- Home
- Skills
- Finance & Accounting
- Receipt and Invoice Checks
Works with the AI Agents you already use
Receipt and Invoice Checks
Your agent reads a receipt or invoice into clean fields. Then a plain Python script checks every number before you use it.
$9
Receipt and Invoice Checks
Example session with this skill installed
Check this restaurant receipt before I add it to my expenses: receipt.png
- Read your context and instructions
- Compiled the receipt and invoice
I read the receipt and ran the check script. One problem was found: the tax is higher than the printed rate.
This example is a real run on a receipt photo from the public CORD dataset. The shop name is blurred in the photo.
What I read
| Field | Value |
|---|---|
| Type | Receipt |
| Grilled Baby Potato | 50,500 |
| Hot Tuna | 67,000 |
| Hot Tuna | 67,000, with a 100% discount (free) |
| Subtotal | 117,500 |
| Service charge (6%) | 11,070 |
| Tax (PB1 10%) | 19,557 |
| Total | 148,127 |
| Currency | Not printed |
Check result
"passed": false
"check": "tax_rate"
"message": "Tax of 10% on 128,570 is at most 12,857.00 but the tax says 19,557.
If the photo shows 19,557, the seller charged too much.
Check with the seller before using it."
All other checks passed: the lines add up to the subtotal, and subtotal plus service plus tax equals the printed total.
What it means
I looked at the photo again and it clearly shows 19,557, so I did not change anything. The receipt adds up, but the tax does not match its own 10% rate. 19,557 is exactly 10% of the bill before the free Hot Tuna was taken off (184,500 + 11,070 = 195,570). So the shop charged tax on the item it gave away, about 6,700 more than 10% of what you paid.
Next steps
- Ask the restaurant to correct the tax, or record the receipt as printed and note the difference.
- If you claim tax back, use the corrected amount.
Connects securely to your tools. The creator never sees your data.
About this skill
Your agent reads the receipt or invoice: vendor, dates, line items, tax and total. Then a bundled Python script checks the numbers. The script does the arithmetic, so your agent never has to add things up in its head.
What it checks
- The line items add up to the subtotal, and subtotal plus tax equals the total
- Quantity times price on every line
- Tax against the printed rate, including tax already included in prices
- Thousands separators read as decimals, like 22.000 read as 22
- Dates that could be day first or month first
- Due dates before the issue date, and dates in the future
- VAT, GSTIN and UK VAT number formats
- Currency codes, missing totals, and a file that holds more than one receipt
How it behaves
- It copies what is printed and never guesses. Hidden or missing values stay empty.
- It never changes a number to make a check pass. It shows you what failed so you can look.
- It needs Python 3.10 or newer. Nothing to install, no API keys, no network.
How it was tested
I tested it blind on public receipt and invoice test sets and scored the results against their answer keys. With Claude Sonnet on 40 documents, 93% of receipts and all invoices were read fully correct. With Claude Opus on an earlier version, 97 of 100 receipts were fully correct, and on 26 invoices the only differences were errors in the answer key. No real misreading got past the checks in these runs. Smaller models such as Claude Haiku made more mistakes and sometimes changed numbers to force a pass, so use a larger model.
"Passed" means the numbers agree with each other. It does not prove every field was read correctly.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Finance & Accounting.
Needs Python 3.10 or newer. Checks one receipt or invoice per image. The checks catch numbers that disagree with each other. A misreading that keeps every sum consistent can still pass, so "passed" does not prove every field is right. Hidden or missing values are left empty for you to fill in. Works best with larger models; not recommended with Claude Haiku.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 17 days ago
- Software
Python 3.10+
No packages to install.