A flaw in the Sixth tool’s automatic terminal command execution logic allows an attacker to inject a malicious command by disguising it as a safe command, causing the model to misclassify it and execute the command without user approval. The vulnerability stems from improper input validation of the model‑derived command classification, enabling the attacker to gain full control over the host system by running arbitrary commands.

The affected product is Sixth, an application that uses a language model to decide which terminal commands to run automatically. No specific vendor or product versions are listed, but any installation that relies on the automatic safe‑command feature is at risk.

The CVSS score is not provided, and the EPSS value is unavailable, however the vulnerability is not listed in CISA’s KEV catalog. Based on the description, the most likely attack vector is a prompt injection performed by the attacker to manipulate the model’s safety assessment. Once the malicious command is classified as safe, the tool executes it automatically, which can lead to a complete compromise of the system if the attacker runs privileged commands. Because no patch is currently documented, the risk remains significant until a vendor update is released.