The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes extensive system diagnostic commands to assess the security posture of the host environment. This includes gathering operating system details (uname, sw_vers), listing network listening ports (ss, lsof), and checking the status of various firewall implementations (ufw, firewall-cmd, nft, socketfilterfw).
[COMMAND_EXECUTION]: The skill generates and offers to execute a remediation plan that can modify critical system settings. This includes changing firewall rules, closing ports, and modifying SSH or RDP configurations. While these are high-privilege operations, the skill implements safety measures such as requiring explicit user approval for each state-changing action, recommending system backups, and providing rollback strategies.
[COMMAND_EXECUTION]: The skill utilizes the 'openclaw' command-line tool for security audits, status checks, and scheduling periodic tasks.
[PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection because it ingests and interprets the output of system commands to determine its next steps and remediation plan. Ingestion points: System command output and tool results (e.g., firewall status, network ports). Boundary markers: None explicitly defined to separate command output from instructions. Capability inventory: Extensive command execution (shell access), system configuration modification, and cron management. Sanitization: The instructions explicitly require the agent to redact secrets and sensitive host details from logs and persistent memory files.

healthcheck