escaping-hardened-containers

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly instructs commands that read sensitive files and environment/metadata endpoints (e.g., /etc/shadow, /proc/*/environ, cloud metadata) which would return secrets that an agent could expose verbatim, creating a high exfiltration risk.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). This document contains explicit, actionable escape techniques and PoCs (e.g., reading /etc/shadow, manipulating cgroups/core_pattern, using the Docker socket, querying cloud metadata, ARP spoofing) intended to enable unauthorized host compromise and data exfiltration.

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 provides explicit, actionable commands to escape containers and modify host system state (mount host devices, write /proc/sys/kernel/core_pattern, create privileged containers via the Docker socket, manipulate cgroups, read /etc/shadow, etc.), which directly instructs compromising the machine.