Workflow Automation
Workflow automation is the infrastructure that makes AI agents reliable. Without durable execution, a network hiccup during a 10-step payment flow means lost money and angry customers. With it, workflows resume exactly where they left off.
This skill covers the platforms (n8n, Temporal, Inngest) and patterns (sequential, parallel, orchestrator-worker) that turn brittle scripts into production-grade automation.
Key insight: The platforms make different tradeoffs. n8n optimizes for accessibility, Temporal for correctness, Inngest for developer experience. Pick based on your actual needs, not hype.
Principles
- Durable execution is non-negotiable for money or state-critical workflows
- Events are the universal language of workflow triggers
- Steps are checkpoints - each should be independently retryable
- Start simple, add complexity only when reliability demands it
- Observability isn't optional - you need to see where workflows fail
- Workflows and agents co-evolve - design for both
Capabilities
- workflow-automation
- workflow-orchestration
- durable-execution
- event-driven-workflows
- step-functions
- job-queues
- background-jobs
- scheduled-tasks
Scope
- multi-agent-coordination → multi-agent-orchestration
- ci-cd-pipelines → devops
- data-pipelines → data-engineer
- api-design → api-designer
Tooling
Platforms
- n8n - When: Low-code automation, quick prototyping, non-technical users Note: Self-hostable, 400+ integrations, great for visual workflows
- Temporal - When: Mission-critical workflows, financial transactions, microservices Note: Strongest durability guarantees, steeper learning curve
- Inngest - When: Event-driven serverless, TypeScript codebases, AI workflows Note: Best developer experience, works with any hosting
- AWS Step Functions - When: AWS-native stacks, existing Lambda functions Note: Tight AWS integration, JSON-based workflow definition
- Azure Durable Functions - When: Azure stacks,.NET or TypeScript Note: Good AI agent support, checkpoint and replay
Patterns
Sequential Workflow Pattern
Steps execute in order, each output becomes next input
When to use: Content pipelines, data processing, ordered operations
SEQUENTIAL WORKFLOW:
""" Step 1 → Step 2 → Step 3 → Output ↓ ↓ ↓ (checkpoint at each step) """
Inngest Example (TypeScript)
""" import {inngest} from "./client";
export const processOrder = inngest.createFunction({id: "process-order"}, {event: "order/created"}, async ({event, step}) => {// Step 1: Validate order const validated = await step.run("validate-order", async () => {return validateOrder(event.data.order);});
// Step 2: Process payment (durable - survives crashes)
const payment = await step.run("process-payment", async () => {
return chargeCard(validated.paymentMethod, validated.total);
});
// Step 3: Create shipment
const shipment = await step.run("create-shipment", async () => {
return createShipment(validated.items, validated.address);
});
// Step 4: Send confirmation
await step.run("send-confirmation", async () => {
return sendEmail(validated.email, { payment, shipment });
});
return { success: true, orderId: event.data.orderId };}); """
Temporal Example (TypeScript)
""" import {proxyActivities} from '@temporalio/workflow'; import type * as activities from './activities';
const {validateOrder, chargeCard, createShipment, sendEmail} = proxyActivities({startToCloseTimeout: '30 seconds', retry: {maximumAttempts: 3, backoffCoefficient: 2,}});
export async function processOrderWorkflow(order: Order): Promise {const validated = await validateOrder(order); const payment = await chargeCard(validated.paymentMethod, validated.total); const shipment = await createShipment(validated.items, validated.address); await sendEmail(validated.email, {payment, shipment});} """
n8n Pattern
""" [Webhook: order.created] ↓ [HTTP Request: Validate Order] ↓ [HTTP Request: Process Payment] ↓ [HTTP Request: Create Shipment] ↓ [Send Email: Confirmation]
