fix: add <thought> stripping to auxiliary_client + tests

auxiliary_client.py had its own regex mirroring _strip_think_blocks
but was missing the <thought> variant. Also adds test coverage for
<thought> paired and orphaned tags.

agent/auxiliary_client.py

@@ -2454,9 +2454,9 @@ def extract_content_or_reasoning(response) -> str:
     if content:
         # Strip inline think/reasoning blocks (mirrors _strip_think_blocks)
         cleaned = re.sub(
-            r"<(?:think|thinking|reasoning|REASONING_SCRATCHPAD)>"
+            r"<(?:think|thinking|reasoning|thought|REASONING_SCRATCHPAD)>"
             r".*?"
-            r"</(?:think|thinking|reasoning|REASONING_SCRATCHPAD)>",
+            r"</(?:think|thinking|reasoning|thought|REASONING_SCRATCHPAD)>",
             "", content, flags=re.DOTALL | re.IGNORECASE,
         ).strip()
         if cleaned:

tests/run_agent/test_run_agent.py

@@ -302,6 +302,17 @@ class TestStripThinkBlocks:
         assert "<think>" not in result
         assert "visible" in result
 
+    def test_thought_block_removed(self, agent):
+        """Gemma 4 uses <thought> tags for inline reasoning."""
+        result = agent._strip_think_blocks("<thought>internal reasoning</thought> answer")
+        assert "internal reasoning" not in result
+        assert "<thought>" not in result
+        assert "answer" in result
+
+    def test_orphaned_thought_tag(self, agent):
+        result = agent._strip_think_blocks("<thought>orphaned reasoning without close")
+        assert "<thought>" not in result
+
 
 class TestExtractReasoning:
     def test_reasoning_field(self, agent):