Make your AI coding assistant ops-smart. 10 battle-tested skills that turn Claude Code / Cursor into a DevOps expert.
English | 简体中文
bash -c "$(curl -fsSL https://raw.githubusercontent.com/tomlen045/ops-skills/main/install.sh)"
One command. 10 skills. Zero configuration. (Unreachable network? The installer auto-falls back to the jsDelivr CDN mirror.)
| Mode | What you get | Best for |
|---|---|---|
1) Native skills → ~/.claude/skills/<name>/SKILL.md |
Auto-activating skills with trigger descriptions | Claude Code v2+ and any harness that reads SKILL.md (Codex, Cursor, Pi…) — recommended |
2) Single-file CLAUDE.md |
All 10 skills concatenated into one context file | Claude Code v1, manual control |
3) .cursorrules |
Same, in Cursor's format | Cursor users |
| 4) Both 2+3 | — | Mixed setups |
| 5) Plain download | ./ops-skills/*.md |
Everything else |
AI coding assistants are great at writing React components and fixing TypeScript errors. But ask them to review a Kubernetes manifest for security issues, or guide you through a production incident, and they give you generic ChatGPT-level advice.
That's because they don't have ops-specific context — the checklists, the prioritization frameworks, the "don't do this or you'll wake up at 3AM" knowledge that senior SREs accumulate over years.
ops-skills fixes this. Each skill is a carefully crafted set of instructions that transforms your AI assistant from a generalist into an ops specialist.
| Skill | What it does | When you need it |
|---|---|---|
| 🚨 SRE Incident Commander | Guides incident response with proper SEV triage, escalation, and structured troubleshooting | Production is down and you're panicking |
| ☸ K8s Security Reviewer | Reviews manifests against 15+ security checks (CIS, NSA guidelines) | Before deploying to production |
| 🔒 Linux Security Hardener | Server hardening with CIS benchmarks, includes rollback plans | Setting up a new server |
| 📊 Prometheus Architect | Generates alert rules that page when they should and don't when they shouldn't | Setting up monitoring |
| 🗺 Infra Diagram Generator | Reads Terraform/compose/K8s → generates Mermaid architecture diagrams | Documentation is stale (always) |
| 📝 Postmortem Writer | Structures incident reports with timeline, root cause, action items | After the fire is out |
| 🏗 Terraform Guardian | Reviews IaC for security, cost, and reliability issues | Before terraform apply |
| 🔍 Log Detective | Pattern-matches log output to identify root causes | Reading logs and going "???" |
| 🐳 Docker Reviewer | Reviews Dockerfiles for security, size, and production readiness | Before building images |
| 📖 On-call Runbook Generator | Generates decision-tree runbooks for each alert | Before going on-call |
bash -c "$(curl -fsSL https://raw.githubusercontent.com/tomlen045/ops-skills/main/install.sh)"This interactive script detects whether you use Claude Code, Cursor, or both, and installs accordingly.
For Cursor users: Download the combined .cursorrules file and place it in your project root.
For Claude Code users: Download CLAUDE.md and place it in your project root.
For any LLM: Download individual skill files from skills/ and use them as system prompts.
Each skill is a standalone .md file in skills/ — read them, copy them, remix them. They work as standalone prompts with any LLM.
Generic AI assistants give generic advice because they lack domain-specific context. Each skill in this pack includes:
- Prioritized checklists — what to check first, not just "check everything"
- Structured output formats — consistent severity/fvidence/fix format
- Anti-patterns — the "never do this" list that prevents 3AM pages
- Exact commands — not "check the logs" but the exact
kubectlcommand with flags - Risk assessment — every suggestion includes what could go wrong
- Rollback plans — because "undo" is important when you're in production
Got a skill that saved your production? PR it. Requirements:
- Solves a real ops problem (not theoretical)
- Structured format with severity levels
- Includes exact commands (not vague instructions)
- Has a "What NOT to do" section
- Written in English
MIT
Built by operators, for operators.
🔍 搜「小薅薅」看更多运维工具
小薅薅 — 年轻人的赛博工具箱。每天为你发现一个好玩、实用、开源的小工具，帮你节省 1 小时探索时间。
📱 微信搜索「小薅薅」 或扫描下方二维码关注
后台回复关键词获取：
- 回复 「工具」 → 获取全部工具合集（离线可用）
- 回复 「运维」 → 获取运维/安全资源包
- 回复 「加群」 → 加入工具交流群
| 🎯 更多作品 | 描述 | Stars |
|---|---|---|
| ops-doctor | 一条命令给Linux服务器做全套体检+健康分 | ⭐ 实用工具 |
| shellmbti | 你的终端历史暴露了你是谁——Shell MBTI人格测试 | ⭐ 病毒传播 |
| naicha-mbti | 8道题测出你的奶茶人格，生成分享卡片 | 🧋 爆款 |
| fafa-generator | 发疯文学生成器——一键生成发疯文案+卡片 | 🔥 热门 |
| life-progress | 人生进度条——把你的时间摆在眼前 | ⏳ 走心 |