OpenClaw Watchdog

v1.3.0

Checks your OpenClaw gateway every 60 seconds. Restarts it automatically if it freezes. Runs in the background, logs events, survives reboots.

0· 22·0 current·0 all-time
by@cromabeltstudios@mirror76
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
VirusTotalVirusTotal
Pending
View report →
OpenClawOpenClaw
Benign
high confidence
Purpose & Capability
The name/description (auto-restart a local OpenClaw gateway) matches the included script: it polls a localhost health endpoint and runs `openclaw gateway restart`. There are no declared env vars, binaries, or config paths that are unrelated to this purpose.
Instruction Scope
SKILL.md and the script scope align: periodic health checks, logging to /tmp, and restarting after 3 failures. Minor inconsistencies in documentation: SKILL.md says logs go to `/tmp/watchdog.log` but the script writes `/tmp/openclaw-watchdog.log`; README references starting the script from a different path and the SKILL.md/SEO mention 'Telegram alerts on restart' although no code implements Telegram. The skill does not read unrelated system files or external credentials.
Install Mechanism
No install spec (instruction-only with an included script). No downloads or external package installs are performed by the skill itself. The code runs locally and will be copied/run by the user/agent — low install risk.
Credentials
No required environment variables or credentials are requested. The script only accesses localhost (127.0.0.1) and writes logs/PID files under /tmp, which is consistent with its stated function.
Persistence & Privilege
The skill is not marked always:true and does not attempt to change other skills or system-wide configuration. It can be started to run as a background daemon and the README suggests users can set up a systemd service (manual step). Note: because the agent (platform) can invoke the skill autonomously, the agent could start the watchdog which will then invoke `openclaw gateway restart` — this is expected behaviour but you should be comfortable with automatic restarts being performed by the skill.
Assessment
This skill appears to do what it says: poll the local OpenClaw health endpoint and run `openclaw gateway restart` when the gateway is unresponsive. Before installing, consider the following: 1) Documentation mismatches — SEO.md mentions Telegram alerts but there is no code for alerts; logs path is inconsistent across files. Treat the bundle as possibly stale/out-of-sync documentation rather than malicious behavior. 2) The script writes logs and a PID to /tmp (world-readable on many systems); if you store sensitive info in logs, move them to a more restrictive location. 3) The restart command runs the external binary `openclaw`; ensure that binary is the expected trusted binary and that the agent has appropriate permissions to restart the gateway. 4) The restart call uses check=True and exceptions from a failing restart may crash the watchdog — review/error-handle if you need high resilience. 5) Because the agent can autonomously start the watchdog, confirm you want automatic restarts (and not manual intervention) in case of failures. If you’re unsure, inspect/run the script in a safe environment first (read the code, run in a non-production container, or request the author to clarify the Telegram claim and the log path inconsistencies).

Like a lobster shell, security has layers — review code before you run it.

auto-restartvk978qw3v25at15dwkkpgmzhc1h84g5kcautomationvk9792f1rhrbmqc2aeyyfbf1jm584e9bvlatestvk978qw3v25at15dwkkpgmzhc1h84g5kcmonitoringvk978qw3v25at15dwkkpgmzhc1h84g5kcopenclawvk978qw3v25at15dwkkpgmzhc1h84g5kcself-healingvk970yyp5fbpfwaf9vpkwg2zqdh84hzhxwatchdogvk978qw3v25at15dwkkpgmzhc1h84g5kc

License

MIT-0
Free to use, modify, and redistribute. No attribution required.

Comments