How Custom Response Grading Python Executes

<customresponse /> elements in "Advanced Problems" can be graded using python provided in <script type="loncapa/python"> tags, such as the example below which basically marks all responses as correct

<script type="loncapa/python">
import re
  
def short_response(expect, ans):
  response = re.search('', ans)
  if response:
  	return 1
  else:
  	return 0
</script>

But what kind of python code can be specified within these script tags? For example, what python libraries are available to this python code?

This code executes in a jailed environment to protect the servers. (Links for the software performing this enforcement, both of which have detailed readmes: codejail repo safe_exec in edx-platform/common/lib/capa.) First, this environment has a restricted python environment. The list of python packages available to this code is listed in the edx platform repo at requirements/edx-sandbox. In addition, python's standard libraries are available.

Also, python code specified in these script tags are subject to OS level restrictions provided by Ubuntu's apparmor package that, among other protections, limit execution time and memory consumption and prevent file and network access.