n8n-skills: Expert Claude Code skills for n8n workflows
A pack of 14 complementary Claude Code skills that teach AI assistants to build production-ready n8n workflows via the n8n-mcp MCP server and deploy self-hosted n8n.
Evidence shows: The repository contains 14 skills, a router skill, and a hooks layer to guide Claude Code in using n8n-mcp. Hooks are described as 'fail open' and never block tool calls, suggesting least privilege, but no specific permission scopes are provided. User confirmation: Skills are instructional but do not explicitly require user confirmation for high-impact actions. Data flow transparency: Skill descriptions mention data flow (e.g., webhook data in $json.body) but do not explicitly state how data is processed or stored. Sensitive data handling: Skills mention credential management but lack specific security handling details. Dependency security: Depends on n8n-mcp but no security audit provided. External effects: Skills guide deployment and configuration, potentially affecting external systems, but not explicitly stated. Rollback: No rollback mechanism mentioned. Source attribution: Author is clear, but publisher is unverified. Deductions: Lack of specific security mechanisms and permission details.
Evidence shows: Skill descriptions are consistent, e.g., n8n Code JavaScript and n8n Code Tool clearly distinguish between Code node and Code Tool, showing self-consistency. Dependency availability: Depends on n8n-mcp but no availability guarantee. Failure messages: Skills include error catalogs and troubleshooting guides, e.g., n8n Validation Expert and n8n Error Handling, providing failure messages. Deductions: Dependency availability unverified; failure messages exist but not tested.
Evidence shows: Audience is clearly Claude Code users; scenarios include building, validating, deploying n8n workflows. Capability boundaries: Skills clearly distinguish Code node vs Code Tool, and Python vs JavaScript limitations. Trigger precision: Skill descriptions are detailed with clear activation conditions. Environment fit: Supports Claude Code, Claude.ai, and API, but Codex plugin not mentioned. Deductions: Environment fit not fully covered.
Evidence shows: Information architecture is clear with README, docs directory, and skill list. Install notes are detailed with multiple methods. Naming stability: Skill names are consistent. Examples and FAQ: README includes usage examples but no FAQ. Known limitations: Skills mention limitations like Python no external libraries. License: MIT license clear. Versioning: No version number or changelog mentioned. Maintenance responsibility: Author is clear but no maintenance plan. Deductions: Missing versioning and changelog.
Evidence shows: Output usability is high with specific guidance. Marginal value: Skills fill a gap in n8n workflow building. Cost benefit: Skills are free but require n8n-mcp and Claude Code. Deductions: No actual effectiveness data.
Evidence shows: Claims like '525+ n8n nodes supported' and '2,653+ workflow templates' lack specific sources. Cross-source: No other sources provided. Fact-inference separation: Skill descriptions contain facts and inferences but not explicitly separated. Deductions: Claims lack traceability.
- Publisher identity is unverified; proceed with caution.
- Dependency on n8n-mcp is not security-audited.
- Skills guide sensitive operations like credential management but lack specific security measures.
- Claims like '525+ n8n nodes' lack verifiable sources.
What does this agent do, and when should you use it?
n8n-skills is a repository containing 14 complementary Claude Code skills, plus an always-on router skill and a hooks enforcement layer, designed to teach AI assistants how to build production-ready n8n workflows using the n8n-mcp MCP server. It also covers deploying the self-hosted n8n instance. The skills cover n8n expression syntax, MCP tool usage, workflow patterns, validation error handling, node configuration, Code nodes (JavaScript/Python), Code Tool, error handling, binary data, sub-workflows, AI agents, multi-instance management, and self-hosting deployment. Installation options include plugin install for Claude Code, marketplace, manual copy to ~/.claude/skills/, or individual skill upload for Claude.ai. Skills activate automatically based on natural language queries, and the hooks layer provides proactive reminders at the point of decision. The project is MIT-licensed and part of the n8n-mcp ecosystem.
