Openclaw Gateway Manager

Manage OpenClaw gateways: query status, modify ports, restart, verify configs, create, delete instances with backups and port scanning.

This package appears coherent for a gateway manager, but it performs destructive and persistent operations — review before use. Specific recommendations: - Inspect scripts (especially gateway-delete.sh and create_service_file in common.sh) before running; deletion uses rm -rf even though backups are taken. - Verify the service file content (LaunchAgent plist / systemd unit). Note the LaunchAgent launches node -e with a JS snippet that spawns openclaw; confirm this behavior and that it doesn't embed unexpected network endpoints. - Ensure openclaw and node binaries invoked are the expected, trusted binaries on your system (check PATH and /usr/local/bin, etc.). - Understand backups are stored under $HOME/.openclaw-deleted-backups/ and verify retention and restore procedure before deleting production data. - Run first in a test account/environment and use the --dry-run / inspect pattern advised by the repo (or add your own dry-run) to confirm actions. - Be aware user-level services will run persistently under your account and will inherit user environment variables; avoid running with sensitive credentials in your environment if you are concerned about process-level exposure. If you cannot inspect the code yourself, do not grant it automatic execution in sensitive environments; run it manually in a controlled test environment first.

Type: OpenClaw Skill Name: openclaw-gateway-manager Version: 1.0.3 The bundle provides a management utility for OpenClaw gateways with capabilities for persistence and destructive file operations. Key scripts like `gateway-create.sh` and `gateway-delete.sh` create user-level services (macOS LaunchAgents and Linux systemd) and use `rm -rf` on configuration directories. Although mitigations like triple-confirmation and backups are implemented, the scripts lack input sanitization for instance names used in path construction (e.g., in `common.sh`), which could allow path traversal. These high-risk functions, combined with the instructions in `SKILL.md` for an AI agent to execute them, pose a potential risk if the agent is subjected to prompt injection.