The Agent Skills Directory

Command Execution (MEDIUM): The MCPConnectionStdio class in scripts/connections.py executes system commands via the stdio_client from the mcp library. While this is the intended transport mechanism for local MCP servers, it allows for arbitrary command execution if the command or args parameters are sourced from untrusted user input without sanitization.- Data Exposure & Exfiltration (LOW): The MCPConnectionSSE and MCPConnectionHTTP classes in scripts/connections.py facilitate outbound network connections to arbitrary URLs. This capability could be leveraged for data exfiltration or connecting to malicious endpoints if the URLs are not restricted to a trusted whitelist.- Indirect Prompt Injection (LOW): The skill is susceptible to indirect prompt injection through tool outputs.
Ingestion points: scripts/connections.py (the call_tool method).
Boundary markers: Absent. Tool results are returned directly without encapsulation or instructions to the model to ignore embedded commands.
Capability inventory: scripts/connections.py (System command execution and network access).
Sanitization: Absent. Data from external tools is processed as-is.

mcp-builder