desktop
CW Coupang Wing

Model

Tasks

Register a product option Resolve a delayed shipment Process a return request Update warehouse stock Download a sales report
1 2 3 4 5
RL env

Coupang Wing environment

5 tasks with a model prompt, seeded environment state, and grader contract.

Task 1

Register a product option

Prompt
Open the seeded Coupang Wing product and add the missing 2-pack option with the SKU from the task brief.
Environment
A draft product is present with single-pack and 3-pack options but no 2-pack option.
Grader
Checks that the 2-pack option exists, uses the correct SKU, and the draft is saved.
Task 2

Resolve a delayed shipment

Prompt
Find order CP-9182, update the shipment date to the approved date, and leave the required seller memo.
Environment
The shipping management page has several delayed orders, including CP-9182.
Grader
Checks the target order date, seller memo, and that other delayed orders were not modified.
Task 3

Process a return request

Prompt
Open the return request for the damaged item, approve return pickup, and choose the prepaid return reason.
Environment
A return queue contains mixed reasons, including one damaged item request awaiting seller action.
Grader
Checks that pickup is approved with the correct reason and refund is not issued prematurely.
Task 4

Update warehouse stock

Prompt
Set the available stock for SKU CW-771-B to the count shown in the warehouse note.
Environment
The inventory page includes a warehouse note and a table of similar SKU variants.
Grader
Checks that only SKU CW-771-B is updated and the stock value matches the note.
Task 5

Download a sales report

Prompt
Filter sales to the current month, include cancelled orders, and export the report as an XLSX file.
Environment
The sales report page defaults to last month and excludes cancelled orders.
Grader
Checks report filters, file format, and whether cancelled orders are included.
UseDesktop Evals

Computer-use agent evals.

RL envs Main site Docs Blog