av-edr-evasion

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt includes plaintext credentials and secret-like values (e.g., "Password123!", ATTACKER_IP/ATTACKBOX addresses and ports) and instructs embedding them verbatim into generated payloads/commands, which requires the LLM to handle/output secrets directly.

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 1.00). The microsoft schema URL is benign, but the ATTACKBOX HTTP links (especially nc.exe, tool.ps1, payload.exe and the generic /file endpoint) point to an attacker-controlled server serving direct executables/scripts over HTTP and are high-risk indicators for malware distribution.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). The document is high-risk: it explicitly instructs on AV/EDR evasion and includes concrete malicious capabilities—custom DLL/EXE payloads, reverse shells, account creation, staged downloaders, AMSI and ETW patching, and LOLBin/techniques for covert execution—clearly facilitating backdoors, remote code execution, credential abuse, and covert data/execution exfiltration paths that can be readily misused.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 1.00). The skill contains explicit runtime download-and-execute commands (e.g., wget/certutil/URLDownloadToFileA/IEX) that fetch and run payloads from http://ATTACKBOX:8080/ (examples: http://ATTACKBOX:8080/nc.exe, http://ATTACKBOX:8080/tool.ps1), so it relies on external content at runtime that directly delivers and executes remote code.

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

• Attempt to modify system services in skill instructions detected (high risk: 1.00). This skill explicitly instructs generating and deploying payloads that change host state—adding admin users, creating/serving/executing custom DLLs/EXEs, opening reverse shells, and patching AMSI/ETW to bypass security—so it pushes the agent to compromise the machine.