The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent to install and execute unverified software from external sources, specifically using pip install crowbar and referencing multiple third-party repositories on GitHub such as JoelGMSec/AutoRDPwn, skelsec/evilrdp, and 0xthirteen/SharpRDP.
[COMMAND_EXECUTION]: Includes a wide array of shell commands for network exploitation, system enumeration, and administrative modification, including nmap, hydra, xfreerdp, tscon, and net localgroup.
[PERSISTENCE]: Provides explicit instructions for creating persistence mechanisms on target systems, such as the Sticky Keys and Utilman backdoors, which involve replacing system executables (sethc.exe, utilman.exe) with cmd.exe.
[DATA_EXFILTRATION]: Outlines techniques for data tunneling and exfiltration using RDP virtual channels with tools like rdp2tcp.
[INDIRECT_PROMPT_INJECTION]: The skill represents an indirect injection surface as it processes external, potentially attacker-controlled data (e.g., target IPs, brute-force wordlists, or scan results) and possesses high-privilege capabilities including file modification and network access.
Ingestion points: Target IP addresses, username lists, and password dictionaries processed by scanning tools.
Boundary markers: None identified; untrusted data is passed directly to command arguments.
Capability inventory: Includes remote network communication, file system modification (reg query), and session manipulation (tscon).
Sanitization: No evidence of input validation or sanitization before interpolation into shell commands.

rdp-pentesting