container-escape-techniques

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly instructs commands that read and write host password hashes and shadow files (e.g., cat /etc/shadow, cat /mnt/host/etc/shadow), which would cause an agent to handle and potentially exfiltrate secret credential data verbatim.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). This is a clearly malicious container-escape playbook that provides step-by-step instructions for host compromise, credential theft, data exfiltration, remote code execution, and persistence/backdoor installation.

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 provides step‑by‑step commands to escape containers and modify the host (mount host filesystems, nsenter into host namespaces, use /var/run/docker.sock to create privileged containers, write cgroup release_agent, exploit runc, create privileged pods), which directly instructs the agent to compromise the machine's state.