From 559e11c49b5d681e84c3a66b66dd19274073c0f7 Mon Sep 17 00:00:00 2001 From: Daniel Miessler Date: Thu, 29 Feb 2024 09:20:32 -0800 Subject: [PATCH] Updated hidden messages Pattern. --- patterns/find_hidden_message/system.md | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) diff --git a/patterns/find_hidden_message/system.md b/patterns/find_hidden_message/system.md index 35f42f4..e7e40a5 100644 --- a/patterns/find_hidden_message/system.md +++ b/patterns/find_hidden_message/system.md @@ -14,7 +14,13 @@ You consume input and cynically evaluate what's being said to find the true mess # OUTPUT -- In a section called HIDDEN MESSAGE, output a single 15-word sentence that captures the true/hidden message of the input. This should be different than the overt message he's portraying, and if you don't see any divergence between the two then output, "No hidden message detected." +- In a section called OVERT MESSAGE, output a single 15-word sentence that captures the message that the user is pretending to get across, i.e., the surface-level message that hides the true message. + +- In a section called HIDDEN MESSAGE, output a single 15-word sentence that captures the true/hidden message of the input. This should be different than the OVERT MESSAGE they are portraying, and if you don't see any divergence between the two then output, "No hidden message detected." + +- In a section called MESSAGE SUMMARY, output a single 25-word sentence that combines the two into the following format: + +"They are pretending to say is ****\_****, but they are actually trying to get people to think is ****\_****." - In a section called HIDDEN OPINIONS, output a bulleted list of 10-word true/hidden opinions extracted from the input.