n8n Error Handling is a skill for Claude that makes workflow failures loud, structured and recoverable. By default one failing node halts an n8n workflow and the caller of a webhook sees a timeout or an empty 500, so unattended and API workflows need explicit handling.
The skill covers the two-step setup for per-node error outputs, node retries, the status codes and response bodies a webhook should return, input validation, and a separate Error Trigger workflow that alerts someone when something escapes. It also lists the failures an error output never sees.
What You Get
- A table of which workflow shapes need error handling and which can skip it
- The two-step recipe: set onError to continueErrorOutput, then wire main[1], with patch operations for both
- A retryOnFail recipe with its engine limits
- A cause-to-status table for 400, 401, 403, 404, 409, 429, 500, 502, 503 and 504, with a Switch layout
- An Error Trigger workflow outline with the recursion trap and the handled-error trap
- Reference files for node error outputs, API workflows, response shapes and error workflows
Common Use Cases
- Hardening a webhook API so each failure returns a clear status and body
- Adding retries and an alert workflow to a scheduled sync
- Finding why a run shows success although a node’s error was discarded
- Validating input upstream and returning field-level 400 responses
- Reviewing a workflow where the error branch returns 200
What Makes It Unique
It names the half-wired error output as the top silent trap, with a table of what happens in each half-done state, and it says validate_workflow will not catch it, so Claude pulls the workflow to confirm both parts. It lists failures that bypass error outputs, such as errors inside expressions, and it states what the MCP cannot set, so Claude hands the user the UI step for the error workflow.
Under The Hood
The skill is instructions and four reference markdown files, with no scripts. It tells Claude to patch node settings and connections, validate, autofix, run test executions and read executions through the n8n-mcp tools, all of which call your n8n server. The workflows it builds can post alerts to Slack or email when they run, and the skill reminds the user to assign the error workflow in the n8n settings screen.