Automation Workflow Helper

Turns your weekly repetitive tasks into automated bots. Identifies high-value automation opportunities, picks the right tool (Dify / Coze / n8n workflow, or...

Security Scan

The skill's stated purpose (helping automate repetitive tasks) matches its instructions, but the runtime guide recommends persisting and reusing authentication artifacts (cookies / bypassing logins) without declaring how secrets will be requested, stored, or protected — that's an unexplained and risky gap.

Purpose & Capability

Name, description, and SKILL.md content line up: the skill audits tasks, recommends workflow platforms or RPA, and explains tradeoffs. Recommending RPA for sites without APIs is coherent with the stated automation purpose.

Instruction Scope

The instructions explicitly recommend 'Log in manually once to save cookies; auto-injected on every future run' and describe bypassing captchas and login walls. That implies collecting and reusing sensitive authentication artifacts (browser cookies/credentials) and possibly reading browser/session state — but the skill text does not describe where or how these artifacts are stored, who can access them, or whether the agent will read local browser profiles. This expands the agent's runtime scope to sensitive data handling without clear limits.

Install Mechanism

No install spec and no code files (instruction-only). This minimizes filesystem footprint and avoids arbitrary downloads, but also reduces transparency/auditability because behavior is purely in the prose instructions the agent will follow.

Credentials

The skill declares no required environment variables or config paths, yet the guidance expects persistent authentication (cookies/login artifacts). There's a mismatch: handling persistent credentials is central to the proposed RPA flows but the skill does not declare any credential needs, storage locations, or protections. That lack of declared secrets/config is disproportionate and unexplained.

Persistence & Privilege

The skill is not force-installed (always:false), but its recommended practice of saving cookies and auto-injecting them across runs implies persistent sensitive state (effectively long-lived credentials). Combined with autonomous invocation being allowed by default, that persistent state could increase blast radius if misused; the SKILL.md gives no guidance about retention, encryption, or user consent for stored artifacts.

What to consider before installing

Before installing or enabling this skill, ask the author to clarify how authentication artifacts are handled: where cookies/credentials are stored, whether they're encrypted, how long they're retained, and who/what can read them. Refuse to paste passwords into a skill chat; prefer OAuth or API keys with limited scope if possible. If you must test RPA that uses saved sessions, do so with a low-privilege or throwaway account. Prefer skills with visible source (homepage/repo), an install spec, and explicit declarations for any credential or config paths. If you enable the skill, consider disabling autonomous invocation until you verify its storage and data-flow practices.

---

Overview

Time is the scarcest resource. Automation lets you scale without hiring. The rule is simple: any task you do more than twice a week that requires no creative judgment is worth automating. This guide helps you spot opportunities, pick the right tool, and ship fast — with fewer Tokens consumed, more stable execution, and more accurate results every time.

---

Step 1: Find Tasks Worth Automating

Not every task should be automated. Start by finding high-value opportunities.

Automation Audit (spend 1 hour on this):

1. List every repetitive task you did this week
2. For each task, note:
   • How many minutes it takes per run
   • How many times per month you do it
   • Whether the steps are always the same (or vary each time)
3. Calculate time cost:

Monthly hours = minutes per run × monthly frequency ÷ 60

Example: a 15-minute task done 20 times/month = 5 hours/month

4. Sort by monthly hours, highest first

Good candidates for automation:

• ✅ Fully fixed steps, same every time
• ✅ High frequency (at least once a week)
• ✅ Pure data shuffling — no creativity or judgment required
• ✅ Syncing information between multiple tools

Poor candidates:

• ❌ Requires creativity, negotiation, or human judgment
• ❌ Very infrequent (a few times a year)
• ❌ Process changes often

Easiest wins to start with:

• Form submission → auto send email / Slack notification
• Form data → auto save to spreadsheet
• Payment received → auto generate invoice
• Scheduled weekly report delivery to clients
• Data sync between tools (CRM ↔ spreadsheet ↔ email)

---

Step 2: Choose Your Automation Approach

This is the most critical decision. There are two routes — picking the wrong one wastes a lot of time.

Route A: Workflow Platforms

Workflow platforms connect SaaS services via their APIs, passing data and triggering actions between tools. No coding required, built visually.

