Fix ToolResult parameters in test

Co-Authored-By: Joe Moura <joao@crewai.com>

src/crewai/utilities/agent_utils.py

@@ -214,8 +214,9 @@ def handle_agent_action_core(
     if show_logs:
         show_logs(formatted_answer)
 
-    if messages is not None:
-        messages.append({"role": "assistant", "content": tool_result.result})
+    # Tool results are already included in the formatted answer with "Observation:" prefix,
+    # so we don't need to append them again to messages to avoid duplication and token bloat.
+    # This fixes issue #2798 where tool results were being duplicated in the LLM prompt.
 
     return formatted_answer
 

tests/test_tool_result_duplication.py

@@ -0,0 +1,93 @@
+from typing import Dict, List, Optional, Union
+from unittest.mock import MagicMock, patch
+
+import pytest
+
+from crewai.agent import Agent
+from crewai.agents.crew_agent_executor import CrewAgentExecutor
+from crewai.agents.tools_handler import ToolsHandler
+from crewai.crew import Crew
+from crewai.task import Task
+from crewai.tools import BaseTool
+from crewai.tools.structured_tool import CrewStructuredTool
+from crewai.utilities.agent_utils import parse_tools
+
+
+class TestTool(BaseTool):
+    name: str = "Test Tool"
+    description: str = "A test tool to verify tool result duplication is fixed"
+
+    def _run(self) -> str:
+        return "Test tool result"
+
+
+def test_tool_result_not_duplicated_in_messages() -> None:
+    """Test that tool results are not duplicated in messages.
+    
+    This test verifies the fix for issue #2798, where tool results were being
+    duplicated in the LLM prompt, increasing token usage and latency.
+    """
+    agent = Agent(
+        role="Test Agent",
+        goal="Test the tool result duplication fix",
+        backstory="I am a test agent",
+        tools=[TestTool()],
+    )
+    
+    task = Task(
+        description="Use the test tool and return the result",
+        expected_output="The test tool result",
+        agent=agent,
+    )
+    
+    crew = Crew(agents=[agent], tasks=[task])
+    
+    structured_tools = parse_tools(agent.tools)
+    tools_names = ", ".join([t.name for t in structured_tools])
+    tools_description = "\n".join([t.description for t in structured_tools])
+    
+    tools_handler = MagicMock(spec=ToolsHandler)
+    
+    with patch.object(CrewAgentExecutor, '_invoke_loop') as mock_invoke_loop:
+        executor = CrewAgentExecutor(
+            agent=agent,
+            task=task,
+            crew=crew,
+            llm=agent.llm,
+            prompt={"system": "System prompt", "user": "User prompt"},
+            max_iter=10,
+            tools=structured_tools,
+            tools_names=tools_names,
+            stop_words=[],
+            tools_description=tools_description,
+            tools_handler=tools_handler,
+            callbacks=[],
+        )
+        
+        executor.messages = [{"role": "user", "content": "Use the test tool"}]
+        
+        from crewai.agents.parser import AgentAction
+        from crewai.tools.tool_types import ToolResult
+        
+        agent_action = AgentAction(
+            tool="Test Tool",
+            tool_input={},
+            thought="I should use the test tool",
+            text="I'll use the Test Tool",
+        )
+        
+        tool_result = ToolResult(
+            result="Test tool result",
+            result_as_answer=False,
+        )
+        
+        executor._handle_agent_action(agent_action, tool_result)
+        
+        tool_result_count = sum(
+            1 for msg in executor.messages if msg.get("content") == "Test tool result"
+        )
+        
+        assert tool_result_count <= 1, "Tool result is duplicated in messages"
+        
+        observation_text = f"Observation: {tool_result.result}"
+        assert observation_text in agent_action.text, "Tool result not properly formatted in agent action text"