CVE-2026-85694: LaVague 0.2.35 Remote Code Execution via eval extraction
LaVague 0.2.35 contains a remote code execution vulnerability in PythonFromMarkdownExtractor.extractasobject that evaluates untrusted language model output derived from web page content. Attackers can inject malicious Python code through web pages using indirect prompt injection to execute arbitrary code on the operator's host without review.
Affected Software
Event History
Frequently Asked Questions
Who is exposed to this issue?
Operators using LaVague 0.2.35 are exposed when PythonFromMarkdownExtractor.extract_as_object processes language model output derived from web page content. An attacker can influence that content through a malicious web page.
Does exploitation require an authenticated account or user interaction?
No authenticated account or separate user interaction is required according to the provided vector. Exploitation depends on inducing the affected workflow to process attacker-controlled web content and its resulting language model output.
What is the impact if exploitation succeeds?
Successful exploitation can execute arbitrary Python code on the operator's host. The reported impact includes compromise of confidentiality, integrity, and availability.
How can I determine whether my deployment is affected?
Check whether the deployment uses LaVague 0.2.35 and invokes PythonFromMarkdownExtractor.extract_as_object on output derived from web pages. Such use of untrusted model output reaches the vulnerable evaluation behavior.