The Agent Skills Directory

[PERSISTENCE]: The skill explicitly defines the creation of persistent access mechanisms as a primary objective.
Evidence: SKILL.md lists "Persistent access mechanisms" in the Outputs/Deliverables section.
Evidence: references/advanced-attacks.md provides specific commands to create backdoor user accounts via Group Policy Objects (GPOs) and WSUS updates using instructions like net user backdoor Password123! /add.
[PRIVILEGE_ESCALATION]: The core functionality of the skill is to guide the user from low-privileged access to full Domain Administrator control.
Evidence: SKILL.md details DCSync attacks using secretsdump.py and forging Golden Tickets with kerberos::golden to gain high-level privileges.
Evidence: references/advanced-attacks.md explains how to abuse GPOs to grant local administrator rights using SharpGPOAbuse.exe --AddLocalAdmin and how to exploit unconstrained delegation.
[COMMAND_EXECUTION]: The skill instructs the agent to execute a wide variety of shell commands and scripts, including exploits for critical system vulnerabilities.
Evidence: SKILL.md provides commands for executing python3 cve-2020-1472-exploit.py (ZeroLogon) and python3 CVE-2021-1675.py (PrintNightmare).
[CREDENTIALS_UNSAFE]: The skill is focused on the unauthorized harvesting and reuse of authentication secrets.
Evidence: Provides comprehensive workflows for Kerberoasting, AS-REP Roasting, and Pass-the-Hash/Ticket attacks to extract hashes, TGS tickets, and TGTs from domain controllers and memory.
[EXTERNAL_DOWNLOADS]: The skill advocates for downloading and running numerous third-party offensive binaries and remote files from non-trusted sources.
Evidence: References binary tools such as SharpHound.exe, Mimikatz.exe, Rubeus.exe, SharpWSUS.exe, and MalSCCM.exe.
Evidence: SKILL.md contains an instruction to execute code from a remote network share: \\attacker\share\evil.dll.
[DATA_EXFILTRATION]: Provides techniques to harvest and export sensitive organizational data from Active Directory.
Evidence: Enumeration of users, domain controllers, and service principal names (SPNs) via tools like BloodHound, PowerView, and GetUserSPNs.py.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface where it ingests untrusted data from an Active Directory environment that could be manipulated to influence agent behavior.
Ingestion points: Processes user properties, group memberships, and SPN data via bloodhound-python and PowerView (SKILL.md).
Boundary markers: None identified.
Capability inventory: Extensive command execution via shell and file system access across all skill files.
Sanitization: None identified.

Active Directory Attacks