Claude Code tutorial: reusable skills and scripts

Austin Marchese presents four rules for using skills in Claude Code, based on his reading of Anthropic engineering material and excerpts from engineering talks. His advice centers on saving repeatable workflows in skill folders rather than rewriting detailed prompts for each task.

The tutorial describes a skill as a description that helps Claude select it, instructions for completing the task, and supporting tools such as scripts, API calls or reference files. Marchese uses domain availability checks to explain why a skill needs useful tools as well as clear prose.

He recommends splitting broad workflows into focused, reusable skills. His example separates content creation into YouTube idea research, script writing and LinkedIn posts. A slide-styling example explains how saving a Python script inside a skill lets later sessions reuse it.

For this coding assistant, the tutorial also discusses invocation controls. Marchese describes user-invocable as controlling visibility in the slash menu and disable-model-invocation as reserving execution for the user. He suggests the latter for actions such as sending messages or deploying code.

The final recommendation is deliberate maintenance: review corrections after using a skill and save recurring rules, examples or edge cases. The improvement he describes depends on updating the skill, rather than assuming each session automatically teaches the model.