Kubernetes Config Error Detective is a skill for diagnosing Kubernetes deployment failures caused by YAML and configuration errors. It targets DevOps engineers, SREs, platform teams, cloud engineers, and AI coding agents who need to understand why a deployment is broken without making live cluster changes.

The skill audits a wide range of configuration inputs: raw manifests, Helm values files, rendered Helm charts, Kustomize overlays, GitOps sync error output, deployment events, and sanitized logs. It also inspects Services, Ingresses, liveness and readiness probes, resource requests and limits, image references, ConfigMap and Secret references, RBAC policies, storage configuration, scheduling constraints, and networking settings.

From that input it identifies likely root causes for common failure modes including CrashLoopBackOff, ImagePullBackOff, Pending Pods, missing Service endpoints, 503 errors, readiness failures, missing configuration references, wrong label selectors, wrong port mappings, misconfigured probes, resource constraint violations, and Helm values mismatches.

Output takes the form of evidence-based diagnosis reports, safe YAML patch suggestions, validation checklists, rollback notes, production readiness reviews, and PR review comments. Because the skill produces analysis and suggested patches rather than executing changes, it can be used safely in any environment without cluster credentials or direct API access.

Use cases

Diagnose a CrashLoopBackOff by submitting the pod manifest and sanitized logs for root cause analysis

Identify why a Service has no endpoints by auditing label selectors and port definitions

Review Helm values and rendered charts for mismatches before a production rollout

Audit Kustomize overlays and GitOps sync errors to find divergence between intended and applied config

Get safe YAML patch suggestions for fixing misconfigured liveness or readiness probes

Run a production readiness review against manifests covering RBAC, resource limits, and networking

When to use it

When a Kubernetes deployment is failing and the likely cause is a YAML or configuration mistake

When reviewing PRs that touch manifests, Helm charts, or Kustomize overlays before merge

When onboarding to an existing cluster and needing a configuration audit without cluster access

When an AI coding agent needs structured diagnosis of deployment errors as part of a CI/CD workflow

When preparing a production readiness checklist for a new service

When not to use it

When the failure requires live cluster inspection such as exec into a pod or real-time metric queries

When the root cause is application-level runtime behaviour rather than configuration

When cluster changes need to be applied automatically — this skill only produces suggestions, not executions

When the deployment tooling is outside Kubernetes entirely, such as ECS or bare-metal Ansible