linux-file-path-abuse

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The prompt explicitly instructs generating and embedding plaintext secrets (passwords, password hashes, SSH keys, and attacker IP/port) into commands and files (e.g., echoing backdoor:, openssl passwd "password123", writing SSH keys), which would require the LLM to output secret values verbatim and thus creates an exfiltration risk.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). This document is a high-risk, intentional privilege-escalation guide providing explicit step-by-step instructions to create persistent backdoors, add/modify root accounts and SSH keys, install SUID binaries, hijack shared libraries and PATH/profile scripts, extract credential files, and establish remote shells—clear malicious intent and abuse patterns.

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 the agent to perform privileged, state-changing attacks (adding UID 0 users, modifying /etc/passwd, /etc/shadow, /etc/sudoers, writing root SSH keys, creating SUID binaries, mounting the host via docker/lxc, etc.), which directly compromise the machine the agent runs on.