Preserve tool_result ordering before reminders

agents/s03_todo_write.py

@@ -186,7 +186,7 @@ def agent_loop(messages: list):
                     used_todo = True
         rounds_since_todo = 0 if used_todo else rounds_since_todo + 1
         if rounds_since_todo >= 3:
-            results.insert(0, {"type": "text", "text": "<reminder>Update your todos.</reminder>"})
+            results.append({"type": "text", "text": "<reminder>Update your todos.</reminder>"})
         messages.append({"role": "user", "content": results})
 
 

agents/s_full.py

@@ -698,7 +698,7 @@ def agent_loop(messages: list):
         # s03: nag reminder (only when todo workflow is active)
         rounds_without_todo = 0 if used_todo else rounds_without_todo + 1
         if TODO.has_open_items() and rounds_without_todo >= 3:
-            results.insert(0, {"type": "text", "text": "<reminder>Update your todos.</reminder>"})
+            results.append({"type": "text", "text": "<reminder>Update your todos.</reminder>"})
         messages.append({"role": "user", "content": results})
         # s06: manual compress
         if manual_compress:

tests/test_tool_result_ordering.py

@@ -0,0 +1,110 @@
+import os
+import sys
+import types
+import unittest
+from pathlib import Path
+from types import SimpleNamespace
+
+
+REPO_ROOT = Path(__file__).resolve().parents[1]
+if str(REPO_ROOT) not in sys.path:
+    sys.path.insert(0, str(REPO_ROOT))
+
+os.environ.setdefault("MODEL_ID", "test-model")
+
+fake_anthropic = types.ModuleType("anthropic")
+
+
+class FakeAnthropic:
+    def __init__(self, *args, **kwargs):
+        self.messages = SimpleNamespace(create=None)
+
+
+setattr(fake_anthropic, "Anthropic", FakeAnthropic)
+sys.modules.setdefault("anthropic", fake_anthropic)
+
+fake_dotenv = types.ModuleType("dotenv")
+setattr(fake_dotenv, "load_dotenv", lambda *args, **kwargs: None)
+sys.modules.setdefault("dotenv", fake_dotenv)
+
+import agents.s03_todo_write as s03_todo_write
+import agents.s_full as s_full
+
+
+class FakeMessagesAPI:
+    def __init__(self, responses):
+        self._responses = iter(responses)
+
+    def create(self, **kwargs):
+        return next(self._responses)
+
+
+def make_tool_use_response(tool_id: str, tool_name: str, tool_input: dict):
+    return SimpleNamespace(
+        stop_reason="tool_use",
+        content=[
+            SimpleNamespace(
+                type="tool_use", id=tool_id, name=tool_name, input=tool_input
+            )
+        ],
+    )
+
+
+class ToolResultOrderingTests(unittest.TestCase):
+    def test_s03_places_tool_results_before_reminders(self):
+        messages = [{"role": "user", "content": "do work"}]
+        fake_api = FakeMessagesAPI(
+            [
+                make_tool_use_response("tool-1", "bash", {"command": "pwd"}),
+                make_tool_use_response("tool-2", "bash", {"command": "pwd"}),
+                make_tool_use_response("tool-3", "bash", {"command": "pwd"}),
+                SimpleNamespace(stop_reason="end_turn", content="done"),
+            ]
+        )
+        original_client = s03_todo_write.client
+        original_handlers = s03_todo_write.TOOL_HANDLERS
+        try:
+            s03_todo_write.client = SimpleNamespace(messages=fake_api)
+            s03_todo_write.TOOL_HANDLERS = {
+                **original_handlers,
+                "bash": lambda **kwargs: "ok",
+            }
+            s03_todo_write.agent_loop(messages)
+        finally:
+            s03_todo_write.client = original_client
+            s03_todo_write.TOOL_HANDLERS = original_handlers
+
+        third_user_message = messages[-2]["content"]
+        self.assertEqual(third_user_message[0]["type"], "tool_result")
+        self.assertEqual(third_user_message[-1]["type"], "text")
+
+    def test_s_full_places_tool_results_before_reminders(self):
+        messages = [{"role": "user", "content": "do work"}]
+        fake_api = FakeMessagesAPI(
+            [
+                make_tool_use_response("tool-1", "bash", {"command": "pwd"}),
+                make_tool_use_response("tool-2", "bash", {"command": "pwd"}),
+                make_tool_use_response("tool-3", "bash", {"command": "pwd"}),
+                SimpleNamespace(stop_reason="end_turn", content="done"),
+            ]
+        )
+        original_client = s_full.client
+        original_handlers = s_full.TOOL_HANDLERS
+        original_has_open_items = s_full.TODO.has_open_items
+        try:
+            s_full.client = SimpleNamespace(messages=fake_api)
+            s_full.TOOL_HANDLERS = {**original_handlers, "bash": lambda **kwargs: "ok"}
+            s_full.TODO.has_open_items = lambda: True
+            s_full.agent_loop(messages)
+        finally:
+            s_full.client = original_client
+            s_full.TOOL_HANDLERS = original_handlers
+            s_full.TODO.has_open_items = original_has_open_items
+
+        third_user_message = messages[-2]["content"]
+        self.assertEqual(third_user_message[0]["type"], "tool_result")
+        self.assertEqual(third_user_message[-1]["type"], "text")
+
+
+if __name__ == "__main__":
+    unittest.main()