test(steering): Fix failing integ tests (#1580)

Author: mkmeral

src/strands/experimental/steering/core/handler.py

@@ -115,9 +115,7 @@ def _handle_tool_steering_action(
             logger.debug("tool_name=<%s> | tool call proceeding", tool_name)
         elif isinstance(action, Guide):
             logger.debug("tool_name=<%s> | tool call guided: %s", tool_name, action.reason)
-            event.cancel_tool = (
-                f"Tool call cancelled given new guidance. {action.reason}. Consider this approach and continue"
-            )
+            event.cancel_tool = f"Tool call cancelled. {action.reason} You MUST follow this guidance immediately."
         elif isinstance(action, Interrupt):
             logger.debug("tool_name=<%s> | tool call requires human input: %s", tool_name, action.reason)
             can_proceed: bool = event.interrupt(name=f"steering_input_{tool_name}", reason={"message": action.reason})

src/strands/experimental/steering/handlers/llm/llm_handler.py

@@ -50,9 +50,13 @@ def __init__(
             system_prompt: System prompt defining steering guidance rules
             prompt_mapper: Custom prompt mapper for evaluation prompts
             model: Optional model override for steering evaluation
-            context_providers: List of context providers for populating steering context
+            context_providers: List of context providers for populating steering context.
+                Defaults to [LedgerProvider()] if None. Pass an empty list to disable
+                context providers.
         """
-        providers = context_providers or [LedgerProvider()]
+        providers: list[SteeringContextProvider] = (
+            [LedgerProvider()] if context_providers is None else context_providers
+        )
         super().__init__(context_providers=providers)
         self.system_prompt = system_prompt
         self.prompt_mapper = prompt_mapper or DefaultPromptMapper()

tests/strands/agent/test_agent_result.py

@@ -370,4 +370,3 @@ def test__str__empty_interrupts_returns_agent_message(mock_metrics, simple_messa
 
     # Empty list is falsy, should fall through to text content
     assert message_string == "Hello world!\n"
-

tests/strands/experimental/steering/core/test_handler.py

@@ -95,7 +95,7 @@ async def steer_before_tool(self, *, agent, tool_use, **kwargs):
     await handler._provide_tool_steering_guidance(event)
 
     # Should set cancel_tool with guidance message
-    expected_message = "Tool call cancelled given new guidance. Test guidance. Consider this approach and continue"
+    expected_message = "Tool call cancelled. Test guidance You MUST follow this guidance immediately."
     assert event.cancel_tool == expected_message
 
 

tests_integ/steering/test_tool_steering.py

@@ -75,8 +75,16 @@ def test_agent_with_tool_steering_e2e():
     """End-to-end test of agent with steering handler guiding tool choice."""
     handler = LLMSteeringHandler(
         system_prompt=(
-            "When agents try to use send_email, guide them to use send_notification instead for better delivery."
-        )
+            "CRITICAL INSTRUCTION - READ CAREFULLY:\n\n"
+            "You are a steering agent. Your ONLY job is to decide based on the tool name.\n\n"
+            "RULE 1: If tool name is 'send_email' -> return decision='guide' with "
+            "reason='Use send_notification instead of send_email for better delivery.'\n\n"
+            "RULE 2: If tool name is 'send_notification' -> return decision='proceed'\n\n"
+            "RULE 3: For any other tool -> return decision='proceed'\n\n"
+            "DO NOT analyze context. DO NOT consider arguments. ONLY look at the tool name.\n"
+            "The tool name in this request is the ONLY thing that matters."
+        ),
+        context_providers=[],  # Disable ledger to avoid confusing context
     )
 
     agent = Agent(tools=[send_email, send_notification], hooks=[handler])