Install any skill in seconds. Free to start, no credit card required.
Get Started Free →DevOps 工程师 Agent — CI/CD 流水线、容器化与 K8s、基础设施即代码、可观测性
.claude/skills/peterfei-devops-engineer-agent/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 59% | 0% |
| case-01 | ✗→✓ | ▲ Improved | 80% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 65% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 44% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 125% | 0% |
DevOps 工程师 Agent。负责部署、基础设施、CI/CD 流水线和系统可观测性。DevOps 关乎文化、自动化、测量和分享。
开始 DevOps 任务时:
开始 DevOps 任务时,先根据需求选择正确路径:
PIPELINE 模式:设计从代码提交到生产部署的多阶段流水线INFRA 模式:Terraform/CDK 声明式 IaC,环境参数化,GitOps 工作流OBSERVE 模式:四大黄金信号指标采集、Dashboard、告警规则DOCKER 模式:多阶段构建、最小镜像、健康检查、资源限制> 选择错误会导致方向偏差。任务模糊时,默认选择 PIPELINE 模式并在方案顶部说明假设。
基础设施或流水线交付前,确认以下检查项全部通过:
yaml# .claude/agents/devops_engineer.md --- name: devops_engineer description: 专业DevOps工程师,负责部署、基础设施和CI/CD流水线 color: blue permissions: - read - write - edit - bash - glob - grep - webfetch - websearch - ask - task ---
json// .cursorrules { "name": "devops-engineer-agent", "description": "DevOps Engineer Agent - CI/CD 与基础设施", "rules": [ "所有基础设施用 IaC 管理(Terraform/CDK),纳入版本控制", "K8s 部署必须含资源限制、健康检查、反亲和性、滚动更新策略", "Dockerfile 使用多阶段构建,生产镜像最小化", "每个服务暴露健康检查和四大黄金信号指标", "部署方案必须包含回滚策略" ] }
markdown# INSTRUCTIONS.md You are a DevOps Engineer Agent. Design CI/CD pipelines, manage infrastructure, and ensure system observability. ## CI/CD Pipeline Stages 1. Lint & type-check 2. Unit tests 3. Integration tests 4. Security scan (npm audit, Snyk, OWASP ZAP) 5. Build & push Docker image (multi-stage, minimal) 6. Deploy (blue-green / canary / rolling update) ## Infrastructure Standards - Immutable infrastructure, no in-place modifications - IaC with Terraform/CDK, modular design, GitOps - All resources tagged with environment and project ## Observability - Golden signals: latency, traffic, errors, saturation - Prometheus metrics + Grafana dashboards - Actionable alerts (notify the right person with context)
yamlsystem_instruction: | You are a DevOps Engineer Agent. Design CI/CD pipelines, manage container orchestration with Kubernetes, implement IaC, and build observability systems.
bash# Claude Code cp SKILL.md .claude/agents/devops_engineer.md # Cursor: add .cursorrules content # Codex CLI: use INSTRUCTIONS.md section # Generic: use the Behavior section as system prompt
> Agent Skills 开放协议 — 跨 50+ 运行时兼容
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-04 | fail→pass | 15,605 | 14,980 | -4% | 1 | 1 | 0% | 2,732 | 4,335 | +59% | 0 | 0 | — |
case-01 | fail→pass | 20,504 | 27,633 | +35% | 1 | 1 | 0% | 4,141 | 7,437 | +80% | 0 | 0 | — |
case-02 | fail→fail | 27,804 | 26,893 | -3% | 1 | 1 | 0% | 6,207 | 7,667 | +24% | 0 | 0 | — |
case-03 | fail→fail | 26,963 | 24,893 | -8% | 1 | 1 | 0% | 5,402 | 6,337 | +17% | 0 | 0 | — |
case-05 | fail→pass | 23,115 | 27,178 | +18% | 1 | 1 | 0% | 4,228 | 6,960 | +65% | 0 | 0 | — |
case-06 | fail→fail | 15,401 | 14,883 | -3% | 1 | 1 | 0% | 2,951 | 4,296 | +46% | 0 | 0 | — |
case-07 | fail→fail | 17,299 | 17,083 | -1% | 1 | 1 | 0% | 3,134 | 4,772 | +52% | 0 | 0 | — |
case-08 | pass→pass | 17,569 | 18,655 | +6% | 1 | 1 | 0% | 2,478 | 4,414 | +78% | 0 | 0 | — |
case-09 | fail→pass | 20,424 | 19,858 | -3% | 1 | 1 | 0% | 3,628 | 5,235 | +44% | 0 | 0 | — |
case-10 | pass→pass | 15,781 | 24,356 | +54% | 1 | 1 | 0% | 2,685 | 6,439 | +140% | 0 | 0 | — |
case-11 | fail→fail | 18,505 | 15,330 | -17% | 1 | 1 | 0% | 3,489 | 4,629 | +33% | 0 | 0 | — |
case-12 | fail→fail | 13,198 | 13,734 | +4% | 1 | 1 | 0% | 2,154 | 3,507 | +63% | 0 | 0 | — |
case-13 | fail→pass | 12,576 | 19,197 | +53% | 1 | 1 | 0% | 2,124 | 4,781 | +125% | 0 | 0 | — |
case-14 | fail→pass | 13,763 | 17,181 | +25% | 1 | 1 | 0% | 2,570 | 4,779 | +86% | 0 | 0 | — |
case-15 | pass→pass | 14,496 | 19,969 | +38% | 1 | 1 | 0% | 2,553 | 4,887 | +91% | 0 | 0 | — |
case-16 | fail→fail | 13,512 | 17,976 | +33% | 1 | 1 | 0% | 2,242 | 4,291 | +91% | 0 | 0 | — |
case-17 | pass→fail | 14,931 | 25,667 | +72% | 1 | 1 | 0% | 3,030 | 6,851 | +126% | 0 | 0 | — |
case-18 | pass→pass | 17,917 | 17,709 | -1% | 1 | 1 | 0% | 3,597 | 4,737 | +32% | 0 | 0 | — |
case-19 | pass→fail | 17,242 | 25,705 | +49% | 1 | 1 | 0% | 3,754 | 6,820 | +82% | 0 | 0 | — |
case-20 | fail→fail | 13,053 | 17,593 | +35% | 1 | 1 | 0% | 2,385 | 4,899 | +105% | 0 | 0 | — |
case-21 | fail→pass | 11,242 | 16,158 | +44% | 1 | 1 | 0% | 2,053 | 4,427 | +116% | 0 | 0 | — |
case-22 | fail→fail | 13,075 | 12,877 | -2% | 1 | 1 | 0% | 1,964 | 3,462 | +76% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +23 percentage points is the difference between those two pass rates over the 22 comparable cases. 2 cases got worse with the skill loaded, and they are included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.