Agent Skill
Agent Skill
A packaged set of instructions and know-how that guides an AI agent through completing a task.
In Simple Terms
An Agent Skill is basically a "how-to manual" or set of instructions that an AI agent refers to when carrying out a specific task. For example, skills might cover things like "steps for editing a PDF file" or "the process for creating a report in a particular format." Based on instructions from a person, the AI loads the right skill, figures out which tools to use and in what order, and carries the work through to the goal on its own. This lets AI go beyond just answering questions in text — it's used in systems where AI handles complex tasks on a person's behalf.
Behind the Name
The name blends "agent" — someone who acts on another's behalf — with "skill," a trained ability to do something well. The idea is that an AI agent is handed a packaged "skill," much like teaching someone a specific procedure, so it can smoothly carry out a particular task.
Take a Closer Look!
An Agent Skill is a package of procedural knowledge and guidelines that an AI agent refers to in order to achieve a given goal.
It teaches the AI more than just which tools to call — it passes along the know-how of which tool to use, in what order, and under what conditions.
Put simply, it's like a "how-to guide" you'd use to train a new employee. By loading up the manual (the skill), the AI understands the workflow and things to watch out for, then picks the right tool (like web search or file operations) to carry out the work.
For example, skills might be defined for things like "steps for investigating a website outage" or "steps for a code security check."
The AI plans out the overall approach while following the steps laid out in the skill.
By giving AI skills, its role expands from simply answering questions to becoming an automation system that completes a whole sequence of tasks on its own.
Adding and sharing specialized know-how as skills makes it easy to broaden the range of work AI can handle.