escaping-containers-to-host
Fail
Audited by Snyk on Aug 23, 2026
Risk Level: CRITICAL
Full Analysis
CRITICAL E005: Suspicious download URL detected in skill instructions.
- Suspicious download URL detected (high risk: 0.80). These links include direct GitHub release binaries and public PoC/pentest repositories (direct binary downloads and exploit code from personal/third‑party repos), which are common distribution vectors for malware or untrusted exploits and should be treated as high risk unless verified.
CRITICAL E006: Malicious code pattern detected in skill scripts.
- Malicious code pattern detected (high risk: 1.00). This repository contains explicit, actionable container escape and host-compromise instructions (cgroup release_agent, Docker socket abuse, hostPath tampering, and runC CVE exploits) including commands to read /etc/shadow and install SSH keys — enabling clear host compromise.
MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).
- Potentially malicious external URL detected (high risk: 0.90). The skill fetches and installs remote executables at runtime (which will be executed) via these URLs: https://github.com/stealthcopter/deepce.git, https://github.com/genuinetools/amicontained/releases/download/v0.4.9/amicontained-linux-amd64, and https://github.com/cdk-team/CDK/releases/latest/download/cdk_linux_amd64, so they are runtime external dependencies that download and run 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 the agent to perform container escape techniques that modify host system files (e.g., write/read /etc/shadow, add SSH keys, drop manifests), run privileged containers via the Docker socket, and otherwise gain host-level code execution, which directly compromises the machine state.
Issues (4)
E005
CRITICALSuspicious download URL detected in skill instructions.
E006
CRITICALMalicious code pattern detected in skill scripts.
W012
MEDIUMUnverifiable external dependency detected (runtime URL that controls agent).
W013
MEDIUMAttempt to modify system services in skill instructions.
Audit Metadata