Test help centre search and the article it opens

The contract this flow keeps

The flow covers find, read and respond, which is what a help centre has to do and in that order.

• Search narrows - the result count drops to the matching article rather than returning everything.
• The article is right - its body explains the refund path and the five day timeline.
• Feedback is offered - a helpfulness widget with an existing score.
• A vote registers - the Yes button becoming the selected option.

Asserting the article body rather than its title is the part that matters. A search that returns the right headline over the wrong content is worse than no result, because the customer acts on it.

The objective in plain English

The vote is cast and the selected state asserted, which proves the feedback registered rather than the button merely responding to a click.

---
mode: testing
url: https://my-testing-repo-main.vercel.app/zendesk-clone-app/help?reset=true
max_steps: 40
tags: [zendesk, support, knowledge]
---

# Zendisk 30.4: Help centre search

Catalog objective: search the help centre and open an article.
Key assertion: the relevant article renders with a feedback widget.

## Search for a topic
Type "refund" into "Search" and verify the result count narrows to the matching article.

## Open the article
Click "How refunds are processed" and verify the article body explains that refunds return to the original payment method and card refunds take 5 working days.

## Verify the feedback widget
Verify a card titled "Was this article helpful?" is shown with "👍 Yes" and "👎 No" buttons and "Found helpful by" of "128 of 134 people".

## Cast a vote
Click "👍 Yes" and verify the Yes button becomes the selected option.

The existing helpfulness score is asserted before the vote, giving the interaction a baseline rather than a value that might always have been there.

Assertion by assertion

A search, an article, a widget and a vote.

A helpfulness score is how a content team decides what to rewrite, so a vote that does not register starves them of the signal without any visible error.

Putting it in CI

kane-cli testmd run tests/zendesk/help-center-search_test.md

Add a search with no results, since that empty state is the moment to offer a ticket and is often left blank.

Deflection is measurable and worth watching. Kane CLI in a pipeline covers scheduled checks, and the test.md format covers keeping related flows together.