Best for:

• Cross-platform data sync (form → CRM → email notification)
• Tools that have API / Webhook integrations
• Scheduled tasks, message pushing, data aggregation, AI processing nodes

Top 3 tools compared:

Tool	Open Source	Cloud	Self-hosted	Pricing	Best for	Link
Dify.ai	✅ Open source (MIT) GitHub	✅ cloud.dify.ai, free tier available	✅ Docker self-host	Free cloud tier; Pro $59/mo	AI-powered flows, smart chatbots, document Q&A, LLM nodes	dify.ai
Coze (ByteDance)	❌ Closed source	✅ coze.com	❌ Not supported	Free, pay-as-you-go	Telegram/Discord bots, content pipelines, no-code quick start	coze.com
n8n	✅ Open source (fair-code) GitHub	✅ n8n.cloud, 14-day free trial	✅ Self-host for free	Cloud from $20/mo; self-host free	Complex branching logic, private deployment, data pipelines, developers	n8n.io

How to choose:

• Need AI / LLM processing nodes → Dify.ai
• Integrating Telegram, Discord, or content pipelines; zero-code quick start → Coze
• Private deployment, data must stay on-prem → n8n or Dify.ai self-hosted
• Tight budget → n8n open-source self-host (completely free) or Dify.ai free cloud

---

Route B: RPA (Robotic Process Automation)

RPA automates tasks by directly controlling the browser UI — like an invisible person clicking, filling forms, scraping data, and downloading files. No API needed on the target website.

What is RPA?

Traditional workflow platforms depend on APIs, but many websites (e-commerce platforms, government portals, internal ERPs, various SaaS tools) don't expose public APIs. RPA drives the browser UI directly, bypassing that limitation — if there's a webpage, it can be automated.

RPA vs Workflow Platforms — Core Differences:

Dimension	Workflow Platform	RPA
How it works	Calls API endpoints	Directly controls browser UI
Prerequisite	Target system has an API	Any webpage works, no API needed
Running cost	Platform subscription + per-call fees	Record once, replay for $0 Token
Execution speed	Subject to API and network latency	Runs locally, extremely fast
Typical use cases	SaaS tool integration, data sync	Price scraping, form filling, screenshots, report export
Stability depends on	API interface staying unchanged	Page structure staying unchanged

Core advantages of RPA:

• Zero-Token replay: AI participates only once during recording; every subsequent run is a pure Python script — no AI compute consumed, cost approaches zero
• Bypass login walls: Log in manually once to save cookies; auto-injected on every future run — no more captchas, SMS codes, or QR scans
• Deterministic execution: The script is fixed at recording time; unlike LLM-driven agents that "improvise" each run, results are predictable

Choose RPA when:

• The target website has no public API
• Steps are fully fixed (login → search → scrape data → save)
• High-frequency execution needed (daily or even hourly)
• Task involves clicking, form filling, screenshots, or file downloads

Recommended tool: OpenClaw RPA

An RPA compiler built for AI Agents — describe the task in plain language, AI drives a real browser to record it once, and auto-generates a standalone Playwright Python script. Every subsequent replay runs with zero Token, zero hallucination, in seconds.

Real-world demo videos include e-commerce price monitoring, financial reconciliation, hotel data collection, and stock news briefings: 👉 View README + Video Demos

Install on the OpenClaw platform in one click: 👉 clawhub.ai/laziobird/openclaw-rpa

Or run in terminal:

openclaw skills install openclaw-rpa

After installing, type #RPA in the chat to launch the guided recording flow.

Route decision tree:

Does the target system have an API?
├── Yes → Workflow platform (Dify / Coze / n8n)
└── No  → RPA (type #RPA to start recording)

Need AI judgment / LLM processing?
├── Yes → Dify.ai
└── No, pure rule-based routing → Coze or n8n

---

Step 3: Design Your Flow

Before you build anything, map the flow clearly.

