redteam

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). 不安全：该提示多处要求将明文密码、NTLM 哈希和凭据内联到命令/调用（例如 impacket/nxc 示例中 -p 'Password'、-hashes :NTHASH 等），意味着模型必须在输出中包含秘密值的逐字文本，存在高风险的数据外泄。

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). This document is a comprehensive offensive toolkit and playbook that explicitly instructs credential theft (Mimikatz, secretsdump, Inveigh), account abuse (Kerberoast, AS-REP), lateral movement and remote code execution (wmiexec/psexec/smbexec/ntlmrelayx), proxy/command-and-control setup to attacker-controlled hosts (chisel/Nc/PowerShell reverse shells), and data exfil/upload-and-exec flows — clearly designed for unauthorized compromise and persistent access.

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

• Attempt to modify system services in skill instructions detected (high risk: 1.00). The prompt instructs the agent to run privileged, state-changing actions (upload/execute binaries, create/start services, modify DNS/AD records, bypass execution policy, run credential-extraction tools and long-lived listeners) which directly modify the machine(s) it runs on or remote hosts and require elevated privileges, so it clearly pushes the agent to compromise system state.