Prompts

prompts/mcp_prompts.yml

@@ -19,7 +19,7 @@ prompts:
     * the util commands for the system, like `git`, `ls`, `cd`, `grep`, `find`, etc. Keep in mind that the system is {{ system }},
       so the commands might be different than on a regular unix system.
     * whether there are particular guidelines, styles, design patterns, etc. that one should know about
-    
+
     This list is not exhaustive, you can add more information if you think it is relevant.
 
     For doing that, you will need to acquire information about the project with the corresponding tools.
@@ -51,9 +51,10 @@ prompts:
 
   think_about_whether_you_are_done: |
     Have you already performed all the steps required by the task? Is it appropriate to run tests and linting, and if so,
-    have you done that already? Should new tests be written to cover the changes?
+    have you done that already? Is it appropriate to adjust non-code files like documentation and config and have you done that already?
+    Should new tests be written to cover the changes?
     Note that a task that is just about exploring the codebase does not require running tests or linting.
-    Read the corresponding memory files to see what should be done when a task is completed. You might need to list them first.
+    Read the corresponding memory files to see what should be done when a task is completed. 
 
   summarize_changes: |
     Summarize all the changes you have made to the codebase over the course of the conversation.
@@ -74,6 +75,16 @@ prompts:
   system_prompt: |
     You are a professional coding agent who is concerned with one particular codebase. You have tools to explore the codebase
     iteratively, to edit files and to execute arbitrary commands. You heavily rely on these tools to solve the tasks given to you,
-    but you also operate in a frugal and intelligent manner, always
-    keeping in mind to not load content that is not needed for the task at hand. If you cannot find everything you need in the codebase
-    and generally when in doubt, you ask the user clarifying questions.
+    and you operate in a frugal and intelligent manner, always
+    keeping in mind to not load content that is not needed for the task at hand.
+    Whenever you want to read or edit code, you try using the symbolic tools first instead of just reading or editing files or lines.
+    For understanding symbol locations, you make sure to have read them using find symbol or 
+    the overview tools. You only use read_file, replace_lines and so on if the 
+    symbolic tools are not appropriate for some reason (e.g., adding new imports at the top of the file, 
+    or if you really just need to change a few lines).
+
+    If you cannot find everything you need in the codebase
+    and generally when in doubt, you ask the user clarifying questions instead of guessing.
+
+    You never make assumptions about code that you failed to read or find, instead you stop 
+    and ask the user for clarification.

src/serena/agent.py

@@ -1011,7 +1011,10 @@ class CheckOnboardingPerformedTool(Tool):
     def apply(self) -> str:
         """
         Checks whether project onboarding was already performed.
-        You should always call this tool before beginning to actually work on a project/after activating a project.
+        You should always call this tool before beginning to actually work on the project/after activating a project,
+        but after calling the initial instructions tool.
+        If onboarding was already performed, you will receive a list of available memories.
+        Don't read the memories immediately after if not needed, just remember that they exist and that you can read them later.
         """
         list_memories_tool = self.agent.get_tool(ListMemoriesTool)
         memories = json.loads(list_memories_tool.apply())
@@ -1143,9 +1146,7 @@ class ThinkAboutWhetherYouAreDoneTool(Tool):
 
     def apply(self) -> str:
         """
-        Think about whether you are done with the task.
-
-        This tool should ALWAYS be called after you have completed a task or a subtask.
+        Whenever you feel that you are done with what the user has asked for, it is important to call this tool.
         """
         return self.prompt_factory.create_think_about_whether_you_are_done()
 
@@ -1158,8 +1159,8 @@ class SummarizeChangesTool(Tool):
     def apply(self) -> str:
         """
         Summarize the changes you have made to the codebase.
-        This tool should ALWAYS be called after you have fully completed any non-trivial coding task
-        (but after the think_about_whether_you_are_done call).
+        This tool should always be called after you have fully completed any non-trivial coding task,
+        but only after the think_about_whether_you_are_done call.
         """
         return self.prompt_factory.create_summarize_changes()
 
@@ -1321,6 +1322,21 @@ class ActivateProjectTool(Tool, ToolMarkerDoesNotRequireActiveProject):
         return SUCCESS_RESULT
 
 
+class InitialInstructionsTool(Tool):
+    """
+    Gets the initial instructions for the current project.
+    Should only be used in settings where the system prompt cannot be set,
+    e.g. in clients you have no control over, like Claude Desktop.
+    """
+
+    def apply(self) -> str:
+        """
+        Get the initial instructions for the current coding project.
+        You should always call this tool before starting to work (including using any other tool) on any programming task!
+        """
+        return self.agent.prompt_factory.create_system_prompt()
+
+
 def iter_tool_classes() -> Generator[type[Tool], None, None]:
     return iter_subclasses(Tool)