- Home
- Skills
- Workflow Automation
- Workflow Architect for n8n
Works with the AI Agents you already use
Workflow Architect for n8n
Your agent plans, writes and checks importable n8n workflow JSON with retries, error workflows and sticky notes built in, and returns the workflow in chat as one code block. It never writes files, even when asked.
$12
Workflow Architect for n8n
Example session with this skill installed
Tell me in Slack when any of my workflows fails, with a link to the failed run. I'm on self-hosted n8n 2.43.8. A quick draft is fine; state your assumptions.
- Read your context and instructions
- Compiled the workflow architect
Plan (quick draft, assumptions stated)
- Outcome: when any workflow that points at this one fails, post the workflow name, failed node, error message and a link to the failed run to Slack.
- Trigger: Error Trigger. Test path: Manual Trigger "Run test" feeding a mock error payload in the same shape.
- Handles both error shapes:
execution(normal failures) andtrigger.error(the trigger itself failed, so no run URL). - Slack channel:
#n8n-alerts(assumption, change it in "Alert Slack"). Retries on the Slack call: 3 tries, 5000 ms apart. - Side effects behind a dry-run switch:
dry_runistrue, so nothing is posted until you flip it.
Workflow JSON (copy it into a .json file or paste it onto the n8n canvas):
{
"name": "Error handler: alert with link to failed run (QE example 03)",
"nodes": [
{
"parameters": {},
"id": "f304613c-340d-50a4-932e-cb7d2ab668d8",
"name": "Run test",
"type": "n8n-nodes-base.manualTrigger",
"typeVersion": 1,
"position": [
0,
160
]
},
{
"parameters": {
"content": "## Shared error workflow\n**What it does:** runs when any workflow that points at it fails, and posts who/what/where with a link to the failed execution.\n\n**Use it:** in each workflow open *Settings* then *Error workflow* and select this one. One error workflow can serve many workflows.\n\n**Test:** *Test workflow* sends a mock error payload (same shape the Error Trigger emits).",
"height": 300,
"width": 460
},
"id": "884e9e19-84ab-5c57-939f-4e411e99ac27",
"name": "Note: Overview",
"type": "n8n-nodes-base.stickyNote",
"typeVersion": 1,
"position": [
-80,
-380
]
},
{
"parameters": {
"content": "### Trigger failures\nIf the *trigger* of the watched workflow fails, n8n sends `trigger.error` instead of `execution`, with no execution URL. `Shape alert` handles both shapes.",
"height": 220,
"width": 340,
"color": 3
},
"id": "38aa035b-2f10-55c5-a000-aa910d2f89ac",
"name": "Note: Trigger errors",
"type": "n8n-nodes-base.stickyNote",
"typeVersion": 1,
"position": [
700,
-380
]
},
{
"parameters": {},
"id": "0311cda8-5504-5bb4-b66f-7a3af91c6d49",
"name": "On workflow error",
"type": "n8n-nodes-base.errorTrigger",
"typeVersion": 1,
"position": [
0,
-60
]
},
{
"parameters": {
"jsCode": "// Mirrors the documented Error Trigger output so the alert can be tested without breaking a real workflow.\nreturn [{ json: {\n execution: { id: '231', url: 'https://n8n.example.com/execution/231', retryOf: null,\n error: { message: 'Request failed with status code 503', stack: '...' },\n lastNodeExecuted: 'Fetch orders', mode: 'trigger' },\n workflow: { id: '17', name: 'Nightly order sync' },\n} }];"
},
"id": "d048b6a5-0cde-57c1-a49e-e9e7ede88dcf",
"name": "Mock error payload",
"type": "n8n-nodes-base.code",
"typeVersion": 2,
"position": [
220,
160
]
},
{
"parameters": {
"assignments": {
"assignments": [
{
"id": "9ef72441-0a45-5a9a-a3e2-ef1d802a8844",
"name": "workflow_name",
"value": "={{ $json.workflow?.name || 'unknown workflow' }}",
"type": "string"
},
{
"id": "a8ce56e2-e921-5cb6-9e91-9f42aa884ee2",
"name": "failed_node",
"value": "={{ $json.execution?.lastNodeExecuted || $json.trigger?.error?.node?.name || 'trigger' }}",
"type": "string"
},
{
"id": "e84a7901-ef44-52cb-86c2-2973d46bfc52",
"name": "error_message",
"value": "={{ $json.execution?.error?.message || $json.trigger?.error?.message || 'no message' }}",
"type": "string"
},
{
"id": "172a540e-1688-5e06-a7c2-043f3eec81ce",
"name": "execution_url",
"value": "={{ $json.execution?.url || 'not saved (trigger error or execution saving disabled)' }}",
"type": "string"
},
{
"id": "0bbbcad5-d272-5684-8df4-ecca266fb6b7",
"name": "is_retry",
"value": "={{ Boolean($json.execution?.retryOf) }}",
"type": "boolean"
},
{
"id": "7b20ad68-a2b5-54c5-abac-5e5c6de2ea53",
"name": "dry_run",
"value": true,
"type": "boolean"
},
{
"id": "9e6f9a93-f5b5-597c-8745-e622fef42742",
"name": "text",
"value": "=:rotating_light: *{{ $json.workflow?.name || 'unknown workflow' }}* failed at `{{ $json.execution?.lastNodeExecuted || $json.trigger?.error?.node?.name || 'trigger' }}`\n> {{ $json.execution?.error?.message || $json.trigger?.error?.message || 'no message' }}\n{{ $json.execution?.url || '' }}",
"type": "string"
}
]
},
"options": {}
},
"id": "e5159edd-a194-5d6d-b944-c91960a8094a",
"name": "Shape alert",
"type": "n8n-nodes-base.set",
"typeVersion": 3.4,
"position": [
480,
40
]
},
{
"parameters": {
"conditions": {
"options": {
"caseSensitive": true,
"leftValue": "",
"typeValidation": "strict",
"version": 2
},
"conditions": [
{
"id": "17a0a2a4-a947-52dd-b2a1-b4a067214cf4",
