The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes common macOS system commands like ifconfig, netstat, scutil, lsof, and ps to gather evidence regarding network interfaces, routing tables, and process-level network activity. These operations are necessary to identify VPN interference or transport protocol fallbacks.
[DYNAMIC_EXECUTION]: The skill employs dynamic execution for high-fidelity diagnostics, including: a standalone Python script scripts/probe_rdp_server.py to test the RDP stack and TLS layer; a Swift snippet to enumerate application windows via CGWindowListCopyWindowInfo; and Python one-liners for STUN connectivity checks and command encoding.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a potential indirect prompt injection surface by ingesting and parsing external application logs. Ingestion points: Local application logs found in ~/Library/Containers/com.microsoft.rdc.macos/Data/Library/Logs/Windows App/. Boundary markers: The skill mitigates risks by using activity GUIDs and specific grep patterns to delimit session data. Capability inventory: Capabilities include command execution (Bash), socket operations (python3), and window state inspection (swift). Sanitization: The skill relies on strict pattern matching to extract relevant data points rather than processing the entire log content as natural language instructions.
[REMOTE_CODE_EXECUTION]: The skill includes a procedure for executing PowerShell commands on a remote Windows host via SSH using Base64-encoded command strings. This capability is used exclusively for correlating client-side connection drops with server-side events such as system reboots or Windows Updates.

windows-remote-desktop-connection-doctor