The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various system utilities including uname, ss, lsof, ufw, and firewall-cmd to assess host security. It also utilizes the openclaw CLI for audits and can configure persistence through openclaw cron for periodic security checks. All state-changing actions require explicit user confirmation.
[DATA_EXFILTRATION]: The skill performs detailed reconnaissance, gathering information on listening ports, network exposure, disk encryption status, and OS versions. This information is collected to inform the security audit, and the skill includes specific instructions to redact tokens and credentials from generated logs.
[EXTERNAL_DOWNLOADS]: Queries the npm registry using npm view to check for the latest versions and update status of the OpenClaw software.
[PROMPT_INJECTION]: The skill instructs the agent to self-assess its model version and recommend switching to more advanced models if necessary. Additionally, it presents an indirect prompt injection surface by processing untrusted output from system commands and audit tools. This is mitigated by instructions to use boundary markers (when memory writes are used) and the requirement for human-in-the-loop confirmation before any action is taken based on that data.

healthcheck