The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill contains multiple detailed procedures for bypassing User Account Control (UAC), including registry hijacking of auto-elevated binaries like fodhelper.exe, eventvwr.exe, computerdefaults.exe, and sdclt.exe. It also includes a "Mock Trusted Directory" technique that leverages a trailing-space directory flaw to spoof system paths.
[PRIVILEGE_ESCALATION]: Extensive instructions for token manipulation using the "Potato" family of exploits (JuicyPotato, GodPotato, PrintSpoofer) to escalate from low-privilege service accounts to NT AUTHORITY\SYSTEM.
[CREDENTIALS_UNSAFE]: Provides commands to harvest credentials by dumping the Local Security Authority Subsystem Service (LSASS) process using procdump and extracting the SAM, SYSTEM, and SECURITY registry hives for offline password cracking.
[DATA_EXFILTRATION]: Instructs on the use of SeBackupPrivilege to bypass standard access controls and read sensitive system files like NTDS.dit (Active Directory database) for extraction.
[PERSISTENCE]: Includes instructions for establishing persistence through the creation of unauthorized scheduled tasks (Register-ScheduledTask) and the modification of registry autorun keys (HKLM\SOFTWARE\Microsoft\Windows\CurrentVersion\Run).
[COMMAND_EXECUTION]: Outlines methods to hijack system services by reconfiguring the binpath of vulnerable services to execute attacker-supplied binaries.
[REMOTE_CODE_EXECUTION]: Details a technique using cmstp.exe to execute a remote .sct payload from an external URL (http://ATTACKER/sct_payload.sct) to bypass security prompts.
[DYNAMIC_EXECUTION]: Provides workflows for compiling malicious DLLs and generating malicious MSI installers using msfvenom to facilitate privilege escalation.
[INDIRECT_PROMPT_INJECTION]: The skill presents a high vulnerability surface for indirect prompt injection.
Ingestion points: The skill relies on parsing output from host enumeration commands like whoami, systeminfo, and sc query to determine the next stage of the attack (SKILL.md).
Boundary markers: None provided; the agent is not instructed to treat system outputs as untrusted data.
Capability inventory: The skill provides the agent with full shell execution, registry manipulation, service reconfiguration, and file system write capabilities across all included files.
Sanitization: None; the skill does not implement validation or sanitization of the enumeration data that drives its logic.

windows-privilege-escalation