OpenClaw Lobster reports lobster timed out, usually a single workflow is too long, external commands are too slow, or the output is too large. Don't increase the timeout time indefinitely, you should first determine which step it is stuck in, and then decide whether to disassemble the process, increase timeoutMs, or reduce the output.
Press Error Handling
| Wrong | Treatment |
|---|---|
| lobster timed out | Increase timeoutMs, or break down a long process into several short workflows |
| output exceeded maxStdoutBytes | Reduce command output, or increase maxStdoutBytes |
| returned invalid JSON | Make sure that the pipeline's tool mode only outputs JSON and does not mix in log text |
| lobster failed | Look for embedded runner errors in openclaw logs --follow |
More stable writing
- Each step has the small CLI output JSON.
- Steps with side effects must be approved.
- "Collection, classification, and execution" are divided into recoverable stages.
- Retry time-consuming external APIs individually, don't blind the entire Lobster pipeline, etc.
If you run the same Lobster process every day, it is recommended to write parameters, timeouts, and output limits into the workflow file, and do not temporarily let the model spell out long commands every time. The stability will be significantly better.
Official open source address: https://github.com/openclaw/openclaw.
If a workflow times out frequently, it may be taking on too many responsibilities. After separating the "pull data", "model judgment" and "execution action", the failure point is clearer, and it is easier to rerun only the failed part.
