BEFORE YOU WRITE · OPEN CHALLENGE

Your agent says it’s ready.
Give it dirty data.

A duplicate product. A missing value. An invoice three cents out. Catch the mistakes before the next import, update or handoff.

3 synthetic cases9 explicit checksFree · no wallet
Take the challenge ↓

CASE 01 · catalog_audit

The supplier feed

A comma is not always a decimal point. Which products need review?

Inspect the input and expected checks
{
  "records": [
    {
      "sku": "A-10",
      "price": "1.234,50"
    },
    {
      "sku": "B-20",
      "price": "29,95"
    },
    {
      "sku": "B-20",
      "price": "29,95"
    },
    {
      "price": "-5,00"
    },
    {
      "sku": "E-50",
      "price": 29.95
    }
  ],
  "id_field": "sku",
  "price_field": "price",
  "locale": "de-DE",
  "currency": "EUR"
}
  • Read the German price as 1234.5
  • Flag both rows with the duplicate ID
  • Catch missing ID, negative price and a non-string price

CASE 02 · record_reconcile

The changing inventory

Is numeric 7 the same ID as text "7"? Is missing the same as null?

Inspect the input and expected checks
{
  "before": [
    {
      "id": 7,
      "stock": 10
    },
    {
      "id": "7",
      "stock": null
    },
    {
      "id": "B",
      "stock": 5
    },
    {
      "id": "D",
      "stock": 1
    }
  ],
  "after": [
    {
      "id": 7,
      "stock": 10
    },
    {
      "id": "7"
    },
    {
      "id": "C",
      "stock": 8
    },
    {
      "id": "D",
      "stock": 2
    }
  ],
  "key": "id",
  "fields": [
    "stock"
  ]
}
  • Find one added, one removed, two changed and one unchanged
  • Keep numeric and string identities distinct
  • Distinguish explicit null from a missing field

CASE 03 · invoice_arithmetic

The three-cent slip

19.99 × 3. The invoice says 60.00. Would your workflow catch it?

Inspect the input and expected checks
{
  "currency": "USD",
  "scale": 2,
  "lines": [
    {
      "quantity": 3,
      "unit_minor": "1999",
      "line_minor": "6000"
    }
  ],
  "subtotal_minor": "6000",
  "tax_minor": "0",
  "shipping_minor": "0",
  "discount_minor": "0",
  "total_minor": "6000"
}
  • Calculate the line as 5997 minor units
  • Calculate the final total as 59.97
  • Flag the line, subtotal and total mismatches

Run it. Inspect it. Keep it.

Run the TaskMint reference engine here, or download the pack to check your own implementation. Browser checks stay on this device.

Download standalone runnerFixtures & output contract
node taskmint-challenge.mjs

This button tests TaskMint’s reference engine on its own fixtures. It does not test your agent.

Test your agent’s output instead

Give your agent the downloadable payloads. Return equivalent TaskMint output objects under catalogue, snapshots and invoice. Compare only the nine published assertions; extra fields are ignored.

node taskmint-challenge.mjs grade your-results.json

Bring one agent. Challenge another.

Run the pack, share what matched or failed, and invite another builder to try the same cases. Include whether you used the reference engine or your own output. There is no sign-up or paid membership.

TaskMint built these open fixtures to demonstrate its tools. Answers are public; they are useful regression checks, not a hidden benchmark or model ranking.

FROM DEMO TO DAILY WORK

Put the check before the write.

Keep the free engine in your workflow, or call TaskMint over HTTP when you need hosted execution. Catalogue audits, snapshot reconciliation and invoice arithmetic each cost 0.005 USDC per completed batch.

Up to 50 records per input or 50 invoice lines. Hosted purchases need your approved compatible wallet, native USDC on Base and an explicit spending limit. No payment is required for this challenge.