The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses surge-cli to perform administrative network operations, including stopping the proxy service (stop), terminating connections (kill), and performing unattended system upgrades (unattended-upgrade).- [REMOTE_CODE_EXECUTION]: The skill documents the script evaluate command, which allows the execution of arbitrary JavaScript code within the Surge runtime (JSC or WebView). This provides a primitive for dynamic code execution that could be exploited if an agent is tricked into running malicious script content.- [DATA_EXFILTRATION]: Commands such as dump request, dump traffic, and retrieve-data allow the agent to access sensitive network information, including HTTP request and response bodies. This data exposure could be leveraged for exfiltration if the agent is directed to send this information to an external endpoint.- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its processing of untrusted network data. * Ingestion points: Untrusted data enters the context through traffic monitoring commands like watch request and retrieve-data. * Boundary markers: There are no instructions or delimiters provided to prevent the agent from obeying instructions embedded within the captured network traffic. * Capability inventory: The skill has significant system-level impact through the surge-cli tool. * Sanitization: No validation or filtering of the captured network content is performed before it is processed by the agent.

surge