av-edr-evasion

Fail

Audited by Gen Agent Trust Hub on Sep 19, 2026

Risk Level: CRITICALEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPRIVILEGE_ESCALATIONPERSISTENCEDYNAMIC_EXECUTIONCREDENTIALS_UNSAFEOBFUSCATIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill references and provides instructions to download resources from external sites flagged as malicious or risky, including loldrivers.io and book.hacktricks.xyz. It also promotes the installation of offensive security tools from numerous non-trusted GitHub repositories.
  • [REMOTE_CODE_EXECUTION]: Multiple guides and the payloads collection provide specific instructions for downloading and executing remote code using built-in Windows utilities such as certutil, mshta, and msiexec, as well as through custom-built loaders.
  • [COMMAND_EXECUTION]: The skill contains a vast library of high-risk command strings designed to generate, encode, and inject malicious payloads into system processes.
  • [PRIVILEGE_ESCALATION]: The skill provides in-depth technical guides for kernel-level tampering, including Direct Kernel Object Manipulation (DKOM), unregistering security callbacks, and the 'Bring Your Own Vulnerable Driver' (BYOVD) technique to bypass system security boundaries.
  • [PERSISTENCE]: Instructions are provided for establishing stealthy persistence on a target system by abusing WMI event subscriptions and hijacking Component Object Model (COM) objects.
  • [DYNAMIC_EXECUTION]: The skill provides functional C# and PowerShell source code intended to be executed at runtime to patch memory buffers and disable security features like the Anti-Malware Scan Interface (AMSI) and Event Tracing for Windows (ETW).
  • [CREDENTIALS_UNSAFE]: Specific methodologies and commands are provided for dumping sensitive system credentials from the LSASS process and the Active Directory NTDS database.
  • [OBFUSCATION]: Several guide files contain base64-encoded binary headers (e.g., 'TVqQAAMAAAAEAAAA...') used as templates for reflectively loading executables in memory to evade static analysis.
  • [INDIRECT_PROMPT_INJECTION]: The skill creates a significant attack surface by instructing the agent to ingest data from untrusted external offensive security resources and then use that data to perform high-privilege operations such as file writing and command execution without boundary markers or sanitization.
Recommendations
  • CRITICAL: 5 infected file(s) detected - DO NOT USE
  • AI detected serious security threats
  • Contains 8 malicious URL(s) - DO NOT USE
Audit Metadata
Risk Level
CRITICAL
Analyzed
Sep 19, 2026, 04:36 PM