Self-healing stability monitor for AI agents - 5 core checks + 15 extended checks, auto-recovery, health scoring

This skill largely does what it says (system monitoring + controlled recovery) but has a few red flags you should address before installing: - Verify dependencies and metadata: the registry metadata lists no required binaries, but the code expects pm2, openclaw-cli and system tools (systemctl, nginx, docker). Ensure those are present and the metadata is updated. - Audit the full source (complete aegis-protocol.py) for any network calls (ssl/socket usage) or commands not in the declared whitelist. The SKILL.md and SECURITY_AUDIT.md claim only hardcoded commands are used — confirm that in the code you run. - Run in a sandbox/non-production host first. Test 'check' and 'status' before enabling 'heal' or scheduled runs. - Restrict privileges: do not run with root unless necessary. If you need automatic recovery, consider granting only the minimal sudo rules for the specific service-restart and session-kill commands rather than unrestricted root. - Disable or review autonomous recovery by default: if you want monitoring only, disable automated 'heal' actions or set conservative thresholds and cooldowns in the config file (~/.openclaw/workspace/.watchdog-config.json). - Monitor logs closely: tail /var/log/aegis-protocol.log and inspect the healing-memory and cache files for unexpected actions. - Pin to a reviewed commit or vendor-signed release and avoid running code from unknown/untrusted sources on critical systems. If you want, I can list the exact commands the code calls (from the file) and produce a minimized whitelist you can copy into a sudoers snippet or an execution-guard wrapper.

Type: OpenClaw Skill Name: aegis-protocol Version: 0.12.9 Aegis Protocol is a system monitoring and self-healing tool that requires high-risk permissions, including 'exec', 'write', and 'process' management. The core script `aegis-protocol.py` uses `subprocess.run` with `shell=True` to execute hardcoded system commands (e.g., `pm2`, `systemctl`, `docker`), which introduces shell injection vulnerabilities if command outputs are ever compromised. While the bundle includes extensive documentation and a 'Security Audit' claiming to be a false positive, the combination of broad system control, the ability to kill sessions, and the pre-emptive defensive documentation designed to bypass security scanners warrants a suspicious classification.