> For the complete documentation index, see [llms.txt](https://www.christiandussol.dev/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://www.christiandussol.dev/ai-native/skills-series/part-1-how-claude-code-skills-work.md).

# Part 1: How Claude Code Skills work

## **Understand what a Skill actually is, before building one**

The first part of the series sits at Layer 4 of the four-layer framework of AI agents (Foundation, Capabilities, MCP, Skills). It explains what makes Skills different from prompts, scripts, or plugins, and why the format matters.

<figure><img src="/files/cETFBOznSc8zHeUozF3k" alt="" width="563"><figcaption></figcaption></figure>

### What the article covers

* The anatomy of a Skill: YAML frontmatter, instructions, references, scripts
* Three-level progressive disclosure and why context window economics matter
* The invocation flow: how Claude decides when to load a Skill
* Why Skills are to AI agents what Helm Charts are to Kubernetes deployments

### Read

* **Medium article**: [From Prompts to Packages: How Claude Code Skills work](https://medium.com/@christian.dussol/from-prompts-to-packages-how-claude-code-skills-work-4609cdefd79f)
* **Series companion**: [The 4-layer framework for understanding AI agents](https://medium.com/@christian.dussol) *(December 2025 background article)*