Daemon that monitors OpenClaw gateway, auto-restarts it on failure, and logs events with configurable intervals on macOS and Linux.

Do not run this script unmodified. Specific recommendations: - Don't trust the SKILL.md description alone: the script only performs a single check-and-restart; add a safe loop and interval and implement restart limits before using as a daemon. - Remove or replace the hard-coded alert recipient ("7533987198"). Confirm where alerts should go and that messaging channels are configured and authorized. - Inspect $HOME/.openclaw/openclaw.json and its backup; the script may overwrite your configuration — back up config elsewhere first. - Run the script interactively first (not with nohup &) to observe behavior and logs, and test restart and messaging commands manually to ensure they do what you expect. - Because source is 'unknown', prefer running in a restricted environment or container until you’re confident. If you want this skill to be accepted, request the maintainer update SKILL.md to match the script, expose configurable CHECK_INTERVAL and MAX_RESTARTS, and remove hard-coded external recipients.

Type: OpenClaw Skill Name: gateway-guardian Version: 1.0.0 The skill bundle 'gateway-guardian' is designed to monitor and auto-restart the OpenClaw gateway. The `guardian.sh` script uses standard shell commands and the `openclaw` CLI to check gateway health, restart it if necessary, and restore configuration from a local backup. It sends status notifications to a hardcoded Telegram ID via the `openclaw message send` command, which is consistent with its watchdog purpose and not indicative of sensitive data exfiltration. There is no evidence of malicious intent, prompt injection against the agent in `SKILL.md`, or significant vulnerabilities.