active-directory-kerberos-attacks
Warn
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONPRIVILEGE_ESCALATIONPERSISTENCEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides numerous shell commands for offensive security tools including Impacket, Rubeus, Mimikatz, and PowerView. These commands are designed to interact with Active Directory services to perform sensitive operations like ticket requesting and hash cracking.
- [PRIVILEGE_ESCALATION]: The playbook details multiple paths to escalate privileges within a domain environment. This includes exploiting unconstrained, constrained, and resource-based constrained delegation, as well as forging Golden, Silver, Diamond, and Sapphire tickets to impersonate high-privileged users.
- [PERSISTENCE]: The instructions cover techniques for maintaining long-term access to a domain, such as the creation of Golden Tickets (using the krbtgt hash) and references to ADCS-based persistence (Golden Certificates).
- [DATA_EXFILTRATION]: The skill includes procedures for harvesting sensitive domain data. Specifically, it provides commands for AS-REP roasting, Kerberoasting, and performing DCSync attacks to extract NTLM hashes for all domain users from a Domain Controller.
- [INDIRECT_PROMPT_INJECTION]: The skill demonstrates an attack surface for indirect prompt injection by processing untrusted data from an Active Directory environment.
- Ingestion points: Commands in SKILL.md and KERBEROS_ATTACK_CHAINS.md ingest data from AD objects (users, SPNs, group memberships, ACLs).
- Boundary markers: The instructions do not define delimiters or warnings to ignore instructions embedded in the ingested data.
- Capability inventory: The skill uses high-privilege tools for command execution, file access (TGT/TGS storage), and network operations.
- Sanitization: There is no mention of sanitizing or validating the content retrieved from the AD environment before the agent acts upon it.
Audit Metadata