fix(backend): tighten ValueError message and test assertion in _generate_schema

- Add stable prefix to ValueError: "Invalid graph schema input: {e}"
  to aid debugging and triage
- Add match="name" to pytest.raises to guard the specific missing-name
  failure path, not just any ValueError

autogpt_platform/backend/backend/data/graph.py

@@ -355,7 +355,7 @@ class BaseGraph(GraphBaseMeta):
                 "required": [p.name for p in schema_fields if p.value is None],
             }
         except AttributeError as e:
-            raise ValueError(str(e)) from e
+            raise ValueError(f"Invalid graph schema input: {e}") from e
 
 
 class GraphTriggerInfo(BaseModel):

autogpt_platform/backend/backend/data/graph_test.py

@@ -1477,5 +1477,5 @@ def test_generate_schema_raises_value_error_when_name_missing():
     and re-raised as ValueError so the existing 400 handler in rest_api.py fires
     instead of falling through to the 500 catch-all.
     """
-    with pytest.raises(ValueError):
+    with pytest.raises(ValueError, match="name"):
         GraphModel._generate_schema((AgentInputBlock.Input, {}))