Sshand MCP Server

This server (without client built-ins) exposes a complete data-exfiltration chain: ssh_download_file → ssh_read_file → ssh_run_command. Untrusted input is ingested, private data is read, and it can be sent to an external sink via the agent composing the tools (→). Static analysis proves the primitive exists, not that a specific run will occur.

Fix: Remove one leg of the trifecta: isolate untrusted-input tools from secret-reading tools and from egress tools, or require human approval between them.

Location: flow ssh_download_file → ssh_read_file → ssh_run_command

In the server's implementation (`host_config.py:62`): Reading private keys / cloud credentials, or serializing the whole environment, is a sensitive-data source that becomes exfiltration when combined with any egress. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: "(e.g., '~/.ssh/id_rsa', '/home/user/.ssh/deploy_key')." ), ) passphrase: str | None = Field(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server host_config.py

In the server's implementation (`manage.py:239`): Reading private keys / cloud credentials, or serializing the whole environment, is a sensitive-data source that becomes exfiltration when combined with any egress. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: path", default="~/.ssh/id_rsa", required=True) passphrase = questionary.password( "Key passphrase (

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server manage.py

In the server's implementation (`server.py:247`): Reading private keys / cloud credentials, or serializing the whole environment, is a sensitive-data source that becomes exfiltration when combined with any egress. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: (e.g., '~/.ssh/id_rsa'). key_passphrase: Optional passphrase to decrypt an encrypted private key.

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server server.py

In the server's implementation (`setup_wizard.py:375`): Reading private keys / cloud credentials, or serializing the whole environment, is a sensitive-data source that becomes exfiltration when combined with any egress. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: vate key path", "~/.ssh/id_rsa") passphrase = _ask("Key passphrase (leave blank if none)") auth = hc.

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server setup_wizard.py

Tool "ssh_run_command" appears to run shell commands or evaluate code (keyword "run_command" in tool name). Arbitrary execution driven by model input is one of the most dangerous MCP capabilities; combined with any untrusted input it becomes RCE.

Fix: Sandbox execution, allowlist commands/arguments, and never pass model output to a shell unescaped.

Location: tool ssh_run_command

In the server's implementation (`platform_utils.py:67`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: try: result = subprocess.run( ["sc", "query", "ssh-agent"], capture_output=True,

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server platform_utils.py

Tool "ssh_write_file" can write, overwrite or delete files (keyword "write_file" in tool name). Verify it is scoped to a safe directory.

Fix: Constrain file operations to an explicit, non-sensitive root; reject path traversal.

Location: tool ssh_write_file

In a packaging/dev/install script (shipped, but not the server runtime) (`tests/test_host_config.py:148`): Reading private keys / cloud credentials, or serializing the whole environment, is a sensitive-data source that becomes exfiltration when combined with any egress. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: key_file = ssh_dir / "id_rsa" key_file.write_text("-----BEGIN OPENSSH PRIVATE KEY-----\n...") monkeypatch.set

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server tests/test_host_config.py

In a packaging/dev/install script (shipped, but not the server runtime) (`tests/test_ssh_client.py:112`): Reading private keys / cloud credentials, or serializing the whole environment, is a sensitive-data source that becomes exfiltration when combined with any egress. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: key_path = key_file / "id_rsa" key_path.write_text("-----BEGIN OPENSSH PRIVATE KEY-----\n...") return str(key_p

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server tests/test_ssh_client.py

Tool "ssh_run_command" can mutate/egress but declares no destructiveHint. Clients that don't default to spec-safe behavior may not prompt before running it.

Fix: Declare accurate annotations, and gate destructive tools on user confirmation regardless.

Location: tool ssh_run_command

Tool "ssh_write_file" can mutate/egress but declares no destructiveHint. Clients that don't default to spec-safe behavior may not prompt before running it.

Fix: Declare accurate annotations, and gate destructive tools on user confirmation regardless.

Location: tool ssh_write_file