n8n Validation Expert covers what to do when validate_node or validate_workflow returns errors, warnings or suggestions in the n8n-mcp server. It explains the five core error types, the four validation profiles and how to read the result structure.
It also covers recovery strategies, the auto-fix tool, known false positives and blind spots, and how to run a workflow after it validates. Three companion guides hold an error catalog, a false-positive guide and a review checklist for existing workflows.
What You Get
- A severity guide that separates blocking errors from warnings and suggestions
- A comparison of the minimal, runtime, ai-friendly and strict profiles, with runtime as the everyday default
- An error catalog with broken configs and their fixes, plus a list of known false positives and blind spots
- Recovery strategies: start fresh, binary search, clean stale connections and auto-fix with confidence levels
- A procedure for running a validated workflow with real triggers or pinned test data, and a checklist for reviewing an existing workflow
Common Use Cases
- Fixing a missing_required or invalid_expression error returned by validate_node
- Deciding whether a best-practice warning such as missing error handling is worth fixing
- Previewing and applying n8n_autofix_workflow fixes at a chosen confidence level
- Running a workflow with pinned data to confirm it behaves after it validates
- Auditing a workflow that already passes validation for silent connection bugs
What Makes It Unique
Most n8n guidance covers how to build a node. This skill covers the feedback side: it records which warnings the validator still raises in error, which mistakes it cannot see at all (such as a JavaScript error inside an expression that resolves to null), and when a warning is acceptable. It states that a passing validation is necessary but not sufficient, and pairs it with a test run.
Under The Hood
The skill is instructions only and ships no scripts. It loads SKILL.md into context and points to three companion markdown guides that the agent reads when needed. Following it, the agent calls n8n-mcp tools: validating nodes and workflows, which is read-only, then auto-fix and partial-update operations that edit and can activate a workflow on the connected n8n instance. The test step runs real nodes, so external calls and writes happen, and the skill tells the agent to ask the user before turning on the workflow’s MCP exposure setting.