active-directory-attacks

Warn

Audited by Gen Agent Trust Hub on Feb 27, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTIONCREDENTIALS_UNSAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes 'sudo' to modify system time ('sudo date -s ...'), which requires elevated system privileges.
  • [COMMAND_EXECUTION]: Provides instructions to establish persistence by abusing Group Policy Objects (GPOs) to add local administrators or create a 'backdoor' user account ('net user backdoor Password123! /add').
  • [REMOTE_CODE_EXECUTION]: Includes exploitation examples like PrintNightmare (CVE-2021-1675) that load and execute DLL files from remote network shares ('\\attacker\share\evil.dll').
  • [REMOTE_CODE_EXECUTION]: References multiple third-party exploit scripts (e.g., 'sam_the_admin.py', 'cve-2020-1472-exploit.py', 'pygpoabuse.py') for execution without verified sources.
  • [DATA_EXFILTRATION]: Facilitates extraction of sensitive domain data, including 'krbtgt' hashes via DCSync and Kerberos tickets for offline cracking.
  • [PROMPT_INJECTION]: Ingests and processes untrusted data from Active Directory (LDAP) and GPO files, creating an indirect prompt injection surface.
  • Ingestion points: SharpHound (LDAP), GPO XML parsing.
  • Boundary markers: Absent.
  • Capability inventory: High-privilege command execution (psexec, wmiexec), file writing.
  • Sanitization: Absent.
  • [CREDENTIALS_UNSAFE]: Hardcodes plain-text passwords (e.g., 'Password123!') within command examples for backdoor account creation and certificate usage.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 27, 2026, 05:24 PM