hardening-linux-endpoint-with-cis-benchmark
Warn
Audited by Gen Agent Trust Hub on Jun 20, 2026
Risk Level: MEDIUMDATA_EXFILTRATIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill contains deceptive instructions. While the metadata and title suggest a 'Hardening' (defensive) purpose, the 'Workflow' section explicitly outlines offensive phases including 'Exploitation' and 'Post-exploitation' (lateral movement and data exposure). Additionally, the 'When NOT to Use' section paradoxically states the skill should not be used for defensive tasks.
- [DATA_EXFILTRATION]: Step 4 ('Audit and Logging') configures the system to send all logs ('.') to an external domain 'syslog-server.corp.com' via rsyslog. This facilitates the continuous exfiltration of sensitive system events and metadata to a remote server not controlled by the user.
- [COMMAND_EXECUTION]: The skill provides numerous commands that require root privileges to modify critical system configuration files, including '/etc/fstab', '/etc/ssh/sshd_config', '/etc/shadow', and PAM configurations. While these are common in hardening, their inclusion alongside offensive workflows increases the risk of misuse.
- [EXTERNAL_DOWNLOADS]: The skill instructions include installing multiple packages ('chrony', 'auditd', 'openscap-scanner') via 'apt install'. While these are legitimate packages, the context of the other findings suggests a need for caution.
Audit Metadata