Flow Design Template:

Trigger: What event starts this flow?
  Example: "Every day at 9 AM" / "New form submission received" / "New email arrives"

Condition (optional): What must be true for it to run?
  Example: "Only process orders with status 'Pending Review'"

Actions:
  Step 1: [Do what]
  Step 2: [Do what]
  Step 3: [Do what]

Error handling: What happens if it fails?
  Example: "Send a Slack message to notify me"

Example Flow (Lead Capture → CRM → Notification):

Trigger: New form submission on website

Condition: Phone number field is not empty

Actions:
  Step 1: Write lead to spreadsheet (name, phone, requirement)
  Step 2: Send welcome email to customer
  Step 3: Send message to sales channel: "New lead: [Name] [Phone]"
  Step 4: Create follow-up task in project management tool, remind in 3 days

Error handling: If Step 1 fails, send alert email to me

Design principles:

• Start with a 2–3 step simple version; get it working before adding complexity
• Test each node individually before chaining them together
• Some APIs respond slowly; add delays between nodes if needed
• Always set up failure alerts — silent failures are more dangerous than no automation

---

Step 4: Build and Test

Example setup steps using Dify.ai:

1. Create a new workflow (Dify console → Studio → Create Workflow)
2. Choose a trigger node (e.g. Webhook, scheduled trigger, form input)
3. Connect accounts (enter API Key / OAuth in the Tool or HTTP Request node)
4. Test the trigger (send a test request, confirm data enters the workflow)
5. Add processing nodes (LLM node / code node / HTTP request node)
6. Map variables (match upstream node outputs to downstream node input fields)
7. Test each node individually (click the node's "Run" button, confirm output is correct)
8. Repeat the above, chaining subsequent nodes one by one
9. Publish and activate (click "Publish"; the workflow starts listening for trigger events)

Testing Checklist:

• Run the full flow with test data
• Test edge cases (empty fields, special characters, very long text)
• Test error handling (intentionally break a step, confirm alerts fire correctly)
• Verify field mappings are correct

Common Issues:

Issue	Cause	Fix
Flow doesn't trigger	Trigger condition too strict	Check filter settings, relax conditions
Action fails	API rate limit or insufficient permissions	Add delay between nodes, re-authorize
Data missing or wrong	Field mapping error	Check each field mapping individually
Duplicate triggers	Trigger source emitting duplicate events	Deduplicate by unique ID

Rule: Always do a full end-to-end run with real data before going live. Don't wait for actual customers to discover bugs.

---

Step 5: Monitor and Maintain

Automation is not "set and forget." Tools change, APIs update, page structures shift.

Weekly check (5 minutes):

• Scan workflow run logs for errors
• Handle failures immediately

Monthly audit (15 minutes):

• Review all active automations: still in use? Still saving time?
• Deactivate flows no longer needed (avoid billing waste and logic confusion)
• Update flows that depend on deprecated tools

Documentation recommendations:

• Write a short description for each automation (store in Notion or a shared doc)
• Include: what it does, when it runs, which tools it connects, how to troubleshoot
• Once you have 10+ automations, this doc will save you

Error alert setup:

• Route all automation failure notifications to one place (Slack channel, email, or team chat)
• Review errors weekly, find root causes — don't just patch symptoms

---

Step 6: 6 Classic Scenarios

Scenario 1: Automation Diagnosis for Your Business (Analysis)

Who it's for: You know you have repetitive work but aren't sure what can be automated or which tool to use

How to use in OpenClaw:

Just describe your daily work in detail — the more specific, the better:

Help me analyze which parts of my work can be automated.

I run a small cross-border e-commerce team. My daily routine looks like this:
1. Morning: open Shopify, Amazon, and our own store backend; check yesterday's orders
   from each; copy everything into an Excel summary sheet for the daily report
2. Check competitor prices on each platform and manually record changes
3. When new orders come in, message the warehouse team in Slack with which orders to ship
4. Every Friday, compile weekly sales data using a fixed template and send the report to my boss
5. When launching new products, manually copy product info into three platform backends

Help me analyze: which can be automated? Workflow or RPA? What's the priority order?

---

Scenario 2: E-commerce Competitor Price Monitoring (RPA)

Pain point: Manually visiting multiple e-commerce platforms every day to check competitor prices and copying data into Excel

Solution: #RPA recording, run on a daily schedule

Trigger: Daily scheduled (e.g. 8 AM)
Action: Open each product page → scrape price, sales volume, rating → append to price_monitor.xlsx

How to use in OpenClaw:

Step 1 — type in the chat:

#RPA

Step 2 — after the Agent asks for task name and capability code, enter (two lines):

Competitor Price Monitor
D

Step 3 — paste your task description:

Open https://www.amazon.com/dp/<ASIN>
Scrape: product name, current price, review count
Open https://www.ebay.com/itm/<ItemID>
Scrape: product name, current price, sold count
Append the data to price_monitor.xlsx on the Desktop.
Each row should include a timestamp. Columns: Platform, Product Name, Price, Sales/Reviews, Timestamp

Why RPA instead of a workflow platform?

