convez.ai
MCP ServerDevOps & InfraestruturaAvançado
seokheejang

Claude Ops Skills

seokheejang/claude-ops-skills

Skills DevOps read-only para K8s, Helm, Terraform, ArgoCD e RPC

Ver no GitHub
InicialPluginScore · de 0 a 100

MCP server com skills e agents para operações DevOps diárias em Kubernetes, Helm, Terraform, ArgoCD e infraestrutura blockchain. Fornece inspeções read-only, auditorias de segurança e pesquisa de boas práticas sem sair do terminal, com bloqueio automático de comandos mutantes.

Instalação

$npx convez-ai seokheejang/claude-ops-skills

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
1
Forks
0
Acompanhando
0
Saúde
42%
Último commit
há 2 semanas
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Inspeção e diagnóstico de clusters Kubernetes
  • Auditoria de segurança (RBAC, NetworkPolicy, Pod Security)
  • Validação e análise de charts Helm e estado de releases
  • Análise de planos Terraform e inspeção de estado
  • Verificação de saúde de nós RPC blockchain
  • Troubleshooting estruturado com evidências observadas

Veredito

Conjunto robusto de skills especializadas para SREs e engenheiros DevOps que gerenciam infraestrutura complexa. Abordagem conservadora (read-only por padrão) com foco em observabilidade, segurança e troubleshooting estruturado é um diferencial sólido.

README

A tradução para o português deste README ainda está na fila.

claude-ops-skills

License: MIT Claude Code

Claude Code slash-command skills and agents for DevOps daily operations. For DevOps engineers and SREs who manage Kubernetes clusters, Helm charts, Terraform IaC, ArgoCD GitOps, and blockchain infrastructure — run read-only inspections, security audits, and best-practice research without leaving the terminal.

All operations are READ-ONLY by default. Mutating commands are blocked at the settings level and never executed.

Table of Contents

Quick Start

git clone https://github.com/seokheejang/claude-ops-skills.git
cd claude-ops-skills
make install

Then restart Claude Code and try:

/k8s-ops my-cluster          # Inspect a K8s cluster
/helm-ops my-release          # Check Helm release status
/best-practice EKS CNI choice # Research industry best practices

Other commands:

make update     # git pull + reinstall
make uninstall  # Uninstall
make test       # Run PreToolUse hook tests

Skills

SkillCommandDescription
k8s-ops/k8s-ops <cluster>K8s cluster inspection (get, describe, logs, top) — paired with k8s-craft for authoring
k8s-craft/k8s-craft <resource or topic>K8s manifest authoring & design (workloads, networking, storage, RBAC, NetworkPolicy, cost, operators, mesh, multi-cluster)
k8s-security/k8s-security <cluster>K8s security audit (RBAC, NetworkPolicy, Pod Security)
ssh-ops/ssh-ops <host>SSH server inspection
troubleshoot/troubleshoot <symptom>Evidence-based failure-diagnosis discipline — every judgment cites observed output, fixes cross-checked from multiple angles, side effects enumerated, remediation is operator-confirmed manual guidance only
rpc-health/rpc-health <endpoint>Blockchain RPC node health check
rpc-agent/rpc-agent <endpoint> <module>RPC analytics agent (block traversal, tx counting)
db-ops/db-ops <database>DB read-only queries (SELECT, SHOW only)
helm-ops/helm-ops <cluster-or-chart>Helm chart validation, release inspection, chart authoring guide
terraform-ops/terraform-ops <directory>Terraform state inspection, plan analysis, IaC authoring guide
argocd-ops/argocd-ops <app-or-context>ArgoCD app status, sync monitoring, drift detection, GitOps manifest authoring
ralph/ralph [N] <task>Self-review loop (iterative verification, default 5 rounds)
mmdraw/mmdraw <target>Mermaid diagram generator from source/docs analysis (→ Excalidraw manual conversion)
compound/compound [completed|paused] <desc>Work synthesis - learnings capture, doc lifecycle, CHANGELOG update
catchup/catchupRead-only session context restore — reads git (recent commits + uncommitted) → CLAUDE.md → README → docs, briefs what was done last & where to start
best-practice/best-practice <topic or question>DevOps best practice research with Citation verification, AI content detection, source independence, domain-aware recency
grill-me/grill-meStress-test a plan or design via relentless one-by-one interview until shared understanding
write-a-skill/write-a-skillInterview-based scaffolding for new skills — enforces description/length/security conventions
transfer-en/transfer-en <한국어>Korean → easy conversational English (CEFR A2) with casual alternative + auto-accumulating learning notes

Agents

AgentDescription
k8s-debuggerSystematic K8s issue debugging
k8s-security-auditorComprehensive K8s security audit (risk analysis + remediation guide)
rpc-monitorRPC node status monitoring
rpc-analyticsRPC analytics engine (block traversal, tx aggregation)
helm-chart-auditorComprehensive Helm chart audit (lint, security, best practices)
argocd-drift-detectorSystematic ArgoCD drift detection across apps and clusters

Prerequisites

ToolRequiredPurpose
Claude CodeYesCLI runtime for skills
makeYesInstall/update/uninstall
kubectlFor K8s skillsK8s cluster inspection
helmFor Helm skillsChart validation, release inspection
terraformFor Terraform skillsState inspection, plan analysis
argocdFor ArgoCD skillsApp status, sync monitoring
sshFor SSH skillsServer inspection

Only Claude Code and make are required. Other tools are needed only for their respective skills.