Configure each node with retry on failure. Use Error Trigger for dead letter handling. """
Parallel Workflow Pattern
Independent steps run simultaneously, aggregate results
When to use: Multiple independent analyses, data from multiple sources
PARALLEL WORKFLOW:
""" ┌→ Step A ─┐ Input ──┼→ Step B ─┼→ Aggregate → Output └→ Step C ─┘ """
Inngest Example
""" export const analyzeDocument = inngest.createFunction({id: "analyze-document"}, {event: "document/uploaded"}, async ({event, step}) => {// Run analyses in parallel const [security, performance, compliance] = await Promise.all([step.run("security-analysis", () => analyzeForSecurityIssues(event.data.document)), step.run("performance-analysis", () => analyzeForPerformance(event.data.document)), step.run("compliance-analysis", () => analyzeForCompliance(event.data.document)),]);
// Aggregate results
const report = await step.run("generate-report", () =>
generateReport({ security, performance, compliance })
);
return report;}); """
AWS Step Functions (Amazon States Language)
""" {"Type": "Parallel", "Branches": [{"StartAt": "SecurityAnalysis", "States": {"SecurityAnalysis": {"Type": "Task", "Resource": "arn:aws:lambda:...:security-analyzer", "End": true}}}, {"StartAt": "PerformanceAnalysis", "States": {"PerformanceAnalysis": {"Type": "Task", "Resource": "arn:aws:lambda:...:performance-analyzer", "End": true}}}], "Next": "AggregateResults"} """
Orchestrator-Worker Pattern
Central coordinator dispatches work to specialized workers
When to use: Complex tasks requiring different expertise, dynamic subtask creation
ORCHESTRATOR-WORKER PATTERN:
""" ┌─────────────────────────────────────┐ │ ORCHESTRATOR │ │ - Analyzes task │ │ - Creates subtasks │ │ - Dispatches to workers │ │ - Aggregates results │ └─────────────────────────────────────┘ │ ┌───────────┼───────────┐ ▼ ▼ ▼ ┌───────┐ ┌───────┐ ┌───────┐ │Worker1│ │Worker2│ │Worker3│ │Create │ │Modify │ │Delete │ └───────┘ └───────┘ └───────┘ """
Temporal Example
""" export async function orchestratorWorkflow(task: ComplexTask) {// Orchestrator decides what work needs to be done const plan = await analyzeTask(task);
// Dispatch to specialized worker workflows const results = await Promise.all(plan.subtasks.map(subtask => {switch (subtask.type) {case 'create': return executeChild(createWorkerWorkflow, {args: [subtask]}); case 'modify': return executeChild(modifyWorkerWorkflow, {args: [subtask]}); case 'delete': return executeChild(deleteWorkerWorkflow, {args: [subtask]});}}));
// Aggregate results return aggregateResults(results);} """
Inngest with AI Orchestration
""" export const aiOrchestrator = inngest.createFunction({id: "ai-orchestrator"}, {event: "task/complex"}, async ({event, step}) => {// AI decides what needs to be done const plan = await step.run("create-plan", async () => {return await llm.chat({messages: [{role: "system", content: "Break this task into subtasks..."}, {role: "user", content: event.data.task}]});});
// Execute each subtask as a durable step
const results = [];
for (const subtask of plan.subtasks) {
const result = await step.run(`execute-${subtask.id}`, async () => {
return executeSubtask(subtask);
});
results.push(result);
}
// Final synthesis
return await step.run("synthesize", async () => {
return synthesizeResults(results);
});}); """
Event-Driven Trigger Pattern
Workflows triggered by events, not schedules
When to use: Reactive systems, user actions, webhook integrations
EVENT-DRIVEN TRIGGERS:
Inngest Event-Based
""" // Define events with TypeScript types type Events = {"user/signed.up": {data: {userId: string; email: string};}; "order/completed": {data: {orderId: string; total: number};};};
// Function triggered by event export const onboardUser = inngest.createFunction({id: "onboard-user"}, {event: "user/signed.up"}, // Trigger on this event async ({event, step}) => {// Wait 1 hour, then send welcome email await step.sleep("wait-for-exploration", "1 hour");
