k8s-review

Reviews Kubernetes manifests, Helm charts, and RBAC for correctness, security, and best-practice issues.

3|Updated Jul 2, 2026
One-click install
npx skills add https://github.com/ArangoGutierrez/claude-toolkit --skill k8s-review-arangogutierrez
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: k8s-review
Source: https://github.com/ArangoGutierrez/claude-toolkit/tree/main/.claude/skills/k8s-review
Command: npx skills add https://github.com/ArangoGutierrez/claude-toolkit --skill k8s-review-arangogutierrez

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Kubernetes pull requests often ship subtle YAML coercion bugs, Helm templating errors, deprecated apiVersions, and over-privileged RBAC that linters miss and reviewers overlook. This Skill runs a systematic Kubernetes-specific review that flags correctness, security, and best-practice issues with file:line precision before human review. ## Core Features & Use Cases - Static analysis: Validates changed manifests with kubeconform or a kubectl server dry-run, lints and renders Helm charts with default and edge values files, and runs kube-linter when available — each tool guarded so the review still works with none installed. - Structured checklist review: Walks a four-section checklist covering YAML correctness (Norway problem, octal modes, duplicate keys), Helm correctness (nindent, quote, selector immutability), K8s API best practices (probes, securityContext, PDBs), and RBAC least-privilege (wildcards, cluster-wide secrets, Role vs ClusterRole). - Findings reporting: Reports each issue as file:line with a category (correctness/security/best-practice), severity (must-fix/should-fix/consider), and the checklist item that flagged it; supports a dispatched mode where a pr-review subagent returns findings as structured data. - Use Case: Before requesting review on a PR that changes a Helm chart and its RBAC Role, run the review to catch an unquoted {{ .Values.enabled }} that renders as a bare yes, a wildcard verb in the Role, and a missing PodDisruptionBudget — each reported with severity and a suggested fix. ## Quick Start Ask the AI to review the Kubernetes manifests and Helm chart changed in this PR for correctness, security, and RBAC least-privilege issues.

Frequently Asked Questions about k8s-review

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I review Kubernetes manifests before a pull request?▼

Run a Kubernetes manifest review that validates changed YAML with kubeconform or a kubectl server dry-run, then walks a checklist covering YAML coercion traps, API deprecations, probes, and securityContext. Findings are reported as file:line with category and severity.

How to check a Helm chart for templating errors?▼

Lint and render the chart with helm lint and helm template using default values plus every values-*.yaml file in the PR. Then check for unquoted values, indent vs nindent misuse, missing required/default functions, and immutable selector labels templated from mutable values.

What tools are needed for Kubernetes static analysis?▼

kubeconform (or a kubectl cluster context for server dry-run) validates manifests against the schema, helm lints and renders charts, and kube-linter adds extra checks. Each tool is optional and guarded by command -v, so the checklist review still runs with none installed.

Does this review check RBAC least-privilege issues?▼

Yes, the RBAC checklist flags wildcard verbs/resources/apiGroups, escalate/bind/impersonate verbs, cluster-wide secrets access, ClusterRole overuse, default ServiceAccount bindings, and aggregated ClusterRole effects. Wildcards and cluster-wide secrets access are must-fix findings.

When should I not use a Kubernetes review checklist?▼

Skip it for YAML style and formatting nits like indentation width or quote style, which yamllint and CI already handle. It also does not redesign charts or the API surface — it flags issues within the existing pattern rather than proposing rewrites.