Using n8n-mcp Skills is a router for agents that build, edit, validate, test or debug n8n workflows through the n8n-mcp server. The detailed rules live in the other skills of the pack, and this one tells the agent which skill to invoke before each action.
It sets three rules (invoke the matching skill first, validate and verify before activating, keep secrets in credentials), a table of when to use native nodes versus Code, and a short summary of the n8n-mcp tools.
What You Get
- Three rules covering skill-first work, validate and verify before activation, and credentials for every secret
- A job-to-node table that prefers IF, Switch, Summarize, Merge, Remove Duplicates and similar native nodes over Code
- A red-flag table that maps a risky thought to the skill to read first
- An index of the fourteen companion skills with when to use each
- A summary of the n8n-mcp tools, node-type naming forms and a step-by-step protocol
Common Use Cases
- Starting any n8n-mcp session so the agent loads the right skill before its first tool call
- Deciding between a native node, an expression and a Code node for a data step
- Catching the habit of activating a workflow straight after a clean validation
- Working with an account that has several n8n instances, where calls must name the target
- Noticing drift between a skill and the live tool schema and reporting it
What Makes It Unique
Other n8n skills teach one area. This one is the map of the whole pack, and it carries strong defaults that apply everywhere: native node first with a written Why Code note for exceptions, no secrets in text fields, and a check of the connections object after every edit. It also tells the agent to trust the live tool over the skill when they disagree.
Under The Hood
The skill is a single instruction file with a README and no scripts. The agent reads it at the start and then loads other skills by name. It describes calling n8n-mcp tools that search node docs, build and update workflows on a connected n8n instance, run tests that execute real nodes, manage credentials and variables and audit security. In the Claude Code plugin install, hooks remind the agent before high-impact tool calls; in other installs nothing nudges it and the skill says so.