Server Mate

Build or extend a lightweight server monitoring and AI operations workflow for Linux hosts running Nginx or Apache. Use when Codex needs to collect psutil me...

This package is internally consistent with a server-monitoring tool, but review and control these items before deployment: - Automation: Keep automation.dry_run: true and auto_ban/auto_heal disabled until you've validated alerts, allowlists, cooldowns, audit logs, and rollback plans. The command templates (iptables, systemctl) will be executed if you enable automation. - Secrets & webhooks: Webhook URLs, Telegram tokens, and any OpenAI key are operator-supplied secrets. Store them securely and do not commit them to Git. The agent will send data to whatever webhook/API endpoints you configure. - GeoIP bootstrap: If no local MaxMind .mmdb is available the report generator will attempt a public mirror download (GEOIP_MIRROR_URL points at a GitHub-hosted mirror). Prefer provisioning GeoIP via your own MaxMind account and geoipupdate; treat public-mirror downloads as an operator-reviewed bootstrap only. - Log paths & scope: The agent can auto-detect auth logs (e.g., /var/log/auth.log or /var/log/secure) if configured that way; verify config.yaml paths to avoid unintentionally reading system logs you don't want processed. Running with defaults on a production host may require root privileges for some operations and to access protected log files. - Network egress: Enabling AI analysis or webhook channels will cause egress to third-party services (OpenAI, Telegram, DingTalk, Feishu, etc.). Audit the content you allow to be sent (raw or excerpted logs) and sanitize sensitive fields if needed. If you want increased assurance before installing: inspect the command_template strings in your config, run the agent in a sandbox with synthetic logs, and prefer local-only config paths (./data, ./logs, ./reports) until you're ready to connect real endpoints.

Type: OpenClaw Skill Name: server-mate Version: 1.3.3 The skill bundle implements a server monitoring and automation system with high-privilege capabilities, specifically 'auto-ban' and 'auto-heal' features in 'server_agent.py' that execute system commands (e.g., iptables, systemctl) via subprocess. While the code includes safety measures like 'dry_run' modes, whitelists, and the use of 'shlex.split' to mitigate shell injection, the pattern of executing commands based on parsed log data (which can be influenced by external attackers) poses a inherent security risk. Additionally, 'report_generator.py' includes logic to download a GeoIP database from a third-party GitHub repository (P3TERX/GeoLite.mmdb) if not found locally, which introduces a supply chain risk. No clear evidence of intentional malice or data exfiltration was found, but the high-risk capabilities warrant a suspicious classification.