diff --git a/scripts/release.py b/scripts/release.py
index 9e7a46be3..55a16edb7 100755
--- a/scripts/release.py
+++ b/scripts/release.py
@@ -46,6 +46,7 @@ ACP_REGISTRY_MANIFEST = REPO_ROOT / "acp_registry" / "agent.json"
 # Auto-extracted from noreply emails + manual overrides
 AUTHOR_MAP = {
     "phanvanhoa@gmail.com": "theAgenticBuilder",  # PR #14180 salvage (route delegate_task progress lines through _safe_print so ACP stdio JSON-RPC frames stay clean)
+    "huangxudong663@gmail.com": "huangxudong663-sys",  # PR #15157 salvage (isinstance(dict) guard on tool-call model_extra; NVIDIA NIM non-dict crash)
     "cypher@augmentl.com": "Nickperillo",  # PR #8008 salvage (Discord channel-name matching + flush pending sends on shutdown)
     "tenoryang@outlook.com": "MarioYounger",  # PR #9028 salvage (bash/sh heredoc approval, NFKC homograph fold, execute_code CREDS/BEARER/APIKEY env filter)
     "peet.wannasarnmetha@gmail.com": "peetwan",  # PR #51841 salvage (loopback ws-ping tuning + token-frame coalescing + loop heartbeat; #48445/#50005)
diff --git a/tests/agent/test_model_extra_type_guard.py b/tests/agent/test_model_extra_type_guard.py
new file mode 100644
index 000000000..aeb37334f
--- /dev/null
+++ b/tests/agent/test_model_extra_type_guard.py
@@ -0,0 +1,89 @@
+"""Regression test for PR #15157: non-dict ``model_extra`` must not crash
+tool-call normalization.
+
+Some OpenAI-compatible providers (observed with NVIDIA NIM + qwen3.5) return a
+**string** for ``model_extra`` on a tool call instead of the dict that pydantic
+``BaseModel.model_extra`` produces.  The original extraction used a falsy
+fallback::
+
+    extra = (tc.model_extra or {}).get("extra_content")
+
+A non-empty string is truthy, so ``(string or {})`` evaluates to the string and
+``.get(...)`` raises ``AttributeError: 'str' object has no attribute 'get'`` —
+which propagates out of ``normalize_response`` and turns every tool call into an
+``[error]``.  The fix replaces the falsy fallback with an explicit
+``isinstance(.., dict)`` guard, so any non-dict ``model_extra`` is treated as
+"no extra_content".
+
+These tests exercise the real ``ChatCompletionsTransport.normalize_response``
+path (the non-streaming site at chat_completions.py), not a local replica of
+the expression.
+"""
+
+from __future__ import annotations
+
+from types import SimpleNamespace
+
+from agent.transports.chat_completions import ChatCompletionsTransport
+
+
+def _make_response(*, model_extra, extra_content_missing=True):
+    """Build a fake OpenAI ChatCompletion with one tool call.
+
+    ``extra_content`` is left unset so ``getattr(tc, "extra_content", None)``
+    returns None and the code falls through to the ``model_extra`` branch —
+    the path that crashed.
+    """
+    func = SimpleNamespace(name="get_weather", arguments='{"city": "SF"}')
+    tc = SimpleNamespace(id="call_1", function=func, model_extra=model_extra)
+    # Deliberately do NOT set tc.extra_content — getattr default None triggers
+    # the model_extra branch.
+    message = SimpleNamespace(
+        tool_calls=[tc],
+        content=None,
+        reasoning=None,
+        reasoning_content=None,
+        reasoning_details=None,
+    )
+    choice = SimpleNamespace(message=message, finish_reason="tool_calls")
+    return SimpleNamespace(choices=[choice], usage=None, model="qwen3.5-397b")
+
+
+class TestModelExtraTypeGuard:
+    """The guard must tolerate every shape of ``model_extra``."""
+
+    def _normalize(self, model_extra):
+        transport = ChatCompletionsTransport()
+        return transport.normalize_response(_make_response(model_extra=model_extra))
+
+    def test_string_model_extra_does_not_crash(self):
+        """A truthy string used to raise AttributeError — must be ignored now."""
+        result = self._normalize("unexpected_string")
+        assert len(result.tool_calls) == 1
+        # No extra_content recovered from a non-dict — provider_data stays empty.
+        assert result.tool_calls[0].provider_data is None
+
+    def test_none_model_extra(self):
+        result = self._normalize(None)
+        assert len(result.tool_calls) == 1
+        assert result.tool_calls[0].provider_data is None
+
+    def test_list_model_extra_does_not_crash(self):
+        """Any non-dict (list) is tolerated, not just strings."""
+        result = self._normalize([1, 2, 3])
+        assert len(result.tool_calls) == 1
+        assert result.tool_calls[0].provider_data is None
+
+    def test_dict_model_extra_still_extracts_extra_content(self):
+        """The valid dict path must keep working — extra_content preserved."""
+        result = self._normalize({"extra_content": {"thought_signature": "sig"}})
+        assert len(result.tool_calls) == 1
+        assert result.tool_calls[0].provider_data == {
+            "extra_content": {"thought_signature": "sig"}
+        }
+
+    def test_dict_without_extra_content_key(self):
+        """A dict that has no extra_content key yields no provider_data."""
+        result = self._normalize({"other_key": "value"})
+        assert len(result.tool_calls) == 1
+        assert result.tool_calls[0].provider_data is None