fix: harden orchestrator parsing defaults

tests/agents/test_orchestrator_chain.py

@@ -0,0 +1,17 @@
+import pytest
+
+from utu.agents.common import TaskRecorder
+from utu.agents.orchestrator.chain import ChainPlanner
+from utu.agents.orchestrator.common import Recorder
+
+
+def test_parse_reports_missing_plan_block() -> None:
+    planner = object.__new__(ChainPlanner)
+    recorder = Recorder(input="summarize the report")
+
+    with pytest.raises(AssertionError, match="No tasks parsed from plan"):
+        planner._parse("<analysis>the response was truncated</analysis>", recorder)
+
+
+def test_task_recorder_defaults_input_to_list() -> None:
+    assert TaskRecorder().input == []

utu/agents/common.py

@@ -92,7 +92,7 @@ def to_dict(self):
 class TaskRecorder(DataClassWithStreamEvents):
     task: str = ""
     trace_id: str = ""
-    input: str | list[TResponseInputItem] = field(default_factory=dict)
+    input: str | list[TResponseInputItem] = field(default_factory=list)
 
     # from RunResultStreaming
     final_output: str = ""

utu/agents/orchestrator/chain.py

@@ -82,7 +82,7 @@ def _parse(self, text: str, recorder: Recorder) -> Plan:
         analysis = match.group(1).strip() if match else ""
 
         match = re.search(r"<plan>\s*\[(.*?)\]\s*</plan>", text, re.DOTALL)
-        plan_content = match.group(1).strip()
+        plan_content = match.group(1).strip() if match else ""
         tasks: list[Task] = []
         task_pattern = r'\{"name":\s*"([^"]+)",\s*"task":\s*"([^"]+)"\s*\}'
         task_matches = re.findall(task_pattern, plan_content, re.IGNORECASE)