n8n-workflow-automation_skill

Overview

This skill designs and outputs n8n workflow JSON and a runbook focused on resilient, auditable automation. It emphasizes robust triggers, idempotency, structured logging, retries with backoff, and human-in-the-loop review queues so automations do not silently fail.

How this skill works

Given a workflow intent and targets, the skill clarifies trigger type and schedule, defines the input data contract, and chooses deduplication keys and storage. It then produces a node-level design with per-node error branches, retry policies, run_id-based observability, status rows, and a review queue for failed items. By default it returns a read-only design spec; it emits importable n8n JSON and a runbook only when explicitly requested and when credentials/destinations are specified.

When to use it

• Schedule recurring jobs (cron) that must be auditable and retry safely (e.g., weekly compliance emails).
• Build webhook-driven flows that log every run, write status rows, and avoid duplicate processing.
• Add idempotency, error handling, and review queues to existing n8n workflows to prevent silent failures.
• Create human-in-the-loop approval steps for any item requiring manual review before reprocessing.

Best practices

• Always specify a deduplication key and backing store (DB or Sheet) to make flows idempotent.
• Reference credentials via environment variable names or n8n credential references; never embed secrets.
• Instrument each run with a generated run_id and write start/end status rows with error details.
• Use per-node retry with exponential backoff and a final failure branch that enqueues items for human review.
• Stop and alert when aggregate thresholds are breached instead of allowing silent degradation.

Example use cases

• Cron every Monday 08:00 Europe/London: aggregate compliance metrics, upload CSV to Drive, email summary, retry on transient failures.
• Webhook: receive inbound JSON, validate schema, deduplicate by external_id, write status row, and log run_id.
• Transform pipeline: fetch records from DB, process, write outputs to target system, and push failures to a review sheet for approval.
• Ad-hoc manual trigger for sensitive tasks that require human approval before final commit.