You will hear skills, agents, and playbooks around here. They are related, but they don't do the same job.
The useful distinction is simple: a skill helps you through a method, an agent takes on more of the work, and a playbook gives you the complete setup for a particular kind of work.
The short version
| Robot | What it does | Best when |
|---|---|---|
| Skill | Guides you through a specific method. | Your judgment needs to stay in the room. |
| Agent | Takes on more of a job and may complete multiple steps. | The job has clear goals, repeatable steps, or research to gather. |
| Playbook | Provides a complete sequence, setup, or working environment. | You need a reliable way to run a larger piece of work from start to finish. |
Skills: a method you can use again
A skill is a reusable set of instructions for one specific way of working. It doesn't just ask AI for an answer. It gives AI a method: what to ask, what information to look for, what order to work in, and what quality checks to make.
You are still part of the process. A good skill asks for the real context, lets you make the decisions that need taste or judgment, and keeps the work from becoming a blank-page prompt every time.
Skills are especially useful for work such as developing an idea, editing a draft, making a strategy decision, or turning a familiar workflow into a repeatable system. You learn the method as you use it, and you can course-correct whenever the work needs you.
Agents: more of the job, with more autonomy
An agent takes on more of a task after you set the direction. It can work through several steps, gather information, use connected tools or data where available, and return something for you to review.
That makes agents useful for jobs with clear criteria: researching a topic, checking a set of documents, finding patterns, or completing a series of routine steps. You still need to set the standard and inspect the result. An agent can move work forward, but it doesn't inherit your judgment by magic.
Playbooks: the whole working environment
A playbook is the fuller setup. It gives you a sequence, the context to prepare, and the environment for doing a larger kind of work well. It may include skills, reference material, decision points, and an order of operations.
Use a playbook when you don't just need one helpful conversation. You need a dependable way to move through a project or recurring piece of work without rebuilding the setup each time.
Which one should you choose?
Start with a skill when the important part is your thinking, taste, or decision-making. Choose an agent when the work can be broken into clear steps and you want AI to take on more of the legwork. Reach for a playbook when you need the complete sequence and working environment, not just one method.
Around here, they are all robots. The point isn't to memorise product taxonomy. The point is to give AI a job that fits the work in front of you.
Where to go next
- Want to turn a method of your own into something reusable? Read Turn one way of working into an AI skill.
- Need a clearer setup before AI can help? Read What AI needs to know before it can help.
- Want to see the library? Enter RobotsOS.