OpenClaw is a multi-channel AI Agent gateway that supports messaging platforms such as WhatsApp, Telegram, and Discord. This skill provides the command reference and troubleshooting capabilities required for operations.
Design Pattern
This skill mainly adopts:
- Tool Wrapper: Provides OpenClaw commands, diagnostic paths, and reference materials
- Runbook / Pipeline: Executes in the order of "status check → problem triage → select repair action → verify result"
- Reviewer (Lightweight): First determine the problem type, then decide whether repair is needed
Gotchas
- Do not restart or auto-repair right away; first look at the status and problem type
- For high-impact operations such as restart, update, repair, and configuration modification, explain to the user before executing
- Do not fabricate OpenClaw CLI subcommands; when uncertain, check the documentation or local reference first
- Do not lump channel problems, model problems, and gateway problems into one catch-all command
- After repairing, be sure to return to the health check; do not stop at "the repair command has been executed"
Standard Diagnostic Process
When the user reports an OpenClaw problem, execute the following steps:
0. First determine whether confirmation is needed
The following operations can be done directly by default:
openclaw statusopenclaw channels status --probeopenclaw logs --limit 200openclaw health
The following operations are high-impact actions and should be explained to the user before execution:
openclaw doctor --repairopenclaw doctor --repair --forceopenclaw gateway restartopenclaw update- Any command that will explicitly modify configuration, reconnect channels, or reset state
# 1. 快速状态检查
openclaw status
# 2. 深度渠道探测
openclaw channels status --probe
# 3. 自动修复
openclaw doctor --repair
# 4. 验证健康状态
openclaw health
Problem Classification Handling
Select the corresponding handling path according to the problem type:
Service cannot start → openclaw gateway status → openclaw doctor --repair → openclaw gateway restart
Channel connection failure → openclaw channels status --probe → openclaw channels login --channel <name> → View channel logs
Message sending failure → openclaw logs --limit 200 → Check channel status → Verify target format
Performance issues → openclaw status --usage → View system logs → Check resource usage
Configuration query → Consult openclaw_commands.md corresponding chapter
Repair Strategy
- Preferred:
openclaw doctor --repairAuto repair - Gateway not running:
openclaw gateway restart - Still unhealthy after repair:
openclaw doctor --repair --force - Repair after update:
openclaw update→openclaw doctor→openclaw health
Success Criteria
openclaw healthReturns no errorsopenclaw channels status --probeAll channel statuses are normalopenclaw statusShows gateway reachable
Most Commonly Used Commands Quick Reference
| Command | Purpose |
|---|---|
openclaw status |
View overall status |
openclaw health |
Health check |
openclaw doctor --repair |
Auto repair |
openclaw gateway restart |
Restart service |
openclaw channels status --probe |
Check channels |
openclaw logs --limit 200 |
View logs |
macOS Notes
- Gateway is managed by the OpenClaw Mac app; do not start it manually with tmux
- Check launchd environment variable overrides:
launchctl getenv OPENCLAW_GATEWAY_TOKEN - System logs:
/tmp/openclaw/openclaw-YYYY-MM-DD.log - Service logs:
~/.openclaw/logs/gateway.log - macOS error logs:
./scripts/clawlog.sh -e
Detailed Reference
- Complete command manual:openclaw_commands.md
- Troubleshooting and repair process:openclaw_fix.md
Reporting Standards
When reporting to the user, it should include: the problems found, the actions taken, the current status, and follow-up recommendations.