The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various system reconnaissance commands to assess host security posture. This includes identifying the OS via uname -a and /etc/os-release, checking listening ports using ss or lsof, and verifying firewall status with ufw, firewall-cmd, or socketfilterfw. These actions are necessary for the skill's primary purpose of security auditing.
[PERSISTENCE]: The skill provides functionality to schedule periodic security audits using the openclaw cron add command. While this creates a persistent task on the host, the instructions mandate that the agent must obtain explicit user approval before creating any scheduled tasks and must use deterministic names (e.g., healthcheck:security-audit).
[DATA_EXPOSURE_&_EXFILTRATION]: The skill reads sensitive system configuration details, such as backup status, disk encryption status, and firewall rules. However, it includes explicit instructions to redact secrets, avoid logging tokens or credentials, and perform these checks in a read-only manner unless the user approves a remediation plan.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from local shell commands and security audit outputs (e.g., openclaw security audit --json). This creates a minor attack surface where malicious system artifacts (like a crafted filename) could influence the agent, but the risk is mitigated by the skill's requirement for human review of all remediation plans before execution.

healthcheck