desktop
NS Naver SmartStore

Model

Tasks

Update a product option Answer a buyer inquiry Prepare a settlement export Handle an exchange request Fix a sale price mismatch
1 2 3 4 5
RL env

Naver SmartStore environment

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

Task 1

Update a product option

Prompt
Open the seeded SmartStore product, add the missing color option, set inventory to 12, and save the product.
Environment
A mock SmartStore admin product already exists with size options and one missing color option.
Grader
Checks that the new option exists, inventory equals 12, and unrelated product fields were not changed.
Task 2

Answer a buyer inquiry

Prompt
Find the unanswered shipping inquiry for order NS-2041 and reply with the provided delivery estimate.
Environment
The inquiry inbox contains multiple messages, including one open question tied to order NS-2041.
Grader
Checks that the correct inquiry is answered, the delivery estimate is included, and the thread is marked complete.
Task 3

Prepare a settlement export

Prompt
Filter settlements for last week, download the CSV, and name it with the store code shown in the dashboard.
Environment
The settlement page contains several date ranges and a visible store code in the account panel.
Grader
Checks that the exported file uses the requested date range and filename convention.
Task 4

Handle an exchange request

Prompt
Open the pending exchange request, confirm the replacement option, and add the internal note from the task brief.
Environment
One exchange request is pending and includes the customer-selected replacement option.
Grader
Checks that only the target exchange is updated and the internal note matches the brief.
Task 5

Fix a sale price mismatch

Prompt
Compare the campaign price against the product detail price and update the product detail page to match the campaign.
Environment
The campaign page and product editor show different prices for the same product.
Grader
Checks that both prices match the campaign value and the campaign settings remain unchanged.
UseDesktop Evals

Computer-use agent evals.

RL envs Main site Docs Blog