Appendix B. Agentic skills for Vanilla Web
Let’s say you want to build your next Vanilla Web app with an AI coding agent, such as Codex or Claude Code. You can describe the screens and features, but there are still decisions to make: how to organize the components, where to keep state, and whether the project needs a router or a library. We’ve explored those decisions throughout the book. Now we can give the agent guidance for making them too.
An Agent Skill is a directory of instructions and supporting references that an agent can read while working on your project. It lets you reuse that guidance across tasks, so you don’t have to explain your development approach in every prompt. This book includes two skills in the accompanying files, under code/skills/.
You can use them without giving the agent a copy of the manuscript. Each skill contains the guidance it needs, with chapter references for you to explore further. The agent still needs to understand your project and choose which parts apply, just as you would when building the app yourself.
B.1 Choosing a skill
The first choice depends on what you want to build or review. One skill looks at the application as a whole; the other focuses on a reusable Web Component: