Nexus Safe

Provides local system health monitoring and controlled service restarts for Docker and PM2 with full privacy and zero external calls.

This skill appears to do what it says (local monitoring + policy-controlled restarts), but there are packaging and documentation gaps you should address before enabling it to perform restarts: - Verify required binaries and env vars: _meta.json and monitor.py require docker, pm2, and NEXUS_SAFE_* environment variables. The registry metadata you were shown contradicted this — confirm which is authoritative and set ALLOWED lists explicitly. - Keep restarts disabled by default: leave NEXUS_SAFE_ALLOW_RESTARTS unset/false until you’ve tested behavior in dry-run mode and reviewed audit.log entries in ~/.nexus-safe/audit.log. - Test with dry-run: use the --dry-run option to confirm which commands the skill would execute and that the allowlist protects only intended services. - Inspect and control inputs: ensure any values passed as service names come from trusted sources (the code uses subprocess.run with argument lists, which mitigates shell injection, but validating service names is still good practice). - Install deps manually from a trusted environment: pip install psutil from an environment you control; there is no installer bundled with the skill. If you want higher assurance, ask the publisher to correct the registry metadata to list required binaries/env vars in the public manifest and to document explicit setup steps for the allowlists and enabling restarts.

Type: OpenClaw Skill Name: nexus-safe Version: 0.1.6 The OpenClaw skill 'nexus-safe' is designed for local system reliability, offering status checks, log retrieval, and policy-controlled service restarts. The `monitor.py` script uses `subprocess.run` with a list of arguments, which is safe against shell injection. Service names are validated against live system processes and explicit allowlists (`NEXUS_SAFE_ALLOWED_DOCKER`/`PM2`), significantly mitigating command injection risks. The skill enforces policies like logs-first and rate limiting. The `SKILL.md` and `AGENT_BRAIN.md` files provide clear, non-malicious instructions for the agent, focusing on safe and reasoned system management. There is no evidence of data exfiltration, malicious execution, persistence mechanisms, or prompt injection attempts against the agent.