diff --git a/lib/crewai/src/crewai/agent/utils.py b/lib/crewai/src/crewai/agent/utils.py
index 20eb06b47..297322a0c 100644
--- a/lib/crewai/src/crewai/agent/utils.py
+++ b/lib/crewai/src/crewai/agent/utils.py
@@ -74,13 +74,17 @@ def build_task_prompt_with_schema(task: Task, task_prompt: str) -> str:
 
     if (task.output_json or task.output_pydantic) and not task.response_model:
         if task.output_json:
-            schema_dict = generate_model_description(task.output_json)
+            schema_dict = generate_model_description(
+                task.output_json, strip_null_types=False
+            )
             schema = json.dumps(schema_dict["json_schema"]["schema"], indent=2)
             task_prompt += "\n" + I18N_DEFAULT.slice(
                 "formatted_task_instructions"
             ).format(output_format=schema)
         elif task.output_pydantic:
-            schema_dict = generate_model_description(task.output_pydantic)
+            schema_dict = generate_model_description(
+                task.output_pydantic, strip_null_types=False
+            )
             schema = json.dumps(schema_dict["json_schema"]["schema"], indent=2)
             task_prompt += "\n" + I18N_DEFAULT.slice(
                 "formatted_task_instructions"
diff --git a/lib/crewai/tests/agents/test_agent_utils.py b/lib/crewai/tests/agents/test_agent_utils.py
new file mode 100644
index 000000000..3d2a2152e
--- /dev/null
+++ b/lib/crewai/tests/agents/test_agent_utils.py
@@ -0,0 +1,41 @@
+"""Tests for crewai.agent.utils prompt-building helpers."""
+
+from __future__ import annotations
+
+import json
+
+import pytest
+from pydantic import BaseModel, Field
+
+from crewai import Task
+from crewai.agent.utils import build_task_prompt_with_schema
+
+
+class _Output(BaseModel):
+    name: str
+    note: str | None = Field(default=None)
+
+
+@pytest.mark.parametrize("output_attribute", ["output_pydantic", "output_json"])
+def test_optional_fields_stay_nullable_in_the_prompt_schema(
+    output_attribute: str,
+) -> None:
+    """An Optional field must still be expressible as null in the prompt schema.
+
+    The provider-side response schema generated from the same model allows null,
+    so stripping it here hands the model two contradictory contracts and leaves
+    it no way to say "not applicable". Both output attributes embed a schema in
+    the prompt, so both are pinned.
+    """
+    task = Task(description="d", expected_output="e", **{output_attribute: _Output})
+
+    prompt = build_task_prompt_with_schema(task, "")
+
+    start = prompt.index("{")
+    end = prompt.rindex("}", start) + 1
+    schema = json.loads(prompt[start:end])
+
+    assert {entry["type"] for entry in schema["properties"]["note"]["anyOf"]} == {
+        "string",
+        "null",
+    }
diff --git a/lib/crewai/tests/llms/bedrock/test_bedrock.py b/lib/crewai/tests/llms/bedrock/test_bedrock.py
index 0127ffd62..87c0d6c09 100644
--- a/lib/crewai/tests/llms/bedrock/test_bedrock.py
+++ b/lib/crewai/tests/llms/bedrock/test_bedrock.py
@@ -602,24 +602,31 @@ def test_bedrock_tool_conversion():
     assert "inputSchema" in bedrock_tools[0]["toolSpec"]
 
 
-def test_bedrock_environment_variable_credentials(bedrock_mocks):
-    """
-    Test that AWS credentials are properly loaded from environment
-    """
-    mock_session_class, _ = bedrock_mocks
+def test_bedrock_environment_variable_credentials():
+    """Pass AWS credentials and region from the environment to boto3."""
+    with (
+        patch.dict(
+            os.environ,
+            {
+                "AWS_ACCESS_KEY_ID": "test-access-key-123",
+                "AWS_SECRET_ACCESS_KEY": "test-secret-key-456",
+                "AWS_DEFAULT_REGION": "eu-west-1",
+            },
+            clear=False,
+        ),
+        patch(
+            "crewai.llms.providers.bedrock.completion.Session"
+        ) as mock_session_class,
+    ):
+        mock_session_class.return_value.client.return_value = MagicMock()
+        LLM(model="bedrock/anthropic.claude-3-5-sonnet-20241022-v2:0")
 
-    mock_session_class.reset_mock()
-
-    with patch.dict(os.environ, {
-        "AWS_ACCESS_KEY_ID": "test-access-key-123",
-        "AWS_SECRET_ACCESS_KEY": "test-secret-key-456"
-    }):
-        llm = LLM(model="bedrock/anthropic.claude-3-5-sonnet-20241022-v2:0")
-
-        assert mock_session_class.called
-        call_kwargs = mock_session_class.call_args[1] if mock_session_class.call_args else {}
-        assert call_kwargs.get('aws_access_key_id') == "test-access-key-123"
-        assert call_kwargs.get('aws_secret_access_key') == "test-secret-key-456"
+    mock_session_class.assert_called_once_with(
+        aws_access_key_id="test-access-key-123",
+        aws_secret_access_key="test-secret-key-456",
+        aws_session_token=None,
+        region_name="eu-west-1",
+    )
 
 
 def test_bedrock_token_usage_tracking():