mediumcall_server_tool
Proxies arbitrary third-party tool execution with unfiltered result pass-through
This tool forwards the agent's call to any MCP server registered in a ~22,000-entry marketplace and returns the result 'content 原样透传' (verbatim). Since the actual tool code and its output are controlled by third parties not vetted by this review, a malicious or compromised listed server could return content containing prompt-injection instructions to the agent, or could exfiltrate whatever arguments (potentially sensitive user data) are passed to it. The gateway provides no described sanitization or sandboxing of returned content before it reaches the agent/user.
low
Not on gate's verified list
This server isn't on gate's handpicked list. That's not necessarily bad, but there's no third-party signal about it — review its tools carefully.
lowcall_server_tool
Broad, un-scoped capability by design
The tool can invoke essentially any tool exposed by any registered endpoint with arbitrary arguments, which is a very broad capability surface (arbitrary code paths on external services) relative to a generic 'call a tool' description. While this is the stated purpose of the marketplace gateway, it means the effective capability is unbounded and depends entirely on the trustworthiness of the 22k-entry directory, which is not independently verified per-call.
lowsearch_servers
Directs agent workflow toward unverified third-party services
The description instructs the agent to always use this marketplace search as the 'first step' for capabilities like weather/exchange rates, funneling tool selection toward a large, largely unaudited directory of external MCP servers rather than vetted/first-party tools, increasing exposure to malicious or low-quality listings.