Auto-Watchdog

WarnAudited by ClawScan on May 10, 2026.

Overview

This watchdog is purpose-aligned, but it asks for hidden persistent automation that can restart processes and clean files with little scoping or approval.

Install only if you intentionally want a background watchdog that can restart services and modify files. Before using it, define exact process/path allowlists, review any guardian scripts, avoid hidden/bypass execution where possible, run with least privilege, enable audit logs, and keep a simple disable/uninstall procedure.

Findings (6)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

What this means

A bad threshold or stale log could cause important services to be killed, restarted repeatedly, or files to be removed without the user noticing first.

Why it was flagged

The skill instructs automatic process termination/restart and file cleanup based on broad conditions, without clear target allowlists, approval gates, dry-run mode, or rollback.

Skill content
Log file not updated in X minutes → kill + restart; Log files growing too large → auto-rotate; Temp file cleanup
Recommendation

Require explicit allowlists for processes and paths, user-approved thresholds, dry-run logging, and confirmation before destructive or service-impacting actions.

What this means

Unreviewed or modified script code could run invisibly in the background with the permissions of the scheduled task.

Why it was flagged

The setup explicitly recommends hidden, non-interactive PowerShell execution with ExecutionPolicy bypass for an unspecified guardian script.

Skill content
guardian-silent.vbs — zero flash process monitor ... powershell.exe -NonInteractive -WindowStyle Hidden -ExecutionPolicy Bypass -File C:\path\to\guardian.ps1
Recommendation

Avoid hidden execution and ExecutionPolicy bypass unless strictly necessary; use signed, reviewed scripts, visible logs, and least-privilege service accounts.

What this means

The watchdog may keep acting on the environment after the user’s immediate task is over, making repeated restarts or cleanup actions hard to notice and stop.

Why it was flagged

The skill is designed for frequent, persistent, silent autonomous operation that continues after initial setup.

Skill content
Register as Task Scheduler job running every 1-5 minutes ... Restart=always ... Auto-Watchdog monitors everything and fixes what it can — silently.
Recommendation

Provide explicit enable/disable/uninstall instructions, visible audit logs, scoped runtime permissions, and a safe mode that alerts before taking action.

What this means

A user may underestimate the risk of allowing automatic process control and cleanup to run silently in the background.

Why it was flagged

The wording encourages users to trust unattended remediation while downplaying the visibility and control needed for high-impact system changes.

Skill content
Set it and forget it ... Fix first, alert second ... Silent = healthy.
Recommendation

Reframe the documentation to clearly disclose risks, required privileges, failure modes, and when user approval is required.

What this means

If installed with excessive privileges, the watchdog could affect more services or files than intended.

Why it was flagged

Creating a systemd service and restarting gateway/processes can require elevated local privileges; this is expected for a watchdog but should be explicitly bounded.

Skill content
/etc/systemd/system/openclaw-watchdog.service ... openclaw gateway status ... Auto-restart if down
Recommendation

Run it under the least-privileged account possible and document exactly which services, paths, and commands it is allowed to manage.

What this means

The safety of the actual watchdog depends on external or user-created scripts not present in this review.

Why it was flagged

The skill references helper scripts that are not included in the provided one-file artifact set, so their behavior cannot be reviewed here.

Skill content
ExecutionPolicy Bypass -File C:\path\to\guardian.ps1 ... ExecStart=/usr/bin/node /path/to/guardian.js
Recommendation

Only use reviewed local scripts from a trusted source, pin their paths and versions, and audit them before scheduling persistent execution.