n8n Automation
Purpose
Operate n8n workflows through connected MCP tools: discover workflows, trigger runs with payloads, inspect executions, and apply focused definition changes safely.
When to Use
- User mentions n8n workflows, nodes, executions, or webhooks.
- Triggering or debugging an automation run.
- Finding why a workflow failed (node error, credentials, payload).
- Enabling/disabling or updating a workflow when write tools exist.
When NOT to Use
- Make or Zapier automations → make, zapier.
- General scripting outside n8n → implementation skills.
- Skill repo authoring → create-skill / publish-skill.
Expected Outcome
- Workflow ID/name and execution status from tool output.
- Error node, message, and I/O snippets when debugging failures.
- Clear boundary when MCP cannot reach the n8n instance.
Inputs to Gather
- Workflow name or ID; environment (prod vs test).
- Trigger payload fields and required credentials.
- Execution ID for failed runs.
- Whether the user wants a one-off test vs production trigger.
Workflow
- List or search workflows if ID unknown.
- Get workflow definition for context (trigger type, key nodes).
- Trigger with validated payload; capture execution ID.
- Poll or fetch execution until terminal state.
- Summarize output data and errors; suggest node-level fixes.
- For edits, confirm impact on scheduled/webhook triggers.
Domain guidance
- Resolve workflow first — name-to-ID lookup before trigger or update.
- Executions are source of truth — never assume success without status from tools.
- Credentials — failures often credential or expression related; cite node name from execution.
- Idempotency — warn on duplicate triggers for billing/side-effect workflows.
- Tool schemas — use only MCP-exposed operations; do not invent REST paths.
Examples
User: "Run the Slack digest workflow for team=platform."
→ Find workflow, trigger with { "team": "platform" }, return execution ID and final status.
User: "Why did last night's sync fail?"
→ Fetch latest failed execution, identify failing node and error message, suggest credential or mapping fix.
Tool Availability Rules
| Access |
Behavior |
| Full tool access |
Trigger, inspect, and apply approved workflow changes. |
| Read-only |
List/get workflows and executions; document change plan for writes. |
| No integration |
Do not fabricate execution results. |
Related tool sets
Review / Decision / Execution Criteria
- Cite execution ID and terminal status in every trigger summary.
- Confirm production triggers that send email, charge, or mutate external systems.
Output Format
- Workflow and action taken.
- Execution ID, status, duration.
- Output summary or error node details.
- Blockers.
- Next steps.
Quality Bar
- No spreadsheet/A1 patterns—n8n is workflow automation only.
Safety and Boundaries
- Do not log secrets from credentials or webhook URLs.
- Confirm destructive or high-volume triggers with the user.
Escalation / Dispatch Rules
- Cross-platform automation comparison → make, zapier, run-automation.
References
skills/old_skills.json (n8n) — legacy catalog (Sheets boilerplate removed).skills/skill.instruction.md, skills/meta.instructions.md
1---2name: n8n3description: Triggers and manages n8n workflows via MCP integration. Use when listing workflows, running executions, inspecting failures, or updating automation definitions in n8n.4---56# n8n Automation78## Purpose910Operate n8n workflows through connected MCP tools: discover workflows, trigger runs with payloads, inspect executions, and apply focused definition changes safely.1112## When to Use1314- User mentions n8n workflows, nodes, executions, or webhooks.15- Triggering or debugging an automation run.16- Finding why a workflow failed (node error, credentials, payload).17- Enabling/disabling or updating a workflow when write tools exist.1819## When NOT to Use2021- Make or Zapier automations → **make**, **zapier**.22- General scripting outside n8n → implementation skills.23- Skill repo authoring → **create-skill** / **publish-skill**.2425## Expected Outcome2627- Workflow ID/name and execution status from tool output.28- Error node, message, and I/O snippets when debugging failures.29- Clear boundary when MCP cannot reach the n8n instance.3031## Inputs to Gather3233- Workflow name or ID; environment (prod vs test).34- Trigger payload fields and required credentials.35- Execution ID for failed runs.36- Whether the user wants a one-off test vs production trigger.3738## Workflow39401. List or search workflows if ID unknown.412. Get workflow definition for context (trigger type, key nodes).423. Trigger with validated payload; capture execution ID.434. Poll or fetch execution until terminal state.445. Summarize output data and errors; suggest node-level fixes.456. For edits, confirm impact on scheduled/webhook triggers.4647## Domain guidance48491. **Resolve workflow first** — name-to-ID lookup before trigger or update.502. **Executions are source of truth** — never assume success without status from tools.513. **Credentials** — failures often credential or expression related; cite node name from execution.524. **Idempotency** — warn on duplicate triggers for billing/side-effect workflows.535. **Tool schemas** — use only MCP-exposed operations; do not invent REST paths.5455### Examples5657**User:** "Run the Slack digest workflow for team=platform."58→ Find workflow, trigger with `{ "team": "platform" }`, return execution ID and final status.5960**User:** "Why did last night's sync fail?"61→ Fetch latest failed execution, identify failing node and error message, suggest credential or mapping fix.6263## Tool Availability Rules6465| Access | Behavior |66|--------|----------|67| Full tool access | Trigger, inspect, and apply approved workflow changes. |68| Read-only | List/get workflows and executions; document change plan for writes. |69| No integration | Do not fabricate execution results. |7071### Related tool sets7273- `n8n`74- `task-manager`7576## Review / Decision / Execution Criteria7778- Cite execution ID and terminal status in every trigger summary.79- Confirm production triggers that send email, charge, or mutate external systems.8081## Output Format82831. Workflow and action taken.842. Execution ID, status, duration.853. Output summary or error node details.864. Blockers.875. Next steps.8889## Quality Bar9091- No spreadsheet/A1 patterns—n8n is workflow automation only.9293## Safety and Boundaries9495- Do not log secrets from credentials or webhook URLs.96- Confirm destructive or high-volume triggers with the user.9798## Escalation / Dispatch Rules99100- Cross-platform automation comparison → **make**, **zapier**, **run-automation**.101102## References103104- `skills/old_skills.json` (`n8n`) — legacy catalog (Sheets boilerplate removed).105- `skills/skill.instruction.md`, `skills/meta.instructions.md`
poly-gents