The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides numerous executable shell snippets for offensive tools such as impacket, crackmapexec, and evil-winrm. These commands allow the agent to perform remote command execution, lateral movement, and system-level manipulation on target machines.
[DATA_EXFILTRATION]: The instructions facilitate the systematic harvesting of sensitive domain data. Examples include dumping the Active Directory database (--ntds), extracting LSA secrets (impacket-secretsdump), and capturing network authentication hashes via protocol poisoning with responder.
[CREDENTIALS_UNSAFE]: The skill extensively uses and manipulates highly sensitive authentication tokens. This includes techniques for Pass-the-Hash, Kerberoasting, and the creation of Golden and Silver Tickets, which grant persistent and unauthorized access to entire network domains.
[EXTERNAL_DOWNLOADS]: The skill references external pivoting tools like chisel and ligolo-ng. These are typically external binaries that must be downloaded to a target environment to establish reverse tunnels and bypass network security boundaries.
[PROMPT_INJECTION]: The metadata and instructional content focus entirely on bypassing security controls and performing unauthorized exploitation, which functions as a task-specific override for general safety principles regarding network security and harmful activities.
[COMMAND_EXECUTION]: The skill includes commands for establishing network tunnels and SOCKS proxies (e.g., chisel, ssh -D), which can be used to bypass firewalls and exfiltrate data from internal networks.
[PROMPT_INJECTION]: The ingestion of untrusted network data via tools like bettercap (sniffing) and responder (poisoning) presents an indirect injection surface where malicious network traffic could potentially influence agent behavior without proper sanitization or boundaries.

network-attack