Agent Skills Deep Dive

An agent skill is a reusable package of instructions, knowledge, resources, and tools that teaches an AI agent how to perform a specific kind of work. Instead of explaining your process again in every conversation, you can capture it once and give your agent a repeatable way to apply it when the situation calls for it.

And a good skill can be much more than a prompt.

It can contain your process and standards. It can reference supporting documents, templates, examples, and external resources. It can include scripts or code for steps that need to be deterministic rather than left to the judgment of an AI model. It can be tested, versioned, improved, and shared.

For knowledge workers and domain experts, this creates an important shift.

The marketer who understands exactly how their company’s brand voice works, the consultant with a research process refined over years, the product manager whose review process consistently catches important problems, or the content strategist with a proven methodology can turn that expertise into something an AI agent can repeatedly apply.

Your expertise stops living only in conversations and starts becoming reusable infrastructure.

In this live, online workshop, we’ll go from the fundamentals of agent skills all the way into designing and building production-quality skills.

We’ll start by understanding what a skill is, how agents discover and load skills, and how progressive disclosure allows a skill to provide the right amount of information at the right time.

Then we’ll go deeper.

We’ll look at how to decide what belongs in instructions versus reference material, when an AI model should make a judgment and when a step should be deterministic, how scripts and tools can extend a skill, how to incorporate external references and examples, how to structure larger skills without overwhelming the agent, and how to compose skills so one can call another instead of growing into a single monolith.

We’ll also cover the lifecycle of a skill: testing it, debugging unreliable behavior, organizing and maintaining its files, using version control, documenting changes, and evolving a skill as your process changes.

And then you’ll build one.

During the hands-on portion of the workshop, you’ll create an agent skill based on something you already know how to do. We’ll build it step by step with Claude (or a similar agent that supports skills), test it against real scenarios, identify where it breaks down, and improve it.

No programming experience is required. The emphasis is not on becoming a developer. It’s on learning how to turn your expertise into a system an AI agent can reliably use.

You’ll leave with a working skill, a framework for designing better ones, and a method for turning more of your knowledge, processes, and standards into reusable capabilities for AI agents.

Agenda

1. What an agent skill is, and what it isn’t
2. How agents discover and load skills
3. Progressive disclosure: the right information at the right time
4. Instructions versus reference material: deciding what belongs where
5. Judgment versus determinism: when a step should be a script
6. Structuring larger skills without overwhelming the agent
7. Composable skills: building skills that call other skills
8. The lifecycle of a skill: testing, debugging, versioning, and evolving it
9. Hands-on: build your own agent skill from scratch
10. Hands-on: test it against real scenarios and improve where it breaks