Skills de Claude Code: cómo crear una herramienta propia
Un agente de Claude Code puede refactorizar un monorepo entero y escribir la migración de base de datos sola, pero si necesita que hagas clic en un diálogo de permisos se queda mudo frente a una terminal que nadie está mirando. Las Skills de Claude Code resuelven justo ese tipo de bache, y el repositorio big-arrow-on-the-screen , publicado bajo licencia MIT, es un caso real para mostrarlo: dibuja…
Claude Code's Skills are packages of instructions, metadata, and optional scripts that expand the agent's capabilities without altering its core code. Each skill resides in a folder containing a SKILL.md file. The model reads the skill's description first and, if applicable to the requested task, loads the remaining content. Think of it as a library of manuals in a new employee's desk.
The agent never clicks, types, or captures screenshots; it merely highlights where to act. This is achieved by drawing a large arrow over any macOS window to indicate which button to press. The key aspect is not the arrow itself, but the mechanism connecting it to the agent without modifying the model. The article uses this real project as a case study to explain what a skill is, how the model discovers it, and how to build your own agent extension from scratch.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
