fengyiqicoder Gmail Receipt Tracker & Expense Categorizer
Overview
Every SMB owner has thousands of receipts buried in email. Come tax time or expense report time, they spend hours searching, screenshotting, and entering data — and still miss deductions. This skill defines a systematic 5-step process to find, categorize, and export receipts so the user can hand a clean file to their bookkeeper / tax prep / accountant.
Core principle: Receipts that aren't categorized at the time of expense get lost. The win is a searchable, categorized, exportable dataset — not just a folder of PDFs.
Workflow
Step 1: Define the scope
Ask the user:
1. Time range? (Last month / quarter / year / specific dates)
2. Email accounts to search? (Personal Gmail, business Gmail, Outlook)
3. Purpose? (Tax prep / monthly expense report / audit / reimbursement)
4. Schedule C or 1120 / 1065? (different categorization)
5. Required output? (CSV for QuickBooks, PDF folder, Excel report)
Step 2: Build the search query strategy
Don't search "receipt". That misses 80%. Use vendor-specific + transaction-trigger words searches.
See references/search-queries.md for the full query library. Key patterns:
# By common trigger words
("your receipt" OR "order confirmation" OR "payment received" OR "invoice" OR "thank you for your purchase" OR "subscription renewed")
# By vendor (top SMB expense categories)
from:(uber OR lyft OR doordash) → Travel/Meals
from:(aws OR digitalocean OR vercel) → Software/Hosting
from:(amazon OR amazonbusiness) → Supplies (parse subject for category)
from:(stripe OR square) → Could be income OR expense — check direction
from:(linkedin OR facebook OR google) → Marketing/Ads
from:(quickbooks OR gusto OR rippling) → Software (subscription)
Step 3: Extract structured data from each email
For each receipt, capture:
| Field | How to extract |
|---|---|
| Date | Email date (or transaction date in body if different) |
| Vendor | From-field domain |
| Description | Subject line + first product line |
| Amount | Look for total; use largest dollar figure if multiple |
| Currency | Default USD; check for symbols/codes |
| Payment method | Last 4 of card if present |
| Category | Map vendor → category (see references) |
| Receipt link | Email permalink or attached PDF |
Step 4: Categorize for tax purposes
Use the IRS Schedule C / 1120 categories. See references/tax-categories.md for the full vendor → category mapping.
Top 8 SMB expense categories (covers 80% of expenses):
- Office supplies & equipment — Amazon, Staples, Best Buy (under $2,500)
- Software subscriptions — SaaS tools
- Cloud hosting & infrastructure — AWS, GCP, Vercel, etc.
- Travel — Flights, hotels, rideshare
- Meals & entertainment — Restaurants (50% deductible federally)
- Marketing & advertising — Google Ads, Meta Ads, LinkedIn Ads
- Professional services — Lawyers, accountants, contractors
- Telecommunications — Phone, internet (business portion)
Step 5: Export the dataset
Default output: CSV that imports cleanly to QuickBooks or Xero.
date,vendor,description,amount,currency,category,payment_method,receipt_link,notes
2026-05-01,Uber,Trip to airport SFO→home,42.50,USD,Travel,Card-1234,gmail://thread/abc,Client meeting
2026-05-02,AWS,May usage,847.23,USD,Software/Hosting,Card-5678,gmail://thread/def,
Also offer:
- PDF zip — all receipt PDFs in one folder, named
YYYY-MM-DD_vendor_amount.pdf - Tax summary: subtotals by category, ready for tax filing
What goes in "Personal" vs "Business"
Ask the user upfront. If they don't have a separate business card, surface every charge for them to flag.
Rules of thumb (US — verify with their CPA):
- Coffee with client = business meal
- Coffee alone = personal
- Home office utilities = % of square footage business use
- Vehicle = mileage log, NOT total payments
- Personal subscription used >50% for business = partial business
Red flags to surface
- 🚨 Vendor that looks unusual / not recognized — could be fraud
- 🚨 Duplicate charges — same vendor, same amount, same day
- 🚨 Subscription you forgot you had — recurring charges with no recent login
- ⚠️ Cash-equivalent transfers (Venmo, Zelle to a person) — need separate accounting
- ⚠️ Large round numbers ($5,000 even) — verify they're not estimates
- ⚠️ Foreign currency charges — surface FX rate and amount in USD
Common mistakes
| Mistake | Why it's bad | Right approach |
|---|---|---|
| Categorizing every Amazon order as "Office Supplies" | Some are personal, some are equipment >$2,500 (capitalized) | Parse subject for product type |
| Missing recurring subscriptions | They look like noise but add up | Group by from: and surface monthly recurring |
| Categorizing Uber as "Travel" always | Uber Eats is "Meals"; commuter rides not deductible | Read subject for "Eats" vs ride |
| Including refunds as expenses | Inflates spending | Search for "refund", "credit" — net them out |
| Not capturing the receipt itself | When IRS asks, summary isn't enough | Always include receipt link / PDF |
Quarterly expense review ritual
Suggest the user run this skill quarterly (not just at tax time):
- Q1: Year-end + January catch-up. Most thorough.
- Q2: Mid-year tax estimate basis
- Q3: Mid-year cleanup
- Q4: Year-end prep — includes any straggler receipts
Annual all-at-once is 3-5x harder than quarterly.
What you don't do
- ❌ Don't make up amounts when you can't read the receipt — flag as "needs verification"
- ❌ Don't categorize personal expenses as business "to save tax" — fraud
- ❌ Don't promise tax deductibility — you're not their CPA
- ❌ Don't delete original emails — keep audit trail
- ❌ Don't ignore receipts <$1 — many small subscriptions add up