await step.run("send-welcome", async () => {
return sendWelcomeEmail(event.data.email);
});
// Wait 3 days for engagement check
await step.sleep("wait-for-engagement", "3 days");
const engaged = await step.run("check-engagement", async () => {
return checkUserEngagement(event.data.userId);
});
if (!engaged) {
await step.run("send-nudge", async () => {
return sendNudgeEmail(event.data.email);
});
}});
// Send events from anywhere await inngest.send({name: "user/signed.up", data: {userId: "123", email: "user@example.com"}}); """
n8n Webhook Trigger
""" [Webhook: POST /api/webhooks/order] ↓ [Switch: event.type] ↓ order.created [Process New Order Subworkflow] ↓ order.cancelled [Handle Cancellation Subworkflow] """
Retry and Recovery Pattern
Automatic retry with backoff, dead letter handling
When to use: Any workflow with external dependencies
RETRY AND RECOVERY:
Temporal Retry Configuration
""" const activities = proxyActivities({startToCloseTimeout: '30 seconds', retry: {initialInterval: '1 second', backoffCoefficient: 2, maximumInterval: '1 minute', maximumAttempts: 5, nonRetryableErrorTypes: ['ValidationError', // Don't retry validation failures 'InsufficientFunds', // Don't retry payment failures]}}); """
Inngest Retry Configuration
""" export const processPayment = inngest.createFunction({id: "process-payment", retries: 5, // Retry up to 5 times}, {event: "payment/initiated"}, async ({event, step, attempt}) => {// attempt is 0-indexed retry count
const result = await step.run("charge-card", async () => {
try {
return await stripe.charges.create({...});
} catch (error) {
if (error.code === 'card_declined') {
// Don't retry card declines
throw new NonRetriableError("Card declined");
}
throw error; // Retry other errors
}
});
return result;}); """
Dead Letter Handling
""" // n8n: Use Error Trigger node [Error Trigger] ↓ [Log to Error Database] ↓ [Send Alert to Slack] ↓ [Create Ticket in Jira]
// Inngest: Handle in onFailure export const myFunction = inngest.createFunction({id: "my-function", onFailure: async ({error, event, step}) => {await step.run("alert-team", async () => {await slack.postMessage({channel: "#errors", text: Function failed: ${error.message}});});}}, {event: "..."}, async ({step}) => {...}); """
Scheduled Workflow Pattern
Time-based triggers for recurring tasks
When to use: Daily reports, periodic sync, batch processing
SCHEDULED WORKFLOWS:
Inngest Cron
""" export const dailyReport = inngest.createFunction({id: "daily-report"}, {cron: "0 9 * * *"}, // Every day at 9 AM async ({step}) => {const data = await step.run("gather-metrics", async () => {return gatherDailyMetrics();});
await step.run("generate-report", async () => {
return generateAndSendReport(data);
});});
export const syncInventory = inngest.createFunction({id: "sync-inventory"}, {cron: "*/15 * * * *"}, // Every 15 minutes async ({step}) => {await step.run("sync", async () => {return syncWithSupplier();});}); """
Temporal Cron Workflow
""" // Schedule workflow to run on cron const handle = await client.workflow.start(dailyReportWorkflow, {taskQueue: 'reports', workflowId: 'daily-report', cronSchedule: '0 9 * * *', // 9 AM daily}); """
n8n Schedule Trigger
""" [Schedule Trigger: Every day at 9:00 AM] ↓ [HTTP Request: Get Metrics] ↓ [Code Node: Generate Report] ↓ [Send Email: Report] """
Sharp Edges
Non-Idempotent Steps in Durable Workflows
Severity: CRITICAL
Situation: Writing workflow steps that modify external state
Symptoms: Customer charged twice. Email sent three times. Database record created multiple times. Workflow retries cause duplicate side effects.
Why this breaks: Durable execution replays workflows from the beginning on restart. If step 3 crashes and the workflow resumes, steps 1 and 2 run again. Without idempotency keys, external services don't know these are retries.
