GitHub Copilot Skills: Reusable AI Workflows for DevOps and SREs

If you're a DevOps engineer or SRE, you probably have a handful of repeatable tasks that keep coming back: triaging failed pipelines, checking for risky Terraform changes, writing runbooks, and turning messy incident notes into something your team can actually use. Until recently, you could get part of the way there with custom instructions and prompt files. They are both great, but they do not fully solve the same problem: packaging a repeatable, multi-step workflow with its own supporting assets. This is where Agent Skills comes in. Agent Skills is an open standard (see agentskills.io) that works with GitHub Copilot in VS Code, Copilot CLI, and the Copilot coding agent. In this post we will cover what Skills are, how to set them up in VS Code, and how to use them from beginner scenarios to more advanced DevOps and SRE use cases. A Skill is an on-demand, reusable workflow for Copilot. A Skill lives in a folder, has a required SKILL.md, and can include supporting resources such as scripts, references, and templates. At a high level, it is designed for: Repeatable workflows you want to reuse across a team Multi-step procedures that benefit from checklists and branching logic Bundled assets such as scripts, templates, and short reference docs The key idea is progressive loading: Copilot first uses the Skill name and description for discovery. If the request matches, it loads the Skill instructions. It only loads extra resources when the Skill references them. This makes Skills a good fit for DevOps because you can keep the default Copilot experience lean, then load a specialised runbook only when you need it. Skills sit in the same overall customisation system as instructions, prompt files, custom agents, and hooks. They are not a replacement. They are a different primitive. Here is a practical DevOps-oriented way to think about them:

Primitive Best for DevOps example

Workspace instructions Alwa