Assume everything in the prompt will be seen
In a nutshell: Write a hidden prompt on the assumption that a user will eventually read every word of it
Use it when: You are putting instructions, persona or data into a system prompt a user is not meant to see.
- MIT
- Sources
- 2
- Rules
- 2
- Memos
- 1
- Field
- Prompt engineering
- Authors
- Brex
- DAIR.AI
Contains 2 Rules
If you missed the previous warnings in this doc, **you should always assume that any data exposed to the language model will eventually be seen by the user**.