openclaw-rescue-gateway-skill_skill

Overview

This skill configures a stable, maintainable Rescue Gateway for OpenClaw that runs independently from the primary gateway. It sets a separate profile, ports, workspace, and launchd label so a secondary Discord Rescue Bot can operate without being stopped by main gateway commands. The configuration ships default full exec permissions and no approval prompts for emergency operations.

How this skill works

The skill creates a dedicated profile directory (~/.openclaw-rescue/openclaw.json) and a distinct launchd label (ai.openclaw.gateway.rescue). It adjusts gateway port, agent workspace, agentDir, exec/tool permissions, and ACPX plugin settings so the rescue instance is isolated. It also documents copying auth-profiles for model API access and provides launchctl-based start/stop/validation commands.

When to use it

• You have a primary OpenClaw gateway installed and need a separate rescue entrypoint.
• You must run a second Discord bot (e.g., OpenClaw Rescue Bot) with independent credentials.
• You require an independent port and workspace (recommended port 19001).
• You need rescue gateway to keep full exec permissions and bypass approval for emergencies.
• You want to ensure main gateway stop commands cannot accidentally stop the rescue instance.

Best practices

• Use the official profile path ~/.openclaw-rescue/openclaw.json for all rescue configs to keep CLI compatibility.
• Assign a dedicated launchd label ai.openclaw.gateway.rescue to fully isolate service lifecycle from the main gateway.
• Choose a port at least 20 ports away from main gateway (default rescue 19001) to avoid derived port conflicts.
• Set tools.exec.security=full and tools.exec.ask=off plus agents.defaults.elevatedDefault=full for emergency readiness.
• Copy main agent auth-profiles.json into the rescue agentDir and set strict file permissions (chmod 600).

Example use cases

• Run a standby Discord Rescue Bot that automatically takes commands when the main gateway is down.
• Maintain an always-available emergency agent with full exec access for rapid incident remediation.
• Operate a rescue environment on a separate macOS user launchd service so openclaw gateway stop won’t affect it.
• Test failover scenarios by bootstrapping and validating the rescue gateway without touching main gateway state.