goad

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt contains explicit passwords and an SSH key path and shows them embedded verbatim in commands and code (e.g., WinRM, Impacket, smbclient, ldapsearch, curl), forcing the agent to handle and output secret values directly.

---

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). This document exposes explicit plaintext credentials and an SSH key path, public jumpbox IP, and step-by-step offensive techniques (WinRM tunneling, Kerberoasting/AS-REP/Kerberoast, pass‑the‑hash, enabling PowerShell remoting via SSM) that directly enable unauthorized access, lateral movement, credential theft, and remote code execution.

---

HIGH W008: Secret detected in skill content (API keys, tokens, passwords).

• Secret detected (high risk: 1.00). I flagged this document because it contains multiple literal, high-entropy passwords that are directly usable credentials for accounts on lab hosts. Per the secret definition, these are not placeholders or simple setup passwords—they are complex-looking strings used for the goadmin account and appear verbatim in credential tables and example commands (WinRM, Kerberos, LDAP, smbclient).

Specifically found and considered secrets:

• 8dCT-DJjgScp (goadmin on DC01) — appears in credentials table and multiple command examples
• NgtI75cKV+Pu (goadmin on DC02 and SRV02) — appears in credentials table
• Ufe-bVXSx9rk (goadmin on DC03) — appears in credentials table
• 978i2pF43UJ- (goadmin on SRV03) — appears in credentials table

Notes on things ignored/not flagged:

• The SSH key path (cyberagent/.../ubuntu-jumpbox.pem) is just a filesystem path to a .pem file, not the private key material itself, so it is not flagged.
• Hostnames, IP addresses, and the username "goadmin" are not secrets by themselves.

Because the document contains multiple directly embedded, high-entropy passwords that would grant access to services, it meets the definition of containing real secrets.

---

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 explicitly instructs obtaining root via "sudo -i" and running system-level provisioning (ansible-playbook) and AWS SSM commands that change system configuration (e.g., enabling PSRemoting), which direct the agent to modify the machine's state and escalate privileges.