Standing Orders: How Agent Instructions Differ From a Prompt

A prompt is a one-time ask: you type it, it runs once, it is gone. Instructions are standing orders. They live with the agent and apply every single time it runs. If you find yourself retyping the same rules every session, you do not have an agent yet.

This is the first of the three things that make something an agent. The other two are tools and triggers, and the overview is in What Is an AI Agent?

The difference in one sentence

A prompt is something you say. Instructions are something the agent has.

That sounds like wordplay until you look at where each one lives. A prompt lives in your head and in a chat box. Instructions live in a file, or a system prompt, or a project, somewhere the agent reads every run whether or not you are thinking about it.

The test that settles it

Ask what happens on a Tuesday when you are busy and nobody opens the app.

If the standards still apply, they are instructions. If the standards only exist because you remembered to type them, they are a prompt, and the quality of the work is now a function of your memory.

What good instructions actually contain

Not a personality. What "done well" means for this job, in the words you would use with a new hire on their first week.

• The standard. What finished looks like, specifically enough that two people would agree on whether it was met.
• The boundaries. What it must never do, and what it has to bring to you instead of deciding.
• The context it cannot infer. Names, preferences, the reason something is done the odd way it is done.

Why this is the piece to write first

Tools without instructions is an agent that can act and does not know what good means. Triggers without instructions is that same agent, running on its own, on a schedule.

Instructions are cheap to write and they are the only one of the three that improves everything downstream, so they are worth doing properly before you hand anything access.