diff --git a/openagent_eval/exceptions/diagnosis.py b/openagent_eval/exceptions/diagnosis.py
index 7cfb1a0..d217d7a 100644
--- a/openagent_eval/exceptions/diagnosis.py
+++ b/openagent_eval/exceptions/diagnosis.py
@@ -37,7 +37,11 @@ def __init__(
             details: Optional dictionary with additional error context.
             step: The diagnosis step that failed (e.g., "blame_attribution").
         """
-        super().__init__(message, details)
+        error_details = details or {}
+        if step is not None:
+            error_details["step"] = step
+
+        super().__init__(message, error_details)
         self.step = step
 
 
diff --git a/tests/unit/test_exceptions.py b/tests/unit/test_exceptions.py
index 9e0c8bc..cb3747e 100644
--- a/tests/unit/test_exceptions.py
+++ b/tests/unit/test_exceptions.py
@@ -9,6 +9,7 @@
     DatasetError,
     DatasetNotFoundError,
     DatasetValidationError,
+    DiagnosisExecutionError,
     InvalidDatasetError,
     MetricError,
     MetricExecutionError,
@@ -130,6 +131,17 @@ def test_timeout_error_preserves_zero_timeout(self) -> None:
         assert error.details["timeout_seconds"] == 0.0
 
 
+class TestDiagnosisError:
+    """Tests for diagnosis errors."""
+
+    def test_execution_error_includes_step_in_details(self) -> None:
+        """Test that the failed step is visible in error details."""
+        error = DiagnosisExecutionError("Failed", step="blame_attribution")
+
+        assert error.step == "blame_attribution"
+        assert error.details["step"] == "blame_attribution"
+
+
 class TestProviderError:
     """Tests for provider errors."""