The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads Kubernetes job specifications from Aqua Security's official GitHub repository (raw.githubusercontent.com/aquasecurity/kube-bench/...) to perform automated CIS benchmark scanning. As this is a well-known security vendor and directly related to the skill's purpose, it is considered a legitimate resource.
[COMMAND_EXECUTION]: The scripts/harden-linux.py and scripts/scan-infrastructure.sh files utilize Python's subprocess module and shell eval to execute various system hardening commands, such as sysctl, systemctl, and sshd -t. These operations are transparently documented and essential for the skill's primary function of system configuration.
[PRIVILEGE_ESCALATION]: The hardening script (harden-linux.py) explicitly requires root privileges (checked via os.geteuid()) to modify critical system files in /etc/ssh/ and /etc/sysctl.d/. This requirement is typical for infrastructure configuration management and is clearly communicated in the documentation.
[INDIRECT_PROMPT_INJECTION]: The skill includes functionality to generate compliance reports by processing external JSON scan results from tools like Trivy and kube-bench. This represents an ingestion surface where untrusted data could influence reporting output, though the current implementation is restricted to local processing for dashboard generation.
Ingestion points: scripts/generate-compliance-report.py (referenced in SKILL.md) reads cis-report.json and vuln-report.json.
Boundary markers: Absent in report processing logic.
Capability inventory: File write (compliance-dashboard.html) and local command execution via related scripts.
Sanitization: Standard JSON parsing is used; specific sanitization of report content before HTML interpolation is not explicitly detailed.

security-hardening