k8s-diagnostics

Skill: k8s-diagnostics

Cluster and workload diagnostics playbook using RootCause MCP tool names.

Focus areas:

• cluster health assessment
• resource usage and pressure
• event chronology
• HPA and VPA behavior
• best-practice interpretation
• capacity planning

Objective

Produce evidence-backed health posture and action priorities.

This skill is mostly read-first. If remediation requires mutation, hand off to k8s-operations.

Primary Tool Set

• k8s.context
• k8s.ping
• k8s.overview
• k8s.resource_usage
• k8s.events
• k8s.events_timeline
• k8s.hpa_debug
• k8s.vpa_debug
• k8s.best_practice
• k8s.list
• k8s.describe
• k8s.graph

Trigger Conditions

Activate for requests like:

• cluster health check
• resource hotspots
• scaling seems wrong
• noisy warnings across namespace
• capacity planning ahead of release
• post-incident verification

Baseline Diagnostic Sequence

1. verify active context with k8s.context
2. verify control plane with k8s.ping
3. collect baseline with k8s.overview
4. detect pressure with k8s.resource_usage
5. correlate warnings with k8s.events_timeline
6. inspect autoscaling using k8s.hpa_debug and k8s.vpa_debug
7. score workload configuration with k8s.best_practice
8. produce prioritized recommendations

Parameter Examples

k8s.context current:

{
  "action": "current"
}

k8s.context list:

{
  "action": "list"
}

k8s.ping:

{}

k8s.overview namespace:

{
  "namespace": "payments"
}

k8s.resource_usage cpu hotspot view:

{
  "namespace": "payments",
  "includePods": true,
  "includeNodes": true,
  "sortBy": "cpu",
  "limit": 40
}

k8s.resource_usage memory hotspot view:

{
  "namespace": "payments",
  "includePods": true,
  "includeNodes": true,
  "sortBy": "memory",
  "limit": 40
}

k8s.events namespace warnings and normals:

{
  "namespace": "payments"
}

k8s.events_timeline warning-only:

{
  "namespace": "payments",
  "includeNormal": false,
  "limit": 150
}