A code injection vulnerability in QWED-AI qwed-verification allows authenticated attackers to execute arbitrary code, potentially compromising the entire verification pipeline.

The infrastructure is vulnerable to code injection (CWE-94) due to improper control over the generation of code. An authenticated attacker with low privileges can leverage this flaw to influence or inject malicious code into the verification process.

Successful exploitation allows an attacker to gain unauthorized control over the verification infrastructure, potentially leading to the manipulation of LLM outputs or schema validation logic. With a CVSS score of 8.8, this vulnerability poses a severe threat to the integrity of AI-driven production environments.

Immediate Action: Update qwed-verification to version 5.1.2 or later to apply the necessary code injection protections.

Proactive Monitoring: Review logs for suspicious tool calls or unexpected code execution patterns within the verification agent state.

Compensating Controls: Apply strict input validation and sandboxing for all inputs processed by the verification infrastructure to prevent arbitrary code generation.

Public Exploit Available: No

Given the critical role of verification infrastructure in AI security, administrators must treat this vulnerability with high urgency. Upgrading to version 5.1.2 is the only definitive way to mitigate the risk of code injection and ensure the ongoing trustworthiness of your automated verification agents.