The pack loads a router skill (using-n8n-mcp-skills) via a SessionStart hook into every Claude Code session, and registers PreToolUse and PostToolUse hooks. PreToolUse hooks fire before high-impact n8n-mcp calls (like get_node for specific node types) and display short reminders pointing to relevant skills; n8n_instances and n8n_manage_credentials calls trigger one-shot reminders. PostToolUse hooks run after validate_workflow and route to skills that own remaining risks, emphasizing that validation passing is necessary but not sufficient. The 14 skills (e.g., n8n Expression Syntax, n8n MCP Tools Expert, n8n Workflow Patterns, etc.) activate based on natural language queries and provide detailed guidance on expression syntax, MCP tool selection, workflow patterns, validation error interpretation, node configuration, JavaScript/Python code writing in Code nodes, Code Tool usage, error handling, binary data handling, sub-workflow design, AI agent construction, multi-instance management, and self-hosting deployment. The content is built from real templates (2,653+), supports 525+ n8n nodes, and includes error catalogs and production-tested patterns.
- A developer using Claude Code who wants to build n8n workflows with the n8n-mcp MCP server but frequently hits validation errors or is unsure how to configure nodes correctly.
- A team standardizing on n8n for workflow automation wants AI assistants to follow proven patterns to reduce errors and improve consistency.
- A developer working on n8n workflows that involve Code nodes needs to know correct data access patterns (e.g., $json.body for webhook data) and avoid common JavaScript/Python pitfalls.
- An ops person deploying self-hosted n8n on a Linux VPS seeks an end-to-end guide covering Docker Compose, Caddy reverse proxy, security hardening, and backup/restore.
- A developer building n8n AI agents (using @n8n/n8n-nodes-langchain nodes) wants guidance on tool calling, structured output, memory, and human-in-the-loop review.
- A team managing multiple n8n instances (production/staging) needs to avoid cross-instance credential miswrites and safely switch instances or copy objects.
What are this agent's strengths and limitations?
- Provides 14 complementary skills covering the full n8n development lifecycle, from expression syntax to self-hosting deployment, with depth and practicality beyond typical ad-hoc tips.
- Includes a hooks enforcement layer (PreToolUse/PostToolUse) that proactively reminds at decision points, and all hooks fail open, never blocking tool calls.
- Content is grounded in real n8n templates (2,653+) and MCP tool responses, supports 525+ nodes, and includes error catalogs and production-tested patterns.
- Dedicated multi-instance skill addresses safe switching and credential protection, preventing silent secret miswrites across instances.
- Built exclusively for Claude Code; hooks only work in Claude Code/Codex plugin installs. On Claude.ai, only skill description activation works, lacking proactive nudges.
- Depends on the n8n-mcp MCP server and a configured n8n instance, adding setup overhead; skills may be tied to specific n8n versions.
- No automated test harness is provided for the skills themselves; quality relies on manual evaluation scenarios (npm test may not be implemented).
- Not compatible with OpenAI/ChatGPT platforms, limiting portability across AI assistants.
How do you install or deploy this agent?
Prerequisites: 1) Install and configure the n8n-mcp MCP server (see its installation guide); 2) Have access to Claude Code, Claude.ai, or Claude API; 3) Configure .mcp.json to point to the n8n-mcp server. Installation methods: Method 1: Plugin Install (recommended) — in Claude Code, run /plugin install czlonkowski/n8n-skills. Method 2: Via Marketplace — run /plugin marketplace add czlonkowski/n8n-skills, then /plugin install and select 'n8n-mcp-skills'. Method 3: Manual — git clone https://github.com/czlonkowski/n8n-skills.git, then cp -r n8n-skills/skills/* ~/.claude/skills/, and reload Claude Code. For Claude.ai, download individual skill folders from skills/ (or prebuilt zips from releases) and upload via Settings → Capabilities → Skills. API/SDK installation is documented in docs/INSTALLATION.md.
How do you use this agent?
After installation, skills activate automatically. For example, typing 'How do I write n8n expressions?' triggers the n8n Expression Syntax skill; 'Find me a Slack node' triggers the n8n MCP Tools Expert; 'Build a webhook workflow' triggers n8n Workflow Patterns. Skills compose: for 'Build and validate a webhook to Slack workflow', Workflow Patterns identifies the pattern, MCP Tools Expert searches for nodes, Node Configuration guides setup, Code JavaScript processes data, Expression Syntax maps data, and Validation Expert validates. The hooks layer fires reminders before node lookups and after validation, routing to relevant skills. You can manually test a skill with claude-code --skill n8n-expression-syntax "Test webhook data access".