The Agent Skills Directory

COMMAND_EXECUTION (MEDIUM): The file scripts/connections.py includes the MCPConnectionStdio class, which utilizes the mcp library to execute local system commands via the stdio_client. This allows for the execution of arbitrary commands on the host system. If an attacker successfully uses prompt injection to control the command or args passed to the create_connection factory, they could execute malicious code with the permissions of the agent.
EXTERNAL_DOWNLOADS (MEDIUM): The MCPConnectionSSE and MCPConnectionHTTP classes in scripts/connections.py enable the agent to establish network connections to arbitrary URLs. This capability could be leveraged for Server-Side Request Forgery (SSRF) or data exfiltration if the agent is directed to connect to an attacker-controlled endpoint.

mcp-builder