适用于 DevOps 和 SRE 工作的 SKILL.md 库。

该仓库设计目的是：

• 适合人类的优秀runbook。
• 一个可预测、机器可读的代理合同（Claude技能，类OpenCode代理）

原则

• 默认安全：技能应为只读，除非明确标注
• 示例中无秘密，无环境特定标识符
• 程序包括验证和回滚/撤销指导

仓库布局

• skills/ / /SKILL.md - 技能档案（每个技能一个）
• skills/_template/SKILL.md - 可复制模板
• schema/ - 前件结构和约定
• scripts/ - 验证器与工具

类别故意设置浅层（例如，kubernetes、aws、terraform、argocd、incident）。偏好标签而非深度文件夹树。

技能赛制

每个SKILL.md开头是YAML前言（在---行之间），随后是Markdown部分。

参见skills/_template/SKILL.md了解典范结构。

与代理人一起使用

• Claude 技能：将技能文件加载为指令包;使用前置文件作为合同（输入/工具/安全）

devops-sre-skills

Agent-ready SKILL.md library for DevOps and SRE work.

This repo is designed to be:

• Good runbooks for humans
• A predictable, machine-readable contract for agents (Claude skills, OpenCode-like agents)

Principles

• Safe by default: skills should be read-only unless explicitly labeled otherwise
• No secrets, no environment-specific identifiers in examples
• Procedures include verification and rollback/undo guidance

Repo layout

• skills///SKILL.md - the skill file (one per skill)
• skills/_template/SKILL.md - copyable template
• schema/ - the front matter schema and conventions
• scripts/ - validators and tooling

Categories are intentionally shallow (e.g., kubernetes, aws, terraform, argocd, incident). Prefer tags over deep folder trees.

Skill format

Each SKILL.md starts with YAML front matter (between --- lines) followed by Markdown sections.

See skills/_template/SKILL.md for the canonical structure.

Using with agents

• Claude skills: load a skill file as an instruction bundle; use the front matter as the contract (inputs/tools/safety)
• OpenCode-like agents: enforce tools_allowed/safety gates from the front matter; present the Markdown procedure during execution

Contributing

See CONTRIBUTING.md.