n8n Skills

In Claude Code: /plugin marketplace add czlonkowski/n8n-skills

What it does

A bundle of seven complementary skills that teach a coding agent how to build flawless n8n workflows through the n8n-mcp MCP server. Instead of letting the model guess at node configs and loop on validation errors, the skills encode the real rules: expression syntax, MCP tool selection, proven workflow patterns, node configuration, validation handling, and Code-node JavaScript and Python.

The seven skills:

1. n8n Expression Syntax, correct {{ }} patterns and the webhook $json.body gotcha.
2. n8n MCP Tools Expert, which n8n-mcp tool to use and the exact parameter formats.
3. n8n Workflow Patterns, proven architectures (webhook, HTTP API, database, AI, scheduled).
4. n8n Node Configuration, operation-aware config and property dependencies.
5. n8n Validation Expert, reading and fixing validation errors without looping.
6. n8n Code (JavaScript), writing Code-node JS with the right $input / $json access.
7. n8n Code (Python), the same for Python, with standard-library limits called out.

VeryCreatives' take

If you build n8n workflows with an agent, this is the difference between a workflow that runs and one that loops on validation errors for twenty minutes. The MCP Tools Expert skill alone earns its place by stopping the agent from misusing n8n-mcp. It is the most popular skill in its niche for good reason. Top pick for any team automating with n8n.

How VeryCreatives uses it

We run n8n on our own server and build automations on it every week, so this is a skill we use in production, not one we tried once.

Before we found it, we built workflows by hand, node by node. Now we build them the agentic way: we talk through what the workflow needs to do, and Claude Code builds it through the n8n-mcp server. The time and hassle that used to go into wiring nodes and debugging a half-working flow mostly disappeared. Building and maintaining workflows is fun again, which is not something we expected to say about internal ops tooling.

A few of the workflows we run with it:

• Ops data bridging: gathering and moving data between Asana and our finance and controlling systems, so the numbers line up without manual exports.
• Content repurposing at scale: turning blog posts into social media pieces.
• Client product-analytics reports: pulling data from PostHog, Google Search Console, and GA4, then posting the report straight into each client's Slack channel.

Building these by agent is far easier than building them by hand, and it scales with how clearly we can describe the outcome.

Two practical notes from running it:

• AI nodes: we route the AI agent nodes through OpenRouter instead of wiring each provider directly. It gives us the flexibility to switch models without signing up for every LLM separately.
• Hosting: we self-host to keep costs down, but the skill works the same on n8n Cloud.

One caveat we hit: when you update an existing workflow with this skill, a node that had a credential set can get switched back to the default, so you re-select the right credential in the node editor afterward. It is a small, affordable cost for what the skill saves you.

When to use it

• You build or maintain n8n workflows and use an agent to do it.
• You have the n8n-mcp MCP server available (these skills assume it).
• You keep hitting validation errors or node-config mistakes the agent cannot self-correct.

When not to

• You do not use n8n. Nothing here transfers to other automation tools.
• You cannot run the n8n-mcp MCP server in your environment.

Cross-agent note

Distributed as a Claude Code plugin, so the one-line install is Claude Code only. The underlying skills are plain SKILL.md folders and port to Codex and OpenCode by copying them into the agent's skills directory and configuring the n8n-mcp MCP server. We have not yet hand-verified those ports, so the compatibility flags are AI-inferred.