スキル一覧に戻る
raulisai

n8n-expert

by raulisai

0🍴 0📅 2026年1月24日
GitHubで見るManusで実行

SKILL.md


name: n8n-expert description: Expert in building, troubleshooting, and optimizing n8n automation workflows. Use when creating new workflows, configuring nodes (Edit Fields, Code, AI Agents), writing JavaScript for transformations, or integrating n8n with external services.

n8n Expert Skill

This skill enables the agent to design, build, and optimize professional automation workflows in n8n. It covers everything from basic node configuration to advanced AI-driven agents and system integrations.

🎯 Core Principles

ALWAYS prioritize native nodes over custom code:

  1. Search for native nodes first: n8n has 400+ integrations.
  2. Centralize Configuration: Use a Set node (labeled "Set: User Config") at the start of the workflow to define IDs, URLs, and toggles.
  3. Multi-Block Documentation: Use separate Sticky Note nodes for:
    • Note: Overview: What the workflow does.
    • Note: Setup: Required credentials and database IDs.
    • Note: Compliance: n8n best practices followed.
  4. Use Edit Fields (Set): Prefer this for 90% of data mappings.
  5. Use Code node only as a last resort: Use it for complex logic, multi-item filtering, or advanced data structures that native nodes can't handle.
  6. Structured Triggers: Always start with a clear entry point (Webhook, Schedule, Chat, etc.).
  7. Documentation: Use Sticky Notes to explain workflow logic for humans.

DOcumentacion resources:

Skill: n8n Management CLI (n8n.bat)

This skill allows you to manage n8n workflows through a simplified command-line interface.

Setup

The CLI is already configured with an API Key and Base URL in the .env file within the workspace.

Available Commands

  • .\n8n list: Lists all workflows in the connected n8n instance.
  • .\n8n get <ID>: Retrieves the JSON structure of a specific workflow.
  • .\n8n create <file.json>: Creates a new workflow from a JSON file.
  • .\n8n update <ID> <file.json>: Updates an existing workflow.
  • .\n8n activate <ID>: Activates a workflow.
  • .\n8n deactivate <ID>: Deactivates a workflow.
  • .\n8n run <ID> [--data '{"key":"value"}']: Executes a workflow with optional input data.
  • .\n8n config --api-key <KEY> --url <URL>: Updates the credentials in .env.

Key Files

  • n8n.bat: The entry point for commands (Windows).
  • n8n_manager.py: The Python script handling logic.
  • .env: Stores N8N_API_KEY and N8N_URL.

Best Practices

  • Always check .\n8n list to find IDs before performing operations.
  • Workflows are stored in the workflow/ directory. Use these files for create or update commands.
  • For automation, use the .\n8n run command to trigger workflows via API.

📦 Essential Node Reference

  • Chat Trigger: @n8n/n8n-nodes-langchain.chatTrigger (Specifically for LangChain-based AI agents). Use version 1.4+.
  • Service Triggers: Telegram, Discord, Google Sheets (on change), etc.

2. Modern AI & LangChain (The Gold Standard)

  • AI Agent Node: @n8n/n8n-nodes-langchain.agent (Use v3.1+).
    • Telegram Trigger Logic (CRITICAL): When connecting a Telegram Trigger directly to an AI Agent, you MUST set:
      • promptType: "define"
      • text: "={{ $json.message.text }}"
    • Without this, the Agent won't receive the user's message as input.
  • Language Model: @n8n/n8n-nodes-langchain.lmChatOpenAi (OpenAI) or @n8n/n8n-nodes-langchain.lmChatGoogleGemini (Gemini v1).
  • Memory: @n8n/n8n-nodes-langchain.memoryBufferWindow (v1.3+).
    • Telegram Pattern: When using Telegram Trigger, you MUST set Session ID to Specify Below (sessionIdType: "customKey") and use {{ $json.message.from.id }} as the sessionKey. Standard Chat Triggers provide sessionId automatically, but Telegram does not.
  • Tools (CRITICAL): MUST use specific Tool node types (e.g., n8n-nodes-base.gmailTool v2.2+).
  • Connection Logic (MAXIMUM CRITICALITY):
    • The keys in the "connections" object MUST be the exact "name" of the source node.
    • Inside the array, the "node" property MUST be the exact "name" of the target node.
    • For AI/LangChain:
      • Source Port: ai_languageModel, ai_memory, or ai_tool.
      • Target Type: ai_languageModel, ai_memory, or ai_tool.
      • Example: "Source Node Name": { "ai_tool": [[{ "node": "Target Node Name", "type": "ai_tool", "index": 0 }]] }.
  • Mapping parameters: Use {{ /*n8n-auto-generated-fromAI-override*/ $fromAI(...) }}.