"leftValue": "={{ $json.dry_run }}",
"rightValue": "",
"operator": {
"type": "boolean",
"operation": "true",
"singleValue": true
}
}
],
"combinator": "and"
},
"options": {}
},
"id": "0086a65e-e9eb-54cb-aa5b-7a399db0f922",
"name": "Dry run?",
"type": "n8n-nodes-base.if",
"typeVersion": 2.2,
"position": [
700,
40
]
},
{
"parameters": {},
"id": "9b67ceed-fac9-559e-8a0a-81699b6cbbda",
"name": "Dry run: not sent",
"type": "n8n-nodes-base.noOp",
"typeVersion": 1,
"position": [
920,
-60
]
},
{
"parameters": {
"select": "channel",
"channelId": {
"__rl": true,
"value": "#n8n-alerts",
"mode": "name"
},
"text": "={{ $json.text }}",
"otherOptions": {}
},
"id": "d32e905a-9c21-5558-b49c-18a7657a2c01",
"name": "Alert Slack",
"type": "n8n-nodes-base.slack",
"typeVersion": 2.3,
"position": [
920,
140
],
"credentials": {
"slackApi": {
"id": "REPLACE_WITH_CREDENTIAL_ID",
"name": "Slack account"
}
},
"retryOnFail": true,
"maxTries": 3,
"waitBetweenTries": 5000
}
],
"connections": {
"On workflow error": {
"main": [
[
{
"node": "Shape alert",
"type": "main",
"index": 0
}
]
]
},
"Run test": {
"main": [
[
{
"node": "Mock error payload",
"type": "main",
"index": 0
}
]
]
},
"Mock error payload": {
"main": [
[
{
"node": "Shape alert",
"type": "main",
"index": 0
}
]
]
},
"Shape alert": {
"main": [
[
{
"node": "Dry run?",
"type": "main",
"index": 0
}
]
]
},
"Dry run?": {
"main": [
[
{
"node": "Dry run: not sent",
"type": "main",
"index": 0
}
],
[
{
"node": "Alert Slack",
"type": "main",
"index": 0
}
]
]
}
},
"settings": {
"executionOrder": "v1"
},
"pinData": {},
"meta": {
"templateCredsSetupCompleted": false
},
"tags": []
}
Import: n8n editor > ... menu > Import from File, or paste onto the canvas.
Setup
- "Alert Slack" (Slack): connect your Slack credential (the JSON has the placeholder
REPLACE_WITH_CREDENTIAL_ID). - Set the channel if it isn't
#n8n-alerts. - In "Shape alert", set
dry_runtofalsewhen you're ready to post. - Publish this workflow, then in each workflow you want watched open Settings > Error workflow and select it.
Test: click Test workflow. The mock payload runs through "Shape alert" and stops at "Dry run: not sent" with the alert text for "Nightly order sync" failing at "Fetch orders" (Request failed with status code 503) and the link https://n8n.example.com/execution/231.
Check result
OK error-handler-alert.json (0 errors, 0 warnings)
INFO [credential-placeholder] node "Alert Slack": credentials.slackApi is a placeholder; select a real credential in n8n after importing.
INFO [summary] 7 nodes, 2 sticky notes, 2 trigger(s).
Checked by hand with the validation checklist (C1-C26).
Not tested here: real Slack posting (dry run), and your instance's credentials.
Connects securely to your tools. The creator never sees your data.
About this skill
An agent skill for building n8n workflows as importable JSON: plan first, build with real node types, then check the result before you import it.
Your agent follows it to:
- Plan with you before building: trigger, data in and out, steps, credentials, failure plan, dedupe key, dry-run switch and test cases, in a fill-in template.
- Write importable workflow JSON using node types and typeVersions checked against n8n 2.43.8 node definitions, with credential placeholders (never real keys) and send/write steps behind a dry-run switch.
- Build in error handling: retry settings within n8n's limits, the three On Error modes, a shared Error Trigger workflow, Stop and Error guards and a back-off loop.
- Add sticky notes so every workflow explains itself.
- Return the workflow in chat as one complete code block for you to paste into n8n (it never writes files).
- Check the JSON step by step with a written checklist: syntax, connections, duplicate names, missing credentials, likely hardcoded secrets, retry values, unreachable nodes and typeVersions
- Debug: map import and runtime errors to likely causes and fixes
- Learn from 5 worked example workflows and 4 snippets (lead intake and dedupe, overdue-invoice digest, error alert workflow, API fetch with retries, batched webhook to a spreadsheet).
What it doesn't do: it doesn't cover AI-agent (LangChain) nodes, and it doesn't connect to your n8n. You import and run the workflows yourself. The checklist catches common structural mistakes; it doesn't prove parameters, credentials or third-party APIs are right.
Requirements: uses the open Agent Skills format supported by Claude Code, Cursor, Codex and others. Instructions only, with no scripts to run. You need your own n8n instance to import and run the results.
How it was made: built with AI assistance. The example workflows were imported into a local self-hosted n8n 2.43.8 and run with mock data by automated tests. It hasn't been run inside an AI agent, on n8n Cloud, or on other n8n versions.
Independent; not affiliated with or endorsed by n8n.
- Node types and typeVersions verified against n8n 2.43.8 only; much older instances may lack some typeVersions.
- Doesn't cover AI-agent (LangChain) nodes.
- The checklist can't verify parameters, credentials or API responses. Only a test run in your n8n shows that.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 25 days ago
Requires