CVE-2026-44461: Zed: Remote Command Injection via Unquoted Environment Variable Keys (SSH / WSL Remote)

Synopsis

Remote command injection in Zed, the code editor, allows an attacker who can influence environment variable keys in project terminal settings to inject shell metacharacters into an SSH or WSL remote command string. The attack requires local access to the victim's machine and a single user interaction (opening a terminal), but no authentication against Zed itself, and a successful exploit achieves arbitrary command execution on the remote host under the victim's account. No fix version has been published yet; HarborGuard is tracking the upstream advisory and will make a patched-image rebuild available as soon as one is released.

HarborGuard Coverage

Exploit Conditions

• Network reachabilityNot required

  The attacker needs an existing shell or process on the host; no over-the-network access to Zed itself is required to plant the malicious environment variable key.

• AuthenticationNot required

  No authentication credential against Zed or its settings interface is required; any party able to influence the project terminal settings can insert a malicious key.

• Victim interactionRequired

  The victim must open a remote terminal session in Zed, which triggers evaluation of the injected shell expression by the remote shell.

• Attack complexityDetail

  Exploit reliability is high and condition-free once the malicious key is in place; no race condition, memory-layout dependency, or environmental timing is needed.

Blast Radius

• Executes arbitrary shell commands on the remote host (SSH or WSL target) under the victim user's account, giving the attacker full control of any process the victim can spawn.
• Reads any file accessible to the victim user on the remote host, including source code, credentials, SSH private keys, and configuration secrets stored in the project directory or home directory.
• Writes or modifies files on the remote host under the victim's permissions, including source files, shell rc scripts, and SSH authorized_keys, enabling persistence.
• Crashes or disrupts processes running under the victim's account on the remote host by terminating services or corrupting state files.