Roo Code Multi-Line Command Injection Vulnerability Allowing Bypass of Execution Allow-List

A command injection vulnerability has been identified in Roo Code versions prior to 3.23.19. The issue arises because the application fails to properly validate line breaks in command input, which can lead to a bypass of the allow-list mechanism. This vulnerability allows for multi-line command injection, where only the first line of a command is considered during execution. As a result, attackers can inject additional commands on subsequent lines, potentially executing disallowed actions.

Impact

Exploiting this vulnerability allows users to execute commands that are not on the allow-list, bypassing command execution restrictions. This could lead to unauthorized actions being performed within the user's environment.

Reproduction

To reproduce this vulnerability, input a command that includes a line break followed by an additional command on the next line. For example, if 'tsc' is allow-listed, input 'tsc' followed by a line break and 'python3 ./ex.py'. The second command will be executed despite not being allow-listed.

Remediation

Users can update to Roo Code version 3.23.19 or later to address this vulnerability.