
n8n Official Skills
Official n8n skill collection for building AI agents and workflows.
About
n8n Official Skills is the official skill repository from n8n, Apache-2.0 licensed, covering the n8n automation platform end to end. It includes 14 skills such as n8n-agents-official for AI Agent nodes, n8n-expressions-official, n8n-error-handling-official, n8n-debugging-official, n8n-subworkflows-official, n8n-credentials-and-security-official, n8n-data-tables-official, and guides for extending n8n with MCP and managing the workflow lifecycle.
The repo ships with Claude Code, Codex, and OpenClaw plugin formats plus hooks, so skills drop straight into the major agent harnesses. It complements rather than replaces the existing community n8n-skills entry in the catalog, which covers a single-maintainer collection; this is the vendor-maintained source of truth that versions with n8n itself.
Honest limitations: it is n8n-specific, so agents working in LangGraph, raw APIs, or other orchestration layers get no value here. Skill quality tracks n8n releases, and fast-moving node schemas can leave some skills slightly behind. Like all workflow automation, the skills assume you run an n8n instance and understand its credential model.
What it can do
- 14 official vendor-maintained skills
- AI Agent node design and LangChain patterns
- Error handling, debugging, and expressions
- Credentials and security best practices
- MCP extension and workflow lifecycle guides
Setup
Install with the skills CLI or clone github.com/n8n-io/skills and copy the skill folders into your agent's skills directory. Plugin formats for Claude Code, Codex, and OpenClaw are included.
Prerequisites
A compatible agent; a running n8n instance for the workflows the skills describe.
Use cases
Works with
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 88; Trending score 78/100 from researcher assessment.
★ 532 GitHub stars
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.