Cross-Cloud Agentic Analytics Architect
Design governed, secure agentic analytics for distributed data
The workflow requires requirements discovery, phase approvals, permission before validation, and permission before writing deployment files; it also calls for role-based security, credentials, and sensitive-data governance. Deducted 9 points because least-privilege role mappings, credential lifecycle controls, complete data-flow disclosure, rollback, external-service data handling boundaries, and dependency security review are not specified.
The four-phase workflow, phase-skipping rules, and contradiction-blocking rules are reasonably explicit, and ambiguous requirements must be clarified. Deducted 3 points because it depends on multiple external MCP servers, documents, and unprovided skills; validation scripts are generated only at runtime, with no committed tests, deterministic reproduction procedure, or concrete failure diagnostics. Static-review cap is 10.
The target scenario, discovery inputs, staged outputs, and approval triggers are fairly clear for multi-cloud, hybrid-cloud, structured, and unstructured data. Deducted 6 points because non-fit boundaries, exclusion triggers, Chinese-language support, and mainland-China reachability are not addressed; core grounding depends on external Google services, which may limit availability for the stated audience.
The documentation has phased information architecture, an output template, supporting references, product-renaming guidance, alternatives, and Apache-2.0 licensing; the README provides installation, active-development, and issue-reporting paths. Deducted 6 points because the skill lacks its own version, changelog, named maintenance owner, dependency installation notes, FAQ, troubleshooting, and resource-availability guidance.
The workflow covers requirements, architecture, design, deployment, validation, and packaging, with specified Markdown, Mermaid, and deployment outputs. Deducted 1 point because no implementation code or validation artifacts are committed; result quality depends on external sources, repeated user approval, and runtime generation, while direct usability and comparative benefit over manual alternatives are not file-verified. Static-review cap is 7.
The skill names many official documents, MCP resources, and citation requirements, and its output template requires supporting citations. Deducted 1 point because the selected skill has no test suite, CI coverage, execution records, or independent corroboration demonstrating key paths; static-review cap is 5.
- Do not treat this skill as a verified deployment-automation solution; deployment steps, permissions, API availability, product status, and generated code require manual review and isolated validation.
- Before using external Google MCP services, documentation, and cross-cloud connectivity, confirm network reachability, data residency, cross-border compliance, credential scope, and costs.
- Add explicit least-privilege IAM mappings, secret/token management, failure rollback, resource cleanup, and reproducible tests.
What does this skill do, and when should you use it?
This skill guides users in designing agentic analytics solutions for structured and unstructured data distributed across Google Cloud, other cloud providers, and on-premises systems. It follows four phases: requirements discovery, solution architecture, solution validation, and packaging. The workflow covers technical decomposition, product selection, Mermaid diagrams, architecture descriptions, design recommendations, deployment guidance, and validation planning. During architecture work, it instructs the agent to use Google Cloud documentation, a developer knowledge MCP service, and related skills.
It asks one question at a time about data sources, hosting locations, metadata federation, analytical needs, natural-language prompts, non-functional requirements, existing multi-cloud or on-premises architecture, and dependencies. It checks for ambiguities and contradictions and pauses architecture recommendations until they are resolved. After technical-decomposition approval, it organizes the solution into user interaction, grounding and trusted data, metadata curation, and data processing and analytics layers; identifies Google Cloud products and features; and produces a Mermaid diagram, architecture description, design recommendations, deployment prerequisites, and deployment instructions. It also creates a validation plan, generates curl or gcloud commands, and runs validation only after permission is granted. Finally, it can consolidate the artifacts into solution-architecture-guide.md and request permission to write code files.
- A data architect needs to analyze inventory data spread across AWS S3, Azure Blob, Google Cloud Storage, databases, and on-premises systems.
- A data team must combine structured sources such as Iceberg data with unstructured sources such as PDF recipes and invoices.
- An enterprise needs a governed, secure architecture with role-based security for agentic analytics across cloud providers or on-premises environments.
- A data science team needs to plan large-scale joins, cleaning, and forecasting workflows accessed through natural-language prompts in an agentic IDE.
- A platform team needs deployment prerequisites, validation plans, and gcloud or curl commands after an architecture has been approved.
What are this skill's strengths and limitations?
- Covers a complete workflow from requirements discovery through validation and delivery.
- Requires ambiguities to be resolved before architecture generation and includes role-based security and credentials in the decomposition.
- Targets distributed structured and unstructured data across clouds and on-premises systems.
- Produces Mermaid diagrams, deployment guidance, validation commands, and a consolidated Markdown guide.
- The repository is Apache-2.0 licensed and can be installed with npx skills add google/skills.
- The SKILL.md does not contain complete product-selection details, implementation code, or a fixed architecture template.
- Architecture phases depend on the Google Developer Knowledge MCP service, official documentation, and other external resources.
- The workflow requires repeated user approvals and does not document an automated approval mechanism.
- No test suite, version compatibility matrix, or verified-platform list is provided.
- Cloud costs, exact IAM roles, and deployment parameters must be determined for the workload; the source gives no universal values.
How do you install this skill?
Run npx skills add google/skills and select the target skill from the repository. Its source path is skills/cloud/google-cloud-solution-agentic-analytics-spark-knowledge-catalog. The source does not document version pinning, a separate installation command, or runtime configuration.
How do you use this skill?
After installing it in an Agent Skills-compatible client, use a prompt such as “Design a governed agentic analytics solution for structured and unstructured data distributed across AWS S3, Google Cloud Storage, and on-premises databases.” Provide requirements phase by phase and approve the technical decomposition, product recommendations, diagram, architecture description, design recommendations, and deployment guidance as requested. Explicit permission is required before validation execution or writing code files.