Recommended fix:
ALWAYS use idempotency keys for external calls:
Stripe example:
await stripe.paymentIntents.create({amount: 1000, currency: 'usd', idempotency_key: order-${orderId}-payment # Critical!});
Email example:
await step.run("send-confirmation", async () => {const alreadySent = await checkEmailSent(orderId); if (alreadySent) return {skipped: true}; return sendEmail(customer, orderId);});
Database example:
await db.query(INSERT INTO orders (id,...) VALUES ($1,...) ON CONFLICT (id) DO NOTHING, [orderId]);
Generate idempotency key from stable inputs, not random values
Workflow Runs for Hours/Days Without Checkpoints
Severity: HIGH
Situation: Long-running workflows with infrequent steps
Symptoms: Memory consumption grows. Worker timeouts. Lost progress after crashes. "Workflow exceeded maximum duration" errors.
Why this breaks: Workflows hold state in memory until checkpointed. A workflow that runs for 24 hours with one step per hour accumulates state for 24h. Workers have memory limits. Functions have execution time limits.
Recommended fix:
Break long workflows into checkpointed steps:
WRONG - one long step:
await step.run("process-all", async () => {for (const item of thousandItems) {await processItem(item); // Hours of work, one checkpoint}});
CORRECT - many small steps:
for (const item of thousandItems) {await step.run(process-${item.id}, async () => {return processItem(item); // Checkpoint after each});}
For very long waits, use sleep:
await step.sleep("wait-for-trial", "14 days"); // Doesn't consume resources while waiting
Consider child workflows for long processes:
await step.invoke("process-batch", {function: batchProcessor, data: {items: batch}});
Activities Without Timeout Configuration
Severity: HIGH
Situation: Calling external services from workflow activities
Symptoms: Workflows hang indefinitely. Worker pool exhausted. Dead workflows that never complete or fail. Manual intervention needed to kill stuck workflows.
Why this breaks: External APIs can hang forever. Without timeout, your workflow waits forever. Unlike HTTP clients, workflow activities don't have default timeouts in most platforms.
Recommended fix:
ALWAYS set timeouts on activities:
Temporal:
const activities = proxyActivities({startToCloseTimeout: '30 seconds', # Required! scheduleToCloseTimeout: '5 minutes', heartbeatTimeout: '10 seconds', # For long activities retry: {maximumAttempts: 3, initialInterval: '1 second',}});
Inngest:
await step.run("call-api", {timeout: "30s"}, async () => {return fetch(url, {signal: AbortSignal.timeout(25000)});});
AWS Step Functions:
{"Type": "Task", "TimeoutSeconds": 30, "HeartbeatSeconds": 10, "Resource": "arn:aws:lambda:..."}
Rule: Activity timeout < Workflow timeout
Side Effects Outside Step/Activity Boundaries
Severity: CRITICAL
Situation: Writing code that runs during workflow replay
Symptoms: Random failures on replay. "Workflow corrupted" errors. Different behavior on replay than initial run. Non-determinism errors.
Why this breaks: Workflow code runs on EVERY replay. If you generate a random ID in workflow code, you get a different ID each replay. If you read the current time, you get a different time. This breaks determinism.
Recommended fix:
WRONG - side effects in workflow code:
export async function orderWorkflow(order) {const orderId = uuid(); // Different every replay! const now = new Date(); // Different every replay! await activities.process(orderId, now);}
CORRECT - side effects in activities:
export async function orderWorkflow(order) {const orderId = await activities.generateOrderId(); # Recorded const now = await activities.getCurrentTime(); # Recorded await activities.process(orderId, now);}
Also CORRECT - Temporal workflow.now() and sideEffect:
import {sideEffect} from '@temporalio/workflow';
const orderId = await sideEffect(() => uuid()); const now = workflow.now(); # Deterministic replay-safe time
Side effects that are safe in workflow code:
- Reading function arguments
- Simple calculations (no randomness)
- Logging (usually)
Retry Configuration Without Exponential Backoff
Severity: MEDIUM
Situation: Configuring retry behavior for failing steps
Symptoms: Overwhelming failing services. Rate limiting. Cascading failures. Retry storms causing outages. Being blocked by external APIs.
Why this breaks: When a service is struggling, immediate retries make it worse. 100 workflows retrying instantly = 100 requests hitting a service that's already failing. Backoff gives the service time to recover.
