fix: harden run-level hook lifecycle

maintainer edit: keep cancellation out of on_error so shutdown paths do not look like run failures, and let the SDK capture hook use the authoritative after_run snapshot.

nanobot/agent/hook.py

@@ -166,7 +166,9 @@ class SDKCaptureHook(AgentHook):
 
     The runner mutates ``context.messages`` in place across iterations, so the
     snapshot is refreshed on every ``after_iteration`` call; the last call
-    reflects the end-of-turn state the SDK caller cares about.
+    reflects the end-of-turn state the SDK caller cares about.  The run-level
+    snapshot is authoritative when available and covers paths without a final
+    per-iteration callback.
     """
 
     def __init__(self) -> None:
@@ -178,3 +180,7 @@ class SDKCaptureHook(AgentHook):
         for call in context.tool_calls:
             self.tools_used.append(call.name)
         self.messages = list(context.messages)
+
+    async def after_run(self, context: AgentRunHookContext) -> None:
+        self.tools_used = list(context.tools_used)
+        self.messages = list(context.messages)