Dimension	Workflow Platform	RPA (OpenClaw RPA)
Feasibility	❌ Amazon/eBay don't expose public price APIs	✅ Drives browser directly — any page works
Cost per run	—	$0 Token, pure local script
ROI (running once/day)	Not possible	2 hours to record; zero cost every day after — payback in 1 day

---

Scenario 3: Sales Lead Auto-entry into CRM (Workflow)

Pain point: Leads from website forms, landing pages, or event registrations must be manually entered into a spreadsheet or CRM

Solution: Coze or n8n workflow

Trigger: Webhook (form submission)
Step 1: Parse fields (name, phone, source channel)
Step 2: Write to CRM / spreadsheet
Step 3: Notify sales rep in Slack with customer info
Step 4: Auto-send welcome email to customer

How to use in OpenClaw:

Describe your scenario directly in the chat:

I want to automate sales lead entry.
I use Typeform to collect customer info (name, phone, needs).
After submission I have to manually copy it into a Google Sheet and post in our sales Slack channel.
Help me design an automation plan and recommend the right tool with setup steps.

The Agent will guide you through: tool selection → mapping triggers and actions → step-by-step Coze / n8n configuration

---

Scenario 4: AI Content Multi-platform Distribution (Workflow)

Pain point: After writing an article or recording a video, you must manually rewrite copy to match each platform's style before publishing

Solution: Dify.ai workflow

Trigger: New content uploaded (RSS / Webhook)
Step 1: LLM node — rewrite title and summary for each platform style (blog / LinkedIn / Twitter)
Step 2: Publish to each platform via API
Step 3: Generate image caption copy, push draft to review queue

How to use in OpenClaw:

Describe your scenario in the chat:

I write one blog post per week. After finishing it, I have to manually rewrite it
in LinkedIn style and Twitter style before posting.
Can you help me design an automation flow that auto-generates versions for each platform
after I finish the original? I'd prefer a tool that supports AI rewriting. I use Dify.ai.

The Agent will guide you: design the Dify workflow node structure → provide LLM Prompt templates → explain how to connect each platform's API

---

Scenario 5: Financial Reconciliation Report (RPA)

Pain point: Every month, export data from an internal system, manually reconcile with an Excel invoice sheet, write a Word report

Solution: #RPA recording

Trigger: End-of-month scheduled run
Action: Log into internal system → export this month's payables list → compare with local invoices.xlsx → generate Word reconciliation report

How to use in OpenClaw:

Step 1 — type in the chat:

#RPA

Step 2 — after the Agent asks for task name and capability code, enter (two lines):

Financial Reconciliation Report
F

Step 3 — paste your task description:

Log in to https://<your-internal-system>, username <username>, password <password>
Go to the "Accounts Payable" page, filter for this month's data, export as CSV and save to Desktop

Read invoices.xlsx on the Desktop (invoice records).
Match against the exported CSV using the "PO Number" field:
- Match found: mark as "Cleared"
- In system only, no invoice: mark as "Invoice Pending"
- Invoice only, not in system: mark as "Exception"

Generate a Word report saved to Desktop, filename: reconciliation_report_YYYYMMDD.docx
Report includes: summary table (total amount, cleared, pending, exception counts) + detail table

Tip: If the system requires an SMS verification code, run #rpa-login <system-url> first to save the login session. Subsequent recordings auto-inject the cookie — no repeated login needed.

Why RPA instead of a workflow platform?

