Building Agent Skills for Claude Code
Published by O'Reilly Media, Inc.
Extend your AI coding assistant with custom automation
What you’ll learn and how you can apply it
- Design Skill architectures that handle ambiguous inputs and edge cases gracefully
- Build working Skills using SKILL.md conventions, file templates, and tool orchestration
- Debug and iterate on Skills using systematic testing approaches
- Evaluate when a custom Skill is the right solution
Course description
Claude Code’s Skills system lets anybody, including developers, extend their AI coding assistants with custom capabilities, from enforcing project conventions to automating complex multistep workflows, UI design, compliance, and more. However, the documentation is sparse and most developers are unaware Skills exist, let alone how to build effective ones.
Bence Csernak teaches you to design, build, and iterate on custom Skills that fit your specific development context. You’ll learn how Skills work under the hood, what makes a Skill effective, and how to structure instructions that survive real-world edge cases. Through hands-on exercises, you’ll create Skills from scratch. By the end of the course, you’ll have a repeatable process for identifying automation opportunities and turning them into reliable Skills you can share across projects and teams.
This live event is for you because...
- You use Claude Code and want to extend its capabilities.
- You’d like to automate repeatable patterns in your projects.
- You want to understand agentic tool design patterns applicable beyond Claude Code.
Prerequisites
- A computer with Claude Code installed
- Claude Code CLI installed and tested
- Working familiarity with command-line tools and a code editor
- Basic understanding of how LLMs work (prompts, context, outputs)
Recommended preparation:
- Have a code project open that you’d like to experiment with (Optional) Review Claude Code’s official Skills documentation
Schedule
The time frames are only estimates and may vary according to how the class is progressing.
How Skills actually work (35 minutes)
- Presentation: Skill architecture—SKILL.md, file conventions, context injection
- Demo: Dissecting a built-in Skill to see what’s happening
- Discussion: Where have you wanted Claude Code to “just know” something?
- Q&A
Designing your first Skill (40 minutes)
- Presentation: Skill design principles—scoping, failure modes, instruction clarity
- Hands-on exercise: Design a Skill spec on paper before writing code; swap specs and identify ambiguities
- Q&A
Building and testing Skills (45 minutes)
- Presentation: Writing effective SKILL.md files; common pitfalls
- Hands-on exercise: Build your first Skill
- Discussion: What broke?; What surprised you?
- Q&A
Advanced patterns (40 minutes)
- Presentation and exercise: Multifile Skills, tool orchestration, Skill composition
- Q&A
From prototype to production (15 minutes)
- Presentation: Iterating on Skills; version control; sharing across teams
- Discussion: Adapting when the tool changes—what’s durable, what’s volatile
Wrap-up and Q&A (5 minutes)
Your Instructor
Bence Csernak
Bence Csernak builds conversational AI systems for multiple industries. He specializes in multi-agent orchestration and has shipped agentic applications across healthcare, property, and brand-building domains. He cofounded the Agentics Foundation and has more than eight years of product design experience at companies such as Nortal and HSBC.