What does this skill do, and when should you use it?
This skill helps users design a conversational search solution for private enterprise content on Google Cloud. It uses AlloyDB to store and index embedding vectors, Cloud Storage for source content, and GKE to host the application components, an open model, and an open-source inference framework. Its workflow covers requirements discovery, architecture, validation, and solution packaging. It is not intended for fully managed RAG, SaaS search services, or non-SQL vector databases.
It asks structured functional and non-functional requirements questions, checks for ambiguities and contradictions, and waits for approval of a technical decomposition before producing architecture work. After approval, it can generate product recommendations, a Mermaid architecture diagram, an architecture description, design recommendations, and deployment guidance including Terraform and Ray-on-GKE or LangChain deployment steps. It also creates a validation plan covering deployment previews, connectivity, vector indexes, embedding pipelines, retrieval latency and accuracy, and security policies, then can run checks after receiving permission.
- A cloud architect designing conversational RAG search over private enterprise content.
- An engineering team choosing AlloyDB for vector storage and indexing while running application components on GKE.
- A technical lead planning ingestion, chunking, embedding, hybrid retrieval, prompt augmentation, and response generation as one workflow.
- A user who has already completed an earlier phase and needs the next architecture, deployment, or validation artifact.
How do you install this skill?
- The core workflow depends on the external Google Developer Knowledge MCP, online documentation, and other skills; the material does not establish reachability or version stability in the target environment.
- The skill may generate Terraform, deployment commands, and validation scripts, but it provides no pinned versions, least-privilege examples, rollback steps, or secure defaults; review is required before execution.
- The citation requirement is not aligned with the output template, which only reserves a References section and does not define per-claim citation or fact-checking procedures.
- The documentation mixes AlloyDB with a Cloud SQL architecture reference; this potential inconsistency should be resolved before product recommendations are accepted.
- Shell / CLI
- Network access
- Local filesystem
- MCP Server
Terraformcurlgcloud CLIGoogle Developer Knowledge MCP server
Run npx skills add google/skills, then select google-cloud-solution-rag-enterprise-search-gke-sqldb from the repository. The source does not document the exact installed folder location.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/google/skills.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/skills/cloud/google-cloud-solution-rag-enterprise-search-gke-sqldb" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
In an Agent Skills-compatible client, ask for an enterprise RAG search design using GKE, AlloyDB, and Cloud Storage, and state which workflow phases are already complete if applicable. During requirements discovery, the skill asks questions first; it will not begin architecture design until the technical decomposition is explicitly approved.
What are this skill's strengths and limitations?
- Covers a workflow from requirements discovery through validation and packaging.
- Targets a clearly defined combination of AlloyDB, Cloud Storage, GKE, an open model, and an open-source inference framework.
- Includes approval gates for requirements resolution, technical decomposition, architecture artifacts, and validation.
- Can produce Mermaid diagrams, Terraform code, deployment guidance, and validation commands or scripts.
- Out of scope for fully managed RAG and SaaS search services.
- Requires a vector-enabled SQL database; non-SQL vector databases are excluded.
- The source provides no test suite, sample outputs, or platform-specific testing evidence.
- Deployment and validation may require Terraform, curl, gcloud, and access to the Google Developer Knowledge MCP server; exact permissions are not specified.
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| GKE Enterprise RAG Search Architect this page ✓ Google · Official | 53 · Use with care | ★ 21k | 12d ago | Apache-2.0 |
| Google Cloud Live Multimodal Streaming Architect ✓ Google · Official | 55 · Use with care | ★ 21k | 12d ago | Apache-2.0 |
| GKE Production Golden Path ✓ Google · Official | 54 · Use with care | ★ 21k | 12d ago | Apache-2.0 |
| Google Cloud AI Agent Builder ✓ Google · Official | 53 · Use with care | ★ 21k | 12d ago | Apache-2.0 |
| AI Workload Migration to GKE Inference ✓ Google · Official | 50 · Use with care | ★ 21k | 12d ago | Apache-2.0 |
How did FollowSkills review this skill?
The skill requires requirements clarification, phase approvals, validation permission, and consent before writing files, while mentioning private networking, encryption, access control, and sensitive-data protection. It does not define least-privilege permissions, concrete data-flow disclosure, credential handling, dependency isolation, or rollback, so points are deducted.
The phase ordering, approval gates, and ambiguity-resolution rules are reasonably clear. However, there are no executable scripts, test suites, pinned dependencies, or detailed failure-diagnosis paths, and reachability of the external MCP and documentation is not established by the static material; the score is conservatively limited.
The intended audience, scenario, and exclusions are explicit: SQL vector storage, open models, and GKE. Multiple approval checkpoints support controlled use. Boundaries for non-fit cases, Chinese-language interaction, and mainland-China network reachability are not documented, so points are deducted.
The information architecture, workflow phases, supporting references, output template, Apache-2.0 license, and Google repository provenance are clear. The skill lacks its own version, changelog, named maintenance responsibility, dependency-installation guidance, FAQ, and systematic limitation disclosure, so points are deducted.
The intended artifacts cover requirements, architecture, diagrams, design, deployment, validation, and packaging, so the core task is plausible. The template remains mostly placeholders and Terraform, scripts, and representative outputs are absent; substantial generation and review would still be required, so the static score is capped at7.
The skill identifies official documentation, related skills, and an MCP source and requires citations for generated guidance. The supplied material contains no actual citations, test results, CI evidence, or independently reproducible outputs, and several sources require runtime access, so only limited evidence credit is justified.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →