diff --git a/src/smallestai/atoms/crew/nodes/output_crew.py b/src/smallestai/atoms/crew/nodes/output_crew.py
index 0f4b860d..bfb86dab 100644
--- a/src/smallestai/atoms/crew/nodes/output_crew.py
+++ b/src/smallestai/atoms/crew/nodes/output_crew.py
@@ -75,16 +75,10 @@ async def _update_settings(self, settings: Dict[str, Any]):
             logger.debug(f"[{self.name}] Updating setting {key} to {value}")
             setattr(self.settings, key, value)
 
-    async def process_event(self, event: SDKEvent):
-        """
-        Route events to appropriate handlers.
-
-        Handles:
-        - LLMRequestEvent -> Start generation
-        """
-
-        await super().process_event(event)
-
+    async def _route_framework_event(self, event: SDKEvent):
+        """Framework-owned event routing. Always runs, independent of any user
+        `on_event` override, so a subclass can't accidentally silence the agent
+        (e.g. by dropping the LLM-request -> generate_response path)."""
         if isinstance(event, SDKSystemLLMRequestEvent):
             await self._handle_llm_request()
         elif isinstance(event, SDKAgentTranscriptUpdateEvent):
@@ -92,6 +86,20 @@ async def process_event(self, event: SDKEvent):
         elif isinstance(event, SDKSystemUpdateOutputAgentSettingsEvent):
             await self._update_settings(event.settings)
 
+    async def on_event(self, event: SDKEvent):
+        """Override this to react to events. This is the safe extension point:
+        the framework routing (interrupts, LLM requests, transcript/context and
+        settings updates) already ran before this is called, so you do NOT need
+        to call super(). Prefer this over overriding `process_event`, which
+        carries that framework routing — dropping it silences the agent."""
+        pass
+
+    async def process_event(self, event: SDKEvent):
+        """Framework dispatch: interrupt handling, output-node routing, the user
+        `on_event` hook, then forward. Prefer overriding `on_event` over this."""
+        await super().process_event(event)
+        await self._route_framework_event(event)
+        await self.on_event(event)
         await self.send_event(event)
 
     async def speak(self, text: str):
diff --git a/tests/custom/test_output_node_on_event_hook.py b/tests/custom/test_output_node_on_event_hook.py
new file mode 100644
index 00000000..fa077ce3
--- /dev/null
+++ b/tests/custom/test_output_node_on_event_hook.py
@@ -0,0 +1,63 @@
+"""OutputCrewNode: framework routing must run even when a subclass overrides the
+user hook without super(). Overriding `on_event` (the documented extension point)
+must NOT silence the LLM-request -> generate_response path, and existing
+`process_event`-with-super() overrides must keep working (backward compat)."""
+import unittest
+from unittest import mock
+
+from smallestai.atoms.crew.nodes import OutputCrewNode
+from smallestai.atoms.crew.events import SDKSystemLLMRequestEvent
+
+
+class _OnEventAgent(OutputCrewNode):
+    """Recommended pattern: override on_event, no super()."""
+    def __init__(self, order):
+        super().__init__(name="a")
+        self._order = order
+
+    async def on_event(self, event):
+        self._order.append("on_event")
+
+    async def generate_response(self):
+        if False:
+            yield ""
+
+
+class _LegacyProcessEventAgent(OutputCrewNode):
+    """Legacy pattern: override process_event and call super()."""
+    def __init__(self, order):
+        super().__init__(name="b")
+        self._order = order
+
+    async def process_event(self, event):
+        await super().process_event(event)
+        self._order.append("legacy_process_event")
+
+    async def generate_response(self):
+        if False:
+            yield ""
+
+
+class OnEventHookTest(unittest.IsolatedAsyncioTestCase):
+    async def test_on_event_override_keeps_framework_routing(self):
+        order = []
+        a = _OnEventAgent(order)
+        a._handle_llm_request = mock.AsyncMock(side_effect=lambda: order.append("llm"))
+        a.send_event = mock.AsyncMock()
+        await a.process_event(SDKSystemLLMRequestEvent())
+        self.assertIn("llm", order, "framework LLM routing must run")
+        self.assertIn("on_event", order, "user hook must run")
+        self.assertLess(order.index("llm"), order.index("on_event"), "framework runs before user hook")
+
+    async def test_legacy_process_event_with_super_still_routes(self):
+        order = []
+        b = _LegacyProcessEventAgent(order)
+        b._handle_llm_request = mock.AsyncMock(side_effect=lambda: order.append("llm"))
+        b.send_event = mock.AsyncMock()
+        await b.process_event(SDKSystemLLMRequestEvent())
+        self.assertIn("llm", order, "backward-compat: super() still routes")
+        self.assertIn("legacy_process_event", order)
+
+
+if __name__ == "__main__":
+    unittest.main()