diff --git a/src/google/adk/plugins/global_instruction_plugin.py b/src/google/adk/plugins/global_instruction_plugin.py
index 4251f691f6..ed2a6d4821 100644
--- a/src/google/adk/plugins/global_instruction_plugin.py
+++ b/src/google/adk/plugins/global_instruction_plugin.py
@@ -79,9 +79,8 @@ async def before_model_callback(
       return None
 
     # Resolve the global instruction (handle both string and InstructionProvider)
-    readonly_context = ReadonlyContext(callback_context.invocation_context)
     final_global_instruction = await self._resolve_global_instruction(
-        readonly_context
+        callback_context
     )
 
     if not final_global_instruction:
diff --git a/tests/unittests/plugins/test_global_instruction_plugin.py b/tests/unittests/plugins/test_global_instruction_plugin.py
index 2253b1fb5a..851f3a9334 100644
--- a/tests/unittests/plugins/test_global_instruction_plugin.py
+++ b/tests/unittests/plugins/test_global_instruction_plugin.py
@@ -43,7 +43,7 @@ async def test_global_instruction_plugin_with_string():
   mock_invocation_context.session = mock_session
 
   mock_callback_context = Mock(spec=CallbackContext)
-  mock_callback_context.invocation_context = mock_invocation_context
+  mock_callback_context._invocation_context = mock_invocation_context
 
   llm_request = LlmRequest(
       model="gemini-1.5-flash",
@@ -80,10 +80,10 @@ async def build_global_instruction(readonly_context: ReadonlyContext) -> str:
   )
 
   mock_invocation_context = Mock(spec=InvocationContext)
-  mock_invocation_context.session = mock_session
 
   mock_callback_context = Mock(spec=CallbackContext)
-  mock_callback_context.invocation_context = mock_invocation_context
+  mock_callback_context._invocation_context = mock_invocation_context
+  mock_callback_context.session = mock_session
 
   llm_request = LlmRequest(
       model="gemini-1.5-flash",
@@ -119,7 +119,7 @@ async def test_global_instruction_plugin_empty_instruction():
   mock_invocation_context.session = mock_session
 
   mock_callback_context = Mock(spec=CallbackContext)
-  mock_callback_context.invocation_context = mock_invocation_context
+  mock_callback_context._invocation_context = mock_invocation_context
 
   llm_request = LlmRequest(
       model="gemini-1.5-flash",
@@ -156,7 +156,7 @@ async def test_global_instruction_plugin_leads_existing():
   mock_invocation_context.session = mock_session
 
   mock_callback_context = Mock(spec=CallbackContext)
-  mock_callback_context.invocation_context = mock_invocation_context
+  mock_callback_context._invocation_context = mock_invocation_context
 
   llm_request = LlmRequest(
       model="gemini-1.5-flash",
@@ -191,7 +191,7 @@ async def test_global_instruction_plugin_prepends_to_list():
   mock_invocation_context.session = mock_session
 
   mock_callback_context = Mock(spec=CallbackContext)
-  mock_callback_context.invocation_context = mock_invocation_context
+  mock_callback_context._invocation_context = mock_invocation_context
 
   llm_request = LlmRequest(
       model="gemini-1.5-flash",