Add an aliases field in clusters.yaml to use short names for clusters:

# skills/k8s-ops/clusters.yaml (local only, .gitignore)
clusters:
  my-cluster:
    kubeconfig: "~/.kube/my_cluster_config"
    aliases: ["dev", "my dev"]     # /k8s-ops dev → my-cluster
  my-prod:
    kubeconfig: "~/.kube/my_prod_config"
    aliases: ["prod"]              # /k8s-security prod → my-prod

Usage:

  • /k8s-ops dev → matches alias dev → my-cluster
  • /k8s-security prod → matches alias prod → my-prod
  • /k8s-ops my-cluster → exact key match → my-cluster

Match priority: exact key name > alias match > partial match (prompts selection if multiple candidates)

clusters.yaml is included in .gitignore so real cluster names are never exposed in the public repo. See clusters.yaml.example for configuration examples.

Structure

├── skills/           # Symlinked to ~/.claude/skills/
│   ├── k8s-ops/      # K8s inspection skill + clusters.yaml (auto-generated locally)
│   │   └── references/ # Troubleshooting deep-dive
│   ├── k8s-craft/    # K8s manifest authoring & design (paired with k8s-ops)
│   │   └── references/ # 8 topics (workloads, networking, storage, ...)
│   ├── k8s-security/ # K8s security audit skill
│   ├── ssh-ops/      # SSH inspection skill
│   ├── troubleshoot/ # Evidence-based failure-diagnosis discipline (no-speculation, manual-remediation guardrails)
│   ├── rpc-health/   # RPC health check skill
│   ├── rpc-agent/    # RPC analytics agent entry point
│   │   └── scripts/  # Bundled analysis scripts (cosmos_total_tx.py, etc.)
│   ├── db-ops/       # DB query skill
│   ├── helm-ops/     # Helm chart inspection + authoring guide
│   │   └── references/ # Chart authoring deep-dive (structure, values, templates, hooks, testing)
│   ├── terraform-ops/ # Terraform state/plan + IaC authoring guide
│   ├── argocd-ops/   # ArgoCD monitoring + GitOps manifest authoring
│   │   └── references/ # GitOps deep-dive (ArgoCD/Flux install, AppSet, sealed secrets)
│   ├── ralph/        # Self-review loop (iterative verification)
│   │   └── references/ # Domain checklists + FAIL/NOTE judgement policy
│   ├── mmdraw/       # Mermaid diagram generator (→ Excalidraw conversion)
│   ├── compound/     # Work synthesis - learnings, doc lifecycle, CHANGELOG
│   │   └── references/ # 3 topics (work-doc, learnings, changelog)
│   ├── catchup/      # Session context restore (git + CLAUDE.md/README/docs)
│   ├── best-practice/ # DevOps best practice research
│   │   └── references/ # 3 topics (sources, verification, output)
│   ├── grill-me/     # Plan/design stress-test via relentless interview
│   ├── write-a-skill/ # Meta-skill: scaffold new skills with conventions
│   └── transfer-en/  # Korean → easy English translator + learning note system (local-only learning.md/pending.md)
├── agents/           # Sub-agent definitions
│   ├── k8s-debugger.md
│   ├── k8s-security-auditor.md
│   ├── rpc-analytics.md  # RPC analytics engine (EVM + Cosmos)
│   ├── rpc-monitor.md
│   ├── helm-chart-auditor.md    # Helm chart comprehensive audit
│   └── argocd-drift-detector.md # ArgoCD drift detection
├── configs/          # Configuration templates
│   ├── settings.json.template     # allow/deny rules
│   ├── claude.md.template         # CLAUDE.md merge content
│   └── settings.local.json.example
├── scripts/          # Install/management scripts
│   ├── install.sh    # Backup → symlink → merge
│   ├── uninstall.sh
│   └── update.sh
└── templates/        # Templates for new skills/agents
  1. Backup: Saves existing files to ~/.claude/backups/claude-ops-skills/{timestamp}/
  2. Symlink skills: skills/* → ~/.claude/skills/* (skips if already linked)
  3. Merge settings.json: Adds allow/deny rules to existing config (preserves hooks, deduplicates)
  4. Merge CLAUDE.md: Marker-based block management (# === claude-ops-skills:start/end ===)
  5. settings.local.json: Never modified

After make install, a Stop hook is registered that sends a macOS notification when Claude finishes a task. If you're already looking at VSCode or iTerm, the notification is suppressed.

To disable, remove the Stop section from ~/.claude/settings.json:

// Delete this block in ~/.claude/settings.json → hooks → Stop
"Stop": [
  {
    "hooks": [
      {
        "type": "command",
        "command": "FRONT=$(osascript -e ...)  ..."
      }
    ]
  }
]

This feature is macOS only (osascript). On Linux, the Stop hook will silently fail with no side effects.

Safety

  • All operations are READ-ONLY — no system-affecting commands are executed
    • K8s: get, describe, logs, top, exec (inspection only)
    • Helm: list, status, get, show, lint, template, history, diff (no install/upgrade/delete)
    • Terraform: state list/show, plan, validate, fmt -check, output (no apply/destroy/init)
    • ArgoCD: app get/list/diff/logs/history, repo list, cluster list (no sync/create/delete)
  • Mutating commands are blocked by deny rules in settings.json + CRITICAL SAFETY sections in each SKILL.md
  • When changes are needed, the skill provides commands as text guidance only — never executes them
  • Chart/IaC authoring uses Write/Edit tools to create files locally — no infrastructure impact
  • Sensitive data (kubeconfig contents, SSH keys) is never included in this repo
  • clusters.yaml is local only — install.sh auto-generates it by scanning ~/.kube/ (included in .gitignore)

Reference

DevOps Skills

Diagram & Visualization

Compound Engineering

Claude Code Ecosystem

Architecture & Patterns

Ver o original no GitHub

Bom para

Relacionados