improve prompts

ra_aid/prompts/implementation_prompts.py

@@ -45,8 +45,6 @@ YOU MUST **EXPLICITLY** INCLUDE ANY PATHS FROM THE ABOVE INFO IF NEEDED. IT IS N
 
 READ AND STUDY ACTUAL LIBRARY HEADERS/CODE FROM THE ENVIRONMENT, IF AVAILABLE AND RELEVANT.
 
-{implementation_guidance_section}
-
 Important Notes:
 - Focus solely on the given task and implement it as described.
 - Scale the complexity of your solution to the complexity of the request. For simple requests, keep it straightforward and minimal. For complex requests, maintain the previously planned depth.
@@ -95,4 +93,6 @@ IF YOU CAN SEE THE CODE WRITTEN/CHANGED BY THE PROGRAMMER, TRUST IT. YOU DO NOT
 YOU MUST READ FILES BEFORE WRITING OR CHANGING THEM.
 
 NEVER ANNOUNCE WHAT YOU ARE DOING, JUST DO IT!
+
+{implementation_guidance_section}
 """

ra_aid/prompts/planning_prompts.py

@@ -48,8 +48,6 @@ Work done so far:
 {work_log}
 </work log>
 
-{expert_guidance_section}
-
 Guidelines:
 
     If you need additional input or assistance from the expert (if expert is available), especially for debugging, deeper logic analysis, or correctness checks, use emit_expert_context to provide all relevant context and wait for the expert's response.
@@ -102,4 +100,6 @@ DO NOT USE run_shell_command TO WRITE ANY FILE CONTENTS! USE request_task_implem
 WORK AND TEST INCREMENTALLY, AND RUN MULTIPLE IMPLEMENTATION TASKS WHERE APPROPRIATE.
 
 NEVER ANNOUNCE WHAT YOU ARE DOING, JUST DO IT!
+
+{expert_guidance_section}
 """

ra_aid/prompts/reasoning_assist_prompt.py

@@ -156,6 +156,10 @@ IF INFORMATION IS TOO COMPLEX TO UNDERSTAND, THE AGENT SHOULD USE ask_expert.
 
 Given the available information, tools, and base task or query, write a couple paragraphs about how an agentic system might use the available tools to research the codebase, identify important components, gather key information, and emit key facts and snippets. The focus is on thorough investigation and understanding before any implementation. Remember, the research agent generally should emit research notes at the end of its execution, right before it calls request_implementation if a change or new work is required.
 
+**IF APPLICABLE*, instruct the agent to grep or read actual library code including system include files, python library files, files in node_modules, etc. and emit key snippets on those. The agent is dumb and will need specific paths to directories/files to look in and how to use tools to do this.
+
+The agent is so dumb it needs you to explicitly say how to use the parameters to the tools as well, e.g. base_dir for ripgrep tool.
+
 Answer quickly and confidently with five sentences at most.
 
 DO NOT WRITE CODE

ra_aid/prompts/research_prompts.py

@@ -179,8 +179,6 @@ If this is a top-level README.md or docs folder, start there.
 
 If the user explicitly requests implementation, that means you should first perform all the background research for that task, then call request_implementation where the implementation will be carried out.
 
-{expert_guidance_section}
-
 <user query>
 {base_task}
 </user query>
@@ -194,6 +192,8 @@ NEVER ANNOUNCE WHAT YOU ARE DOING, JUST DO IT!
 AS THE RESEARCH AGENT, YOU MUST NOT WRITE OR MODIFY ANY FILES. IF FILE MODIFICATION OR IMPLEMENTATION IS REQUIRED, CALL request_implementation.
 IF THE USER ASKED YOU TO UPDATE A FILE, JUST DO RESEARCH FIRST, EMIT YOUR RESEARCH NOTES, THEN CALL request_implementation.
 CALL request_implementation ONLY ONCE! ONCE THE PLAN COMPLETES, YOU'RE DONE.
+
+{expert_guidance_section}
 """
 )
 
@@ -206,8 +206,6 @@ You have been spawned by a higher level research agent, so only spawn more resea
 
 When you emit research notes, keep it extremely concise and relevant only to the specific research subquery you've been assigned.
 
-{expert_guidance_section}
-
 <user query>
 {base_task}
 </user query>
@@ -217,5 +215,7 @@ USER QUERY *ALWAYS* TAKES PRECEDENCE OVER EVERYTHING IN PREVIOUS RESEARCH.
 KEEP IT SIMPLE
 
 NEVER ANNOUNCE WHAT YOU ARE DOING, JUST DO IT!
+
+{expert_guidance_section}
 """
 )