Google Cloud Agentic Data Science Architect
Design and validate a multi-product Google Cloud architecture for agent-based data analytics and machine learning workloads.
The skill requires requirements confirmation, decomposition confirmation, and approval of architecture and deployment plans. It recommends least-privilege IAM, Secret Manager, DLP, Model Armor, private networking, and sandboxed code execution, providing basic data-flow and sensitive-data safeguards. Points are deducted because concrete permission scopes, retention and cross-border data handling, per-change confirmation before IaC execution, rollback procedures, and dependency-integrity checks are unspecified; generated deployment actions may affect external infrastructure.
The four-phase workflow, prerequisites, validation checks, timeouts, retries, failover, and error-handling recommendations make the happy path reasonably coherent. Points are deducted because there is no committed test suite or reproducible execution evidence, several steps depend on external documents and tools, and diagnosis guidance is thin for invalid input, missing resources, unreachable documentation, or deployment failures.
The skill clearly targets multi-product Google Cloud architectures for agent-based data analytics or machine-learning workloads and states non-fit cases such as simple queries, non-agent pipelines, general cloud reviews, and agent coding. Points are deducted because input/output boundaries, trigger rules, and non-fit ranges remain broad, Chinese-language support is absent, and reachability in mainland China is not addressed despite dependence on Google Cloud, GitHub, external documentation, and the MCP ecosystem.
The documentation is phase-oriented and includes a template, product-mapping guidance, design recommendations, installation information, support and contribution paths. The repository states Apache-2.0 licensing, official organizational provenance, and active development. Points are deducted because the skill lacks its own version, changelog, explicit maintainer, or update commitment, and provides limited compatibility-maintenance guidance for referenced paths, product renaming, and external resources.
The skill covers discovery, architecture design, implementation planning, and validation, and specifies a structured solution-architecture-guide.md output. Under static calibration, this supports the core-task score ceiling. Points are deducted because IaC, deployment scripts, and verification scripts are requested outputs rather than verified assets; product decisions depend on external material, and results may require substantial human review and environment adaptation.
Evidence includes a fixed revision, repository templates, product-mapping and design guidance, and references to official or relevant technical materials, giving limited auditability. Points are deducted because the supplied files contain no CI, test suite, execution logs, or third-party reproduction results, so key product claims and naming mappings cannot be independently verified from the files alone.
- Before running generated Terraform, deployment commands, or verification scripts, review IAM, networking, cost, and resource changes individually and establish a recoverable rollback plan.
- Verify the current validity and mainland-China reachability of the external Google Cloud documentation, GitHub samples, MCP Toolbox, and product-renaming information.
- The skill provides no verified IaC, deployment scripts, or test results; its outputs should not be treated as deployed or security-validated.
What does this skill do, and when should you use it?
This skill is for users planning agentic data science solutions on Google Cloud. It covers requirements discovery, technical decomposition, solution design, implementation planning, and deployment validation. Possible outputs include product mappings, Mermaid architecture diagrams, a Markdown architecture guide, infrastructure as code, deployment instructions, and verification scripts. It is not intended for simple queries, non-agentic pipelines, general cloud reviews, or writing agent code.
It asks about data sources, users, access models, query types, and performance, security, and governance constraints; analyzes components and relationships; selects an agentic design pattern; maps components to Google Cloud products and features; creates a Mermaid architecture diagram and a Markdown file named solution-architecture-guide.md; plans infrastructure as code such as Terraform and deployment steps; and defines validation checks and scripts using examples such as terraform plan, curl, or gcloud. The skill also directs the agent to consult specified Google Cloud documentation, sample code, deployment resources, and MCP Toolbox configuration documentation.
- A data science team needs a Google Cloud architecture for a natural-language analytics system spanning multiple data sources.
- A cloud architect needs to decompose an agent-based machine learning workload and select matching Google Cloud products.
- A platform engineer needs to turn an approved agentic data science design into Terraform and deployment instructions.
- A security or operations team needs to verify network routing, service endpoints, and IAM restrictions after deployment.
What are this skill's strengths and limitations?
- Covers the workflow from requirements discovery through deployment validation.
- Requires confirmation of the technical decomposition before solution design.
- Produces concrete artifacts, including Mermaid diagrams, a Markdown architecture guide, IaC, deployment instructions, and validation scripts.
- Targets multi-product agentic data analytics and machine learning workloads on Google Cloud.
- Its scope is limited to multi-product architectures for agent-based data analytics or machine learning.
- It relies on external Google Cloud documentation, examples, and deployment resources for design and implementation guidance.
- The source provides no test suite, concrete Terraform modules, or complete ready-to-deploy implementation.
- User approval is required at several stages, which can lengthen the interaction.
How do you install this skill?
Run npx skills add google/skills, then select google-cloud-solution-agentic-ai-data-science-workflow during installation. The source does not document a specific installation directory or a separate command for installing only this skill.
How do you use this skill?
In an Agent Skills-compatible client with the skill installed, provide a concrete request such as: “Design a Google Cloud agentic data science workload that accesses multiple data sources, supports natural-language analytics, and meets enterprise security and governance requirements.” Prepare the data sources, target users, network access model, expected queries, and performance, security, and governance constraints. The source requires user confirmation at key workflow stages but does not define a client-specific trigger syntax.
How does this skill compare with similar options?
The source names single-agent, sequential or parallel, and review-and-critique patterns as alternatives, but it does not compare this skill with specific competing products or repositories.