Recommended fix:
ALWAYS use exponential backoff:
Temporal:
const activities = proxyActivities({retry: {initialInterval: '1 second', backoffCoefficient: 2, # 1s, 2s, 4s, 8s, 16s... maximumInterval: '1 minute', # Cap the backoff maximumAttempts: 5,}});
Inngest (built-in backoff):
{id: "my-function", retries: 5, # Uses exponential backoff by default}
Manual backoff:
const backoff = (attempt) => {const base = 1000; const max = 60000; const delay = Math.min(base * Math.pow(2, attempt), max); const jitter = delay * 0.1 * Math.random(); return delay + jitter;};
Add jitter to prevent thundering herd
Storing Large Data in Workflow State
Severity: HIGH
Situation: Passing large payloads between workflow steps
Symptoms: Slow workflow execution. Memory errors. "Payload too large" errors. Expensive storage costs. Slow replays.
Why this breaks: Workflow state is persisted and replayed. A 10MB payload is stored, serialized, and deserialized on every step. This adds latency and cost. Some platforms have hard limits (e.g., Step Functions 256KB).
Recommended fix:
WRONG - large data in workflow:
await step.run("fetch-data", async () => {const largeDataset = await fetchAllRecords(); // 100MB! return largeDataset; // Stored in workflow state});
CORRECT - store reference, not data:
await step.run("fetch-data", async () => {const largeDataset = await fetchAllRecords(); const s3Key = await uploadToS3(largeDataset); return {s3Key}; // Just the reference});
const processed = await step.run("process-data", async () => {const data = await downloadFromS3(fetchResult.s3Key); return processData(data);});
For Step Functions, use S3 for large payloads:
{"Type": "Task", "Resource": "arn:aws:states:::s3:putObject", "Parameters": {"Bucket": "my-bucket", "Key.$": "$.outputKey", "Body.$": "$.largeData"}}
Missing Dead Letter Queue or Failure Handler
Severity: HIGH
Situation: Workflows that exhaust all retries
Symptoms: Failed workflows silently disappear. No alerts when things break. Customer issues discovered days later. Manual recovery impossible.
Why this breaks: Even with retries, some workflows will fail permanently. Without dead letter handling, you don't know they failed. The customer waits forever, you're unaware, and there's no data to debug.
Recommended fix:
Inngest onFailure handler:
export const myFunction = inngest.createFunction({id: "process-order", onFailure: async ({error, event, step}) => {// Log to error tracking await step.run("log-error", () => sentry.captureException(error, {extra: {event}}));
// Alert team
await step.run("alert", () =>
slack.postMessage({
channel: "#alerts",
text: `Order ${event.data.orderId} failed: ${error.message}`
})
);
// Queue for manual review
await step.run("queue-review", () =>
db.insert(failedOrders, { orderId, error, event })
);
}}, {event: "order/created"}, async ({event, step}) => {...});
n8n Error Trigger:
[Error Trigger] → [Log to DB] → [Slack Alert] → [Create Ticket]
Temporal: Use workflow.failed or workflow signals
n8n Workflow Without Error Trigger
Severity: MEDIUM
Situation: Building production n8n workflows
Symptoms: Workflow fails silently. Errors only visible in execution logs. No alerts, no recovery, no visibility until someone notices.
Why this breaks: n8n doesn't notify on failure by default. Without an Error Trigger node connected to alerting, failures are only visible in the UI. Production failures go unnoticed.
Recommended fix:
Every production n8n workflow needs:
- Error Trigger node
- Catches any node failure in the workflow - Provides error details and context
- Connected error handling: [Error Trigger] ↓ [Set: Extract Error Details] ↓ [HTTP: Log to Error Service] ↓ [Slack/Email: Alert Team]
- Consider dead letter pattern: [Error Trigger] ↓ [Redis/Postgres: Store Failed Job] ↓ [Separate Recovery Workflow]
Also use:
- Retry on node failures (built-in)
- Node timeout settings
- Workflow timeout
Long-Running Temporal Activities Without Heartbeat
Severity: MEDIUM
Situation: Activities that run for more than a few seconds
Symptoms: Activity timeouts even when work is progressing. Lost work when workers restart. Can't cancel long-running activities.
