GHSA-p4h8-56qp-hpgv

Impact

A crafted hostAlias argument such as -oProxyCommand=... was passed to ssh/scp without an argument terminator. SSH interprets arguments starting with - as options regardless of position, so the option-injection caused SSH to execute the attacker-supplied ProxyCommand locally on the machine running the MCP server — before any network connection. This bypassed the documented protection of # @password: annotations and exposed local SSH keys, browser cookies, other MCP server credentials, and anything else readable by the server process.

A second local-RCE vector existed on Windows: spawn(..., { shell: true }) was used so that ssh.exe/scp.exe could be found via PATH. With shell: true, every argument is re-parsed by cmd.exe, so shell metacharacters (&, |, ^, >, ", ;, …) in hostAlias, command, localPath or remotePath would have been interpreted by cmd.exe and could have triggered arbitrary local command execution on Windows.

The MCP server runs locally over STDIO, but the LLM driving it is not trusted: its tool arguments can be steered by prompt injection from any untrusted text the LLM ingests (web pages, e-mails, repository files, output of other MCP servers). The attack does not require a malicious user — only that the LLM ingests attacker-controlled text at any point during the session.

Patches

Fixed in 1.3.5.

• Add -- argument terminator to all ssh/scp invocations.
• Strict whitelist for hostAlias (rejects leading - and shell metacharacters).
• Known-host check: every hostAlias must be defined in ~/.ssh/config (including Include directives) or present in ~/.ssh/known_hosts.
• Resolve ssh.exe/scp.exe to absolute paths and use shell: false everywhere on Windows.

Workarounds

None. Upgrade to 1.3.5.

Credit

Reported by Pico (@piiiico) as part of an MCP server security audit.