CVE-2026-110206: Roo Code command approval bypass via command substitution

CVE-2026-110206 · Severity: high · CVSS 8.8 · Published 2026-10-13

Roo Code is an AI coding assistant that automatically approves and executes shell commands when the "always approve" setting is enabled. An attacker can bypass the approval allowlist by wrapping a malicious command in shell command substitution syntax (such as $(echo malicious_command)), causing the system to approve the inner harmless command but execute the malicious output instead, achieving arbitrary command execution on the developer's machine.

Technologies: Roo Code Inc Roo Code, Roocode Roo Code. Vendors: Roo Code Inc, Roocode.

Quick answers

Is CVE-2026-110206 exploited in the wild?
No. Junglewise Threat Intelligence has no report of CVE-2026-110206 being exploited in the wild.
How severe is CVE-2026-110206?
High severity, with a CVSS score of 8.8, published 13 October 2026.
Which products and versions are affected?
Roo Code Inc Roo Code through 3.54.0.
Is there a fix for CVE-2026-110206?
No fixed version is recorded yet. Check the vendor's advisory in the references below. If you run it, Junglewise drafts a mitigation that fits your setup and applies it once you approve, with senior security engineers behind it.
How is CVE-2026-110206 attacked?
It is attacked over the network.
Am I affected by CVE-2026-110206?
Check whether you run Roo Code Inc Roo Code or Roocode Roo Code at an affected version. Junglewise does that check for you around the clock: connect your code, cloud and domains, and it matches new vulnerabilities like this one against what you actually run and tells you when one reaches you.

Technical details

The vulnerability is a logic error in the command approval parser (parseCommandLine() and getCommandDecision()): when a command substitution like $(echo touch /tmp/x) appears at command position, the approval check extracts and validates only the inner command (echo touch /tmp/x), which matches the echo allowlist. However, bash executes the substitution by first running the inner command and then executing its standard output as a new command, so the attacker's touch /tmp/x is executed without approval. This represents a fundamental gap in the approval model's understanding of shell semantics. The vulnerability requires that "always approve execute" be enabled and that the attacker can inject commands into the agent's prompt (indirect prompt injection).

Affected products

  • Roo Code Inc Roo Code through 3.54.0

Timeline

  • 2026-10-13: disclosed

References

This page as Markdown. Free to reuse under CC BY 4.0 with a link back.