Why this breaks: Temporal detects stuck activities via heartbeat. Without heartbeat, Temporal can't tell if activity is working or stuck. Long activities appear hung, may timeout, and can't be gracefully cancelled.
Recommended fix:
For any activity > 10 seconds, add heartbeat:
import {heartbeat, activityInfo} from '@temporalio/activity';
export async function processLargeFile(fileUrl: string): Promise {const chunks = await downloadChunks(fileUrl);
for (let i = 0; i < chunks.length; i++) {// Check for cancellation const {cancelled} = activityInfo(); if (cancelled) {throw new CancelledFailure('Activity cancelled');}
await processChunk(chunks[i]);
// Report progress
heartbeat({ progress: (i + 1) / chunks.length });}}
Configure heartbeat timeout:
const activities = proxyActivities({startToCloseTimeout: '10 minutes', heartbeatTimeout: '30 seconds', # Must heartbeat every 30s});
If no heartbeat for 30s, activity is considered stuck
Validation Checks
External Calls Without Idempotency Key
Severity: ERROR
Stripe/payment calls should use idempotency keys
Message: Payment call without idempotency_key. Add idempotency key to prevent duplicate charges on retry.
Email Sending Without Deduplication
Severity: WARNING
Email sends in workflows should check for already-sent
Message: Email sent in workflow without deduplication check. Retries may send duplicate emails.
Temporal Activities Without Timeout
Severity: ERROR
All Temporal activities need timeout configuration
Message: proxyActivities without timeout. Add startToCloseTimeout to prevent indefinite hangs.
Inngest Steps Calling External APIs Without Timeout
Severity: WARNING
External API calls should have timeouts
Message: External API call in step without timeout. Add timeout to prevent workflow hangs.
Random Values in Workflow Code
Severity: ERROR
Random values break determinism on replay
Message: Random value in workflow code. Move to activity/step or use sideEffect.
Date.now() in Workflow Code
Severity: ERROR
Current time breaks determinism on replay
Message: Current time in workflow code. Use workflow.now() or move to activity/step.
Inngest Function Without onFailure Handler
Severity: WARNING
Production functions should have failure handlers
Message: Inngest function without onFailure handler. Add failure handling for production reliability.
Step Without Error Handling
Severity: WARNING
Steps should handle errors gracefully
Message: Step without try/catch. Consider handling specific error cases.
Potentially Large Data Returned from Step
Severity: INFO
Large data in workflow state slows execution
Message: Returning potentially large data from step. Consider storing in S3/DB and returning reference.
Retry Without Backoff Configuration
Severity: WARNING
Retries should use exponential backoff
Message: Retry configured without backoff. Add backoffCoefficient and initialInterval.
Collaboration
Delegation Triggers
- user needs multi-agent coordination -> multi-agent-orchestration (Workflow provides infrastructure, orchestration provides patterns)
- user needs tool building for workflows -> agent-tool-builder (Tools that workflows can invoke)
- user needs Zapier/Make integration -> zapier-make-patterns (No-code automation platforms)
- user needs browser automation in workflow -> browser-automation (Playwright/Puppeteer activities)
- user needs computer control in workflow -> computer-use-agents (Desktop automation activities)
- user needs LLM integration in workflow -> llm-architect (AI-powered workflow steps)
Related Skills
Works well with: multi-agent-orchestration, agent-tool-builder, backend, devops
When to Use
- User mentions or implies: workflow
- User mentions or implies: automation
- User mentions or implies: n8n
- User mentions or implies: temporal
- User mentions or implies: inngest
- User mentions or implies: step function
- User mentions or implies: background job
- User mentions or implies: durable execution
- User mentions or implies: event-driven
- User mentions or implies: scheduled task
- User mentions or implies: job queue
- User mentions or implies: cron
- User mentions or implies: trigger
Limitations
- Use this skill only when the task clearly matches the scope described above.
- Do not treat the output as a substitute for environment-specific validation, testing, or expert review.
- Stop and ask for clarification if required inputs, permissions, safety boundaries, or success criteria are missing.