3. Core Workflow Nodes (Native)

  • Data Transformation:
    • n8n-nodes-base.set (v3.4+): Use for assigning values.
    • n8n-nodes-base.code (v2): Use for complex JS.
  • Integration:
    • n8n-nodes-base.httpRequest (v4.2+): Support for genericCredentialType and various auth methods.
    • n8n-nodes-base.googleSheets (v4.5+): For reading/updating sheets with OAuth2.
  • Control & Logic:
    • n8n-nodes-base.if (v2.2): Use for conditional branching.
    • n8n-nodes-base.wait (v1.1): Essential for polling loops.
  • Documentation:
    • n8n-nodes-base.stickyNote (v1): Use for detailed markdown guides within the workflow.

🛠️ MCP Integration (The "Agentic" Way)

When interacting with n8n via the n8n-mcp server, use these tools to bridge the gap between code and automation:

  • search_workflows: Find existing automation logic by name or description.
  • get_workflow_details: Retrieve the full JSON, nodes, and trigger URLs of a specific workflow.
  • execute_workflow: Trigger a workflow remotely (useful for starting an automation from a chat).
  • list_workflows: Overview of all available automations.

Workflow Management Patterns

  1. Audit: Use get_workflow_details to verify if a workflow follows best practices (error handling, documentation).
  2. Onboarding: When asked to "learn" a setup, list and get details of key workflows.
  3. Execution: After building a logic, use execute_workflow to test the end-to-end flow.

🛠️ Advanced Patterns & Best Practices

The "$fromAI" Pattern

When building tools for an AI Agent, use the $fromAI expression to map agent decisions to node parameters.

// Example for a Gmail tool parameter
"={{ /*n8n-auto-generated-fromAI-override*/ $fromAI('Subject', 'The subject of the email', 'string') }}"

Error Handling

  • Use the Error Trigger node to catch failures globally.
  • Enable Continue on Fail on individual nodes where 100% success isn't critical.
  • Use IF nodes after critical steps to check for success flags.

Polling Pattern (Async API)

  1. Initiate: HTTP Request to start a job (returns request_id).
  2. Loop Start: Wait node (e.g., 60s).
  3. Check Status: HTTP Request using {{ $json.request_id }}.
  4. Decision: IF node checking {{ $json.status }} == 'COMPLETED'.
    • False: Connect back to the Wait node.
    • True: Proceed to Fetch Result.

Ritual Design Pattern

  1. Trigger: Cron, Schedule, or Chat Trigger.
  2. Config: Set node with all user-changeable variables.
  3. Intelligence: AI Agent or specialized nodes (Notion, Telegram).
  4. Documentation: 3-block Sticky Note system (Overview, Setup, Compliance).

Compliance Standards

  • Standardized Labels: Use clear labels like Set: User Config, IF: Success?, Telegram: Notify.
  • No Hardcoded Keys: All IDs should come from the Config node or Credentials.
  • Rich Documentation: Use H2 headers and emojis in notes.

🚀 Common Workflow Templates

  1. Personal AI Assistant: Telegram Trigger -> AI Agent (with Memory + Tools for Gmail/Calendar).
  2. Data Analyst: Web Chat Trigger -> AI Agent (with Google Sheets Tool).
  3. Event Router: Webhook -> Switch (by severity) -> Slack/Email/Database.
  4. Scheduled Reporter: Schedule -> HTTP Request -> Edit Fields -> Gmail Send.

📝 Checklists

  • Does a native node exist for this service?
  • Are all expressions using the latest $json syntax?
  • Is there error handling for critical API calls?
  • Did I use Sticky Notes to document the "why"?
  • If using AI, is the system prompt clear and the tools correctly mapped with $fromAI?

スコア

総合スコア

50/100

リポジトリの品質指標に基づく評価

SKILL.md

SKILL.mdファイルが含まれている

+20
LICENSE

ライセンスが設定されている

0/10
説明文

100文字以上の説明がある

0/10
人気

GitHub Stars 100以上

0/15
最近の活動

3ヶ月以内に更新がある

0/10
フォーク

10回以上フォークされている

0/5
Issue管理

オープンIssueが50未満

+5
言語

プログラミング言語が設定されている

+5
タグ

1つ以上のタグが設定されている

0/5

レビュー

💬

レビュー機能は近日公開予定です