trust-attacks

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt explicitly instructs the agent to include and report discovered credentials/tokens in its return summary and evidence files and shows commands that require inserting secret values (passwords, krbtgt hashes, trust keys, NTLM/AES keys) verbatim into CLI/PowerShell arguments, creating a direct secret-handling/exfiltration requirement.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). This document is an explicit offensive playbook for abusing Active Directory trust relationships — it details credential theft (DCSync/secretsdump), forging and installing Kerberos tickets (golden/diamond/trust tickets), extracting krbtgt and trust keys, coercing and capturing TGTs, and modifying AD objects (shadow principals/RBCD) to achieve cross-domain/forest privilege escalation and persistent access.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 0.90). The skill explicitly instructs modifying system files and running privileged commands on the host (e.g., "Add /etc/hosts entries" and "Fix requires root: sudo ntpdate TARGET_DC") and contains many commands that perform state-changing actions (Set-ADObject, Set-ADComputer, DCSync, trust key extraction) which together encourage privileged/administrative modifications to the machine and network.