diff --git a/common/lib/xmodule/xmodule/capa_module.py b/common/lib/xmodule/xmodule/capa_module.py
index fb2ddfff27..ff717acdf6 100644
--- a/common/lib/xmodule/xmodule/capa_module.py
+++ b/common/lib/xmodule/xmodule/capa_module.py
@@ -658,6 +658,8 @@ class CapaDescriptor(RawDescriptor):
     # TODO (vshnayder): do problems have any other metadata?  Do they
     # actually use type and points?
     metadata_attributes = RawDescriptor.metadata_attributes + ('type', 'points')
+    
+    system_metadata_fields = RawDescriptor.system_metadata_fields + ['markdown']
 
     def get_context(self):
         _context = RawDescriptor.get_context(self)
diff --git a/common/lib/xmodule/xmodule/templates/problem/numericalresponse.yaml b/common/lib/xmodule/xmodule/templates/problem/numericalresponse.yaml
index 18b2af6d51..40c81d90e5 100644
--- a/common/lib/xmodule/xmodule/templates/problem/numericalresponse.yaml
+++ b/common/lib/xmodule/xmodule/templates/problem/numericalresponse.yaml
@@ -3,6 +3,19 @@ metadata:
     display_name: Numerical Response
     rerandomize: never
     showanswer: always
+    markdown: "A numerical response problem accepts a line of text input from the
+       student, and evaluates the input for correctness based on its
+       numerical value.  
+
+       The answer is correct if it is within a specified numerical tolerance
+       of the expected answer.
+
+       Enter the numerical value of Pi: = 3.14159 +- .02
+       
+       Enter the approximate value of 502*9: = 4518 +- 15%
+    
+       Enter the number of fingers on a human hand: = 5 +- 0
+        "
 data: |
    <problem>
    <p>