Dimension	Workflow Platform (with AI nodes)	RPA (OpenClaw RPA)
Feasibility	❌ Internal ERP has no external API	✅ Drives browser to log in and export directly
Token cost per run	LLM node ~$0.05–$0.50/run	$0, pure local script
Cost for 20 runs/month	$1–$10/month (LLM pricing)	$0, record once and run forever
Output format	Requires extra processing for Excel/Word	✅ Natively outputs .xlsx + .docx
ROI	Ongoing cost	3 hours to record, pays back same month, then zero cost forever

---

Scenario 6: New Client Onboarding Automation (Workflow)

Pain point: After signing a contract, you must manually create a project, send a welcome email, provision accounts, and add the client to a group — miss one step and it's embarrassing

Solution: n8n workflow

Trigger: Contract signed (DocuSign / HelloSign Webhook)
Step 1: Create project space in Notion / Google Drive
Step 2: Send welcome email + link to onboarding docs
Step 3: Create system account, send activation email
Step 4: Add client to relevant Slack / Teams channel
Step 5: Update CRM status to "Active"

How to use in OpenClaw:

Describe your scenario in the chat:

After we sign a contract, there's a whole onboarding checklist:
create a project doc in Notion, send a welcome email, add the client to the service Slack channel,
create an account in our system and send an activation email, and update the CRM status.
Right now it's all manual and we keep missing steps.
Help me design an n8n workflow that auto-triggers when the contract is signed and completes all these steps.

The Agent will guide you: confirm which tools have APIs → design the n8n workflow nodes → provide Webhook configuration and field mapping

---

What the Agent outputs:

For each task you describe, it analyzes:

• ✅ Can be automated / ⚠️ Partially automatable / ❌ Keep manual
• Recommended route: Workflow (Dify / Coze / n8n) or RPA (#RPA)
• Priority ranking (by ROI, highest first)
• Quick-start method for each item (prompt or tool configuration entry)

---

Step 7: Calculate Automation ROI

Not every automation is worth the investment. Use this formula to prioritize.

ROI Formula:

Monthly time saved (hours) = minutes per run ÷ 60 × monthly frequency
Build cost = build time (hours) × hourly rate + monthly tool fee
Payback period (months) = build cost ÷ monthly value saved

Payback < 3 months → prioritize
Payback > 6 months → evaluate necessity first

Rule: Focus on automations with a payback period under 3 months — that's the highest ROI zone.

How to use in OpenClaw:

Just tell the Agent your situation and let it do the math:

Help me figure out if this is worth automating.

I check competitor prices on three e-commerce platforms every day — about 30 minutes each time,
roughly 20 times a month.
I plan to use RPA recording, estimate 2 hours to set up, tool is free.
My hourly rate is about $50.

Calculate the payback period and tell me whether it's worth doing.

Agent output:

📊 ROI Calculation

Monthly time saved: 30 min ÷ 60 × 20 runs = 10 hours/month
Monthly value saved: 10 hours × $50 = $500/month

Build cost: 2 hours × $50 = $100 (tool is free, no monthly fee)

Payback period: $100 ÷ $500 = 0.2 months (~6 days)

✅ Verdict: Strongly recommended. Pays back in 6 days; saves $500/month after that.
Recommended: RPA (e-commerce platforms have no public API). Type #RPA to start recording.

You can also ask the Agent to compare priorities across multiple tasks:

I have three things I want to automate. Help me figure out which to do first:

1. Daily competitor price check: 30 min/run, 20 runs/month, ~2 hours to set up (RPA, free)
2. Form lead entry into CRM: 5 min/run, 60 runs/month, ~3 hours to set up (Coze, free)
3. Weekly sales report: 45 min/run, 4 runs/month, ~4 hours to set up (Dify, $15/month)

My hourly rate is $75. Rank them by payback period and give me a recommendation.

---

Common Mistakes to Avoid

• Automating before optimizing: automating a broken process just makes errors happen faster. Fix the process first, then automate.
• Over-automating: low-frequency tasks that require judgment are fine to do manually — don't automate for the sake of automating.
• No error handling: silent failures are more dangerous than doing things manually — data will quietly go wrong. Always set up alerts.
• Insufficient testing: a broken automation is worse than no automation — it creates bad data and missed tasks.
• Starting too complex: get a 2–3 step simple version running first, then add logic gradually.
• No documentation: three months later you won't remember what this workflow does.

---

Professional Services

If you need hands-on help analyzing and implementing your automation:

Service	Price	What's included
Consultation & Diagnosis	$19	Identify automation opportunities and recommend the right AI tools
Single Scenario Solution	$69	Full automation flow design and implementation for one use case
Full Package	$299	Enterprise-grade automation system

Contact Email: [e-mail]