System Prompts and Custom Instructions
Write standing instructions that shape every reply, and test them so they hold up in real use.
A taste of a lesson
I want our team assistant to never share our pricing strategy. Can I just say that in the system prompt?
You can add it, but do not rely on it. Instructions are guidance, not a lock. A determined user can often get an assistant to ignore or reveal them, and adherence is never perfect. The safer rule is simpler: if the assistant must not share the pricing strategy, do not give it access to that document at all. Use the instruction as a second layer, for example 'If asked about pricing strategy, say you do not have that information'. Which documents does the assistant currently have access to?
Written by the teacher as an example. In your lesson the tutor answers your own questions, and like any AI it can be wrong.
What you will be able to do
- Draft concise standing instructions with scope, format and uncertainty rules
- Avoid conflicts between standing instructions and everyday requests
- Test instructions with typical, edge and adversarial requests
- Explain why instructions are neither secret nor a security control
Lesson plan
- 1 What standing instructions are Understand system prompts and their consumer versions across different tools. Start
- 2 What belongs and what does not Choose content that improves most replies without overloading the model. Start
- 3 Writing short, clear rules Phrase instructions so they are followed consistently. Start
- 4 Testing your instructions Check instructions against varied requests before relying on them. Start
- 5 Limits and shared assistants Know what instructions cannot do, especially for team assistants. Start
Try asking
About this tutor
Most assistants let you set instructions that apply to every conversation: custom instructions, project instructions, or the system prompt of a custom assistant you build for your team. The names differ by tool but the craft is the same. This tutor teaches what belongs in standing instructions and what does not, how to keep them short and stable, how they interact with what you type in the chat, and how to test them with varied requests. You also learn their limits: they drift in long chats, they are not secret, and they are not a security boundary. You leave with a tested set of instructions for yourself or a shared assistant.
Reviews
4.5
2 ratingsSample
- Ana C.Sample
My custom instructions went from a messy wish list to six clear lines. The test set idea caught a conflict I had between brief and thorough.
- Brian K.Sample
Very useful for our team assistant. The point that instructions are not a security control changed how we set up file access.
About the teacher
Prompt workflows for heavy users: chaining, standing instructions, long documents and reasoning models
9 tutors 269 lessons taught Sample
I work with people who already use AI assistants every day and want more dependable results. My background is in operations and process design, which taught me to treat a prompt like a small procedure: inputs, steps, checks and a clear output. I teach chaining, reusable instructions, long document work and how to test whether a prompt change actually helped....
See Darius's profile and tutorsMore like this
Other tutors on the same or nearby topics.