AI agents call ssh_download_file to retrieve information from Ssh Agent without modifying anything — typically the context-gathering step in research, monitoring, and reporting workflows, before the agent takes action elsewhere.
While file download is a Read operation (retrieves data with no side effects), the severity is elevated to medium rather than low because: (1) the tool operates over SSH which grants broad system access, (2) an AI agent could misuse it to exfiltrate sensitive data, credentials, or intellectual property from a remote server, and (3) the context of the ssh-agent-mcp server shows the tool is part of a suite enabling…
From the tool's definition Tool name is 'ssh_download_file' and description indicates it 'downloads files from a remote server' (从远程服务器下载文件). This is fundamentally a retrieval operation that reads and transfers data without modifying or deleting anything on the remote system.
Attacks that exploit this kind of access
从远程服务器下载文件. It is categorised as a Read tool in the Ssh Agent MCP Server, which means it retrieves data without modifying state.
Register the Ssh Agent MCP server in PolicyLayer and add a rule for ssh_download_file: allow, deny, rate-limit, or require approval. Point your MCP client at the PolicyLayer proxy URL and the rule is enforced on every call, before it reaches Ssh Agent. Nothing to install.
ssh_download_file is a Read tool with low risk. Read-only tools are generally safe to allow by default.
Yes. Add a rate_limit block to the ssh_download_file rule in your PolicyLayer policy. For example, setting max: 10 and window: 60 limits the tool to 10 calls per minute. Rate limits are tracked per agent session and reset automatically.
Set action: deny in the PolicyLayer policy for ssh_download_file. The AI agent will receive a policy violation error and cannot call the tool. You can also include a reason field to explain why the tool is blocked.
ssh_download_file is provided by the Ssh Agent MCP server (zhijun/ssh-mcp). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.
Every MCP server has a record like this.
Type a name, get the same breakdown: verified identity, auth posture, risk grade, capabilities, recommended policy.
Teams ship this data inside their own products. See what a licence covers →