Google Cloud Borderless Lakehouse Architect
Designs governed lakehouses that connect multicloud data to AI agents.
The workflow requires requirements confirmation, architecture review, deployment review, and validation, and it includes IAM, Secret Manager, credential vending, sensitive-data protection, and private-network guidance. However, it does not specify concrete least-privilege roles, complete data-flow disclosure, rollback procedures, or explicit security confirmation before executing generated IaC/scripts, so points are deducted.
The four-phase workflow and required artifacts are broadly consistent and include dry-run, connectivity, and security-policy checks. However, there are no committed tests, reproducible examples, dependency-availability evidence, or diagnostic failure handling for abnormal inputs, so the static score remains modest.
The audience, supported borderless lakehouse scenarios, discovery questions, Mermaid diagram requirements, and Markdown outputs are reasonably clear, and non-fit cases are stated. Chinese-language interaction is not addressed, and the core workflow depends on Google Cloud and external documentation services without mainland-China reachability or fallback guidance, so points are deducted.
The material is organized progressively by phase and supplies a template, references, renaming guidance, product mapping, and design recommendations. Repository context provides Apache-2.0 licensing, installation, issue reporting, and active-development signals. The skill itself lacks versioning policy, changelog, explicit maintenance ownership, FAQs, a dependency list, and a clear known-limitations section, so it is not full marks.
The skill defines outputs covering requirements, product mapping, architecture diagrams, an architecture guide, IaC, deployment instructions, and validation reporting, making the core design task plausible. However, there is no completed representative output or execution evidence; users must still review, fill in environment-specific details, and run the generated artifacts, so the static score is conservative.
Several official Google Cloud documentation links, a template, and repository governance materials provide some traceability for key guidance. There is no CI, committed test suite covering key paths, independently reproducible result, or third-party corroboration, so the score remains below the static ceiling.
- Do not apply generated Terraform, deployment scripts, or validation commands directly to production; review IAM, networking, credentials, data egress, and destructive changes first, and define rollback procedures.
- The core workflow depends on Google Cloud services and external Google documentation; mainland-China deployments require separate validation of reachability, product availability, regional restrictions, and fallbacks.
- The skill lacks comprehensive error handling, test coverage, changelog detail, and demonstrated successful cases, so outputs require human technical review and execution-based validation.
What does this skill do, and when should you use it?
This skill guides users through designing a secure, governed open data lakehouse for cross-cloud, hybrid, and on-premises data, with agentic AI integration. It covers requirements discovery, solution design, implementation planning, and validation. Its outputs can include a technical decomposition, Google Cloud product mapping, Mermaid architecture diagram, Markdown architecture guide, IaC, deployment instructions, and validation scripts. It is intended for complex multi-product data architectures, not simple single-cloud warehouses or non-AI workloads.
Collects requirements about data sources, metadata federation, security and credentials, cross-environment analytics, and natural-language queries; identifies and decomposes hybrid, borderless, and on-premises components; retrieves specified Google Cloud documentation, maps components to products, and creates a Mermaid diagram separating ingestion and serving while showing Managed Service for Apache Spark as a shared bridge; generates solution-architecture-guide.md, Terraform or similar IaC, deployment instructions, validation checks, and lightweight scripts; requests confirmation at key stages and iterates on user feedback.
- A cloud architect needs to connect data silos across Google Cloud, another cloud, and on-premises systems to AI agents.
- A data engineering team needs a governed, secure design for joining and transforming distributed multicloud data.
- An AI or analytics team needs to plan natural-language agent workflows over federated data sources.
- A platform team needs generated IaC, deployment guidance, and validation procedures for a multi-product Google Cloud data solution.
What are this skill's strengths and limitations?
- Provides an end-to-end workflow from requirements discovery through validation.
- Explicitly covers multicloud, hybrid, on-premises, and federated-query scenarios.
- Requires specified Google Cloud documentation to ground design and implementation guidance.
- Defines deliverables including Mermaid diagrams, IaC, deployment instructions, and validation materials.
- The SKILL.md contains no actual Terraform, scripts, test suite, or sample output.
- The client needs search or fetch capability to retrieve the referenced Google Cloud resources.
- Product choices, permissions, APIs, and deployment details must be generated interactively; the skill file alone does not deploy a solution.
- It explicitly excludes simple single-cloud data warehouses and non-AI workloads.
How do you install this skill?
Run npx skills add google/skills, then select google-cloud-solution-agentic-ai-borderless-data-lakehouse during installation. The README does not document a local destination directory or a separate installation command for this skill.
How do you use this skill?
In an Agent Skills-compatible client, provide a concrete request such as: “Design a governed open data lakehouse for an agentic analytics system connecting Google Cloud, another cloud provider, and on-premises data.” Supply data-source, metadata, security, analytics, and query requirements, then confirm or revise the generated decomposition, architecture, and implementation plan.