A practical guide can help you think through a task you want an AI agent to perform. But a guide and an installable Skill are not the same thing. The guide is a published tutorial to consult; it is not a listed Skill package. Treat it as a reference for planning the work, not as something the agent can install.
This distinction keeps expectations clear. You can use a guide to identify useful steps and questions, then describe those steps directly in your instructions. The agent still needs the task, relevant context, and a clear definition of what a useful result should look like.
Start with the task, not the guide
Before referring to a tutorial, write down what you want the agent to accomplish. Make the goal specific enough that the agent can tell what work is in scope and what result you expect. Then use the guide to help organize the process—not to substitute for your own task instructions.
For example, rather than asking the agent to “use this guide,” explain the task you need done and identify which parts of the guide may be relevant. If the guide describes a process, translate that process into steps that apply to your situation. Do not assume that every step fits your task.
Turn the reference into clear instructions
Break the work into a sequence the agent can follow. Depending on the task, that might mean gathering the necessary information, completing a defined action, and checking the result. Use concrete verbs and make the boundaries explicit: what should the agent do, what should it leave alone, and when should it stop?
A useful instruction also says how to handle uncertainty. Ask the agent to flag a step it cannot carry out with the information provided, rather than filling the gap with a guess. If a step from the guide does not apply, the agent should explain the issue instead of forcing it into the workflow.
Provide context and define the output
A guide offers general direction, but your task may depend on details that are not in the tutorial. Supply the relevant background, source material, constraints, and any terms the agent needs to understand. Be clear about which information it should rely on and whether it should ask before making a consequential choice.
Describe the output you need, too. Specify its form and the elements it should include—for instance, a short summary, a checklist, or a structured draft. When the output has a clear shape, you can more easily assess whether the agent completed the task as intended.
Review the result and handle unsupported steps
Check the agent’s work against your original goal and instructions. Look for missing steps, unclear claims, and places where the result goes beyond the context you supplied. If the agent could not complete part of the task, review its explanation and decide what additional information or direction is needed.
The guide can help you plan and review a workflow, but it does not guarantee that the agent’s result is correct or that every step is suitable. Keep your instructions tied to the actual task, and treat unresolved questions as questions to address—not as facts the agent can safely assume.
You can use the catalog’s beginner’s guide as a starting reference for thinking through an AI task: https://www.browse-ai.tools/blog/how-to-use-ai-tools-step-by-step-guide-beginners-2026. Use it to support your planning, while keeping the distinction clear: a published tutorial can inform your instructions, but it is not itself an installable Skill.