In short
What you send a model to get something back. In a chat, a message. In an agent, the instruction file plus the situation. Save them. Improve them. Never retype them.
A prompt is the input. In a chat window it is the message you type. In an agent it is assembled automatically: the instruction file, the rules for this job, the material to work on. Same idea, different scale.
The skill people call prompt engineering is mostly two habits. State what you want clearly, with the shape of the answer you expect. And treat the output as information: read where the model misunderstood, adjust, run again. The second habit is the one that separates people who get good results from people who stop after two tries.
The habit that matters most for a business: save the prompt. A prompt that worked once, typed into a chat and lost, is a chore you will do again tomorrow. A prompt saved in a file and refined over time is an asset. That file is an instruction file.
Example
What it looks like.
Lost: typing 'summarize this transcript into action items' every Monday.Kept: a file that says how action items are formatted, who gets assigned, what counts as a decision, run by an agent every Monday.
Related
Read next.
- instruction fileThe Markdown file that tells an agent who it is, what it owns, and what it may never do. The agent's job description, readable by the agent.Read the definition
- output is informationA bad result is data about what the model understood. Read it, adjust, go again. Beginners stop at two tries. Builders go fifteen.Read the definition
- program in txt and mdThe instructions are the code. Write what you want in plain language, keep it in a file, update the file forever. No Python, no Bash.Read the definition