[bug] Arbitrary code execution in from_rail_string

[AlejandroEsquivel]

Describe the bug
It seems certain crafted RAIL specs and using from_rail_string can allow arbitrary code execution.

To Reproduce
Steps to reproduce the behavior:

from guardrails import Guard

railstring = """<rail version="0.1">
<output
type="string"
validators="test:{print('\n\narbitrary code execution\n\n')};"
/>
</rail>"""

Guard.from_rail_string(railstring)

Expected behavior
Prevent arbitrary code execution from RAIL specs

Library version:
0.5.0 And likely others

[CalebCourier]

Yes, this is a known issue but also a large concern we need to address before we ever offer a managed service. Since this always occurs on the client, it's slightly less serious than a traditional system, but I agree it should still be addressed. The limiting factor here is data. With RAIL we don't know what the arguments are intended to be which makes them more difficult to parse.
See this comment for a couple options for how to try to address this.
https://github.com/guardrails-ai/guardrails/blob/main/guardrails/utils/validator_utils.py#L30-L35

[edit]: update comment link