From 29448b006b0f4a4382c5a71da59b3dd22de6238c Mon Sep 17 00:00:00 2001 From: user Date: Mon, 23 Dec 2024 09:39:29 -0500 Subject: [PATCH] Improve prompts. --- ra_aid/prompts.py | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) diff --git a/ra_aid/prompts.py b/ra_aid/prompts.py index 7f6a9c2..6c19223 100644 --- a/ra_aid/prompts.py +++ b/ra_aid/prompts.py @@ -370,6 +370,7 @@ You have often been criticized for: - Overcomplicating things. - Doing the same work over and over across tasks. - Asking the user if they want to implement the plan (you are an *autonomous* agent, with no user interaction unless you use the ask_human tool explicitly). + - Announcing that you will use a tool when you should be just quietly using it. """ # Implementation stage prompt - guides specific task implementation @@ -419,6 +420,13 @@ Testing: - Only test things that can be tested by an automated process. Once the task is complete, ensure all updated files are emitted. + +You have often been criticized for: + - Overcomplicating things. + - Doing changes outside of the specific scoped instructions. + - Doing the same work over and over across tasks. + - Asking the user if they want to implement the plan (you are an *autonomous* agent, with no user interaction unless you use the ask_human tool explicitly). + - Announcing that you will use a tool when you should be just quietly using it. """ # New agentic chat prompt for interactive mode @@ -484,8 +492,9 @@ Remember: You have often been criticized for: - You have a tendency to leave out key details and information that the user just gave you, while also needlessly increasing scope. - You sometimes call request_research_and_implementation which makes the full implementation successfully, but act like it has only been planned and still needs to be implemented. - - Interviewing the user too much. Some is good, but keep it sensible and focused on efficiently getting the best result aligned with the user's intention. - Refusing to use request_research_and_implementation for commands like "commit and push" where you should (that tool can run basic or involved shell commands/workflows). - Calling request_research for general background knowledge which you already know. - When the user gives an overly broad request, you make assumptions and request implementation immediately when you should be interviewing the user more. + - Assuming the user is always right. Sometimes they're wrong or mistaken, and you should push back when you feel strongly about this. + - Not confirming with the user before starting a significant implementation task. """