The Agent Skills Directory

[PRIVILEGE_ESCALATION]: Provides numerous methods for escalating privileges to root or Administrator, including abusing sudo misconfigurations (e.g., vim, find, and awk escapes), SUID binaries, Linux capabilities, and Windows token impersonation techniques (e.g., SeImpersonatePrivilege, SeBackupPrivilege).
[REMOTE_CODE_EXECUTION]: Includes command patterns for downloading and executing remote scripts from external sources, specifically using PowerShell's Invoke-Expression (IEX) to execute a shell script from a remote URL.
[DATA_EXFILTRATION]: Outlines techniques for accessing and copying sensitive files such as SSH private keys (/root/.ssh/id_rsa), the Active Directory database (ntds.dit), and system registry hives, which can be used to harvest credentials.
[PERSISTENCE]: Details how to establish long-term unauthorized access using Linux cron jobs, Windows scheduled tasks, and the creation of Kerberos Golden Tickets in Active Directory environments.
[COMMAND_EXECUTION]: Contains instructions for performing high-risk system operations like mounting file systems, creating volume shadow copies to bypass file locks, and modifying domain user accounts.
[CREDENTIALS_UNSAFE]: Uses hardcoded example passwords (e.g., 'Password123', 'Password!@#') within command templates, which may encourage the use of weak or predictable credentials.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface where user-supplied targets are interpolated into powerful shell commands.
Ingestion points: User-provided target URLs, IP addresses, and account names.
Boundary markers: Includes a mandatory confirmation gate requiring the user to confirm authorization and scope before execution.
Capability inventory: Wide-ranging capabilities including sudo exploitation, remote script execution, and persistent credential dumping.
Sanitization: Lacks explicit instructions for sanitizing or validating user-provided target strings before they are incorporated into executable commands.

privilege-escalation-methods