Across the industry developers and builders have discovered the potential to create software using AI agents. As an agent enabled software engineer, you probably have your stack of favorite prompts and tools. Would it be cool to have short cuts for your favorite agent prompts? It turns out that Anthropic has proposed an elegant open standard to empower agent users to encapsulate their favorite prompts, processes, and workflows into re-usable elements called skills.
I want to give a shout out to Orlando Devs for a fun and insightful meetup yesterday. Make sure to join their Discord and helpful community. Thank you to Will Gorman of Nucleus Security for leading a hands-on workshop exploring practical use of AI agents. I enjoyed getting to explore opencode.ai and exploring some common developer scenarios. If you’re looking for a free option for exploring agents, OpenCode looks neat. It was neat to learn how Mr. Gorman’s company has explored building agent skills to improve security for their clients. It seemed like a cool use case for agent skills for process encapsulation, data capture and tool integration.
In this post, I’d like to explore a few helpful links so that you can get started with agents skills. Since agent skills have become an open standard for multiple agent tools, you can adapt these insights to your favorite tools. While these links will focus on Github Copilot, you can adapt these insights for sure.
Benefits of Agent Skills
- Automate repetitive, boilerplate, or multi-step tasks that traditionally break your focus.
- Eliminate the need to switch between your editor, terminal, browser, or documentation for common operations. Your IDE becomes your command center.
- Delegate mundane tasks, freeing up your mental energy for complex problem-solving and creative coding.
- Make advanced tooling or multi-step processes accessible via simple natural language commands.
How do I get started?
Agent Skills Overview & Open Specification
- Link: Agent Skills Official Site
- Author Anthropic
- Summary: This is the home base for the Agent Skills format. It outlines a lightweight, open standard designed to extend AI agent capabilities by packaging procedural knowledge, scripts, and templates into portable, version-controlled folders. It explains the concept of “progressive disclosure,” where an agent loads a skill’s full instructions into its context window only when a specific task triggers a match. The site contains full format specifications, quickstart guides, and an impressive showcase of tools that natively support the standard (including Claude Code, Cursor, and GitHub Copilot).
Teach Copilot How Your Team Works in Visual Studio
- Link: Agent Skills in Visual Studio (Microsoft Dev Blogs)
- Author: Simona Liao (Microsoft)
- Summary: Microsoft officially announced support for Agent Skills inside Visual Studio. This blog post details how developers can use a dedicated UI skills panel to create, edit, and discover reusable instructions (using the
SKILL.mdformat) right inside their IDE. It features a highly practical breakdown of Skills vs. Custom Instructions, explaining that while custom instructions are “always-on” global preferences (e.g., coding styles), Agent Skills are dynamically loaded, task-specific workflows (e.g., executing build pipelines or drafting specific GitHub issue templates).
Official Guide: Adding Agent Skills for GitHub Copilot
- Link: Adding Agent Skills for GitHub Copilot (GitHub Docs)
- Summary: A comprehensive, technical deep-dive into creating and configuring custom skills for GitHub Copilot cloud agents, CLI, and code reviews. This piece outlines exactly where to locate project-specific directories (like
.github/skills/) versus personal directories. It provides template examples of writing the necessary YAML frontmatter (defining name and trigger descriptions) and highlights how to safely pre-approve supplementary bash or shell scripts using theallowed-toolsfield to automate local environments without security risks.
Community Power: The Awesome Copilot Skills Repository
- Link: Awesome Copilot Agent Skills Library (GitHub)
- Summary: If you don’t want to build every workflow from scratch, this is your repository. It acts as a curated directory of community-built agent skills that you can install instantly using the GitHub CLI (
gh skills install). The index contains dozens of highly specific, production-ready skill folders ranging from repository architectural blueprint generators and OWASP security compliance checkers, to niche automation scripts for tools like Adobe Illustrator and Azure App Insights telemetry instrumentation.
Leave a Reply