytical needs, and the natural language prompts your team will run in its IDE. It then covers non-functional needs such as security and compliance, reliability, performance, operations, and cost.

If your workload already runs on other clouds or on-premises, the agent asks you to describe that architecture. If it finds contradictions, for example between zero-copy access and copying data, it explains the conflict and asks how you want to resolve it. You can delegate the decision, and the agent then proposes an option and asks you to approve it.

Review the architecture, validation plan and solution guide

Once requirements are settled, the agent builds a technical decomposition in four layers: user-interaction (IDE), grounding and trusted data, metadata curation, and data processing and analytics. The decomposition must address role-based security and credentials.

Validation then produces a plan, instructions and test scripts. Before any gcloud, REST API or MCP command runs, the skill requires the agent to follow its usage attribution instructions, which tag commands with a label that identifies only the skill and its version. The final output follows an assets/output-template.md structure with sections for executive summary, requirements, and architecture.