<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Katya Pereverzeva on Medium]]></title>
        <description><![CDATA[Stories by Katya Pereverzeva on Medium]]></description>
        <link>https://medium.com/@katyapereverzeva?source=rss-d7e6421e2633------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*3lSDSSxa5seIcwi6_QtPug@2x.jpeg</url>
            <title>Stories by Katya Pereverzeva on Medium</title>
            <link>https://medium.com/@katyapereverzeva?source=rss-d7e6421e2633------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 10:33:06 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@katyapereverzeva/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Prompt frameworks to boost QA efficiency with AI]]></title>
            <link>https://medium.com/@katyapereverzeva/prompt-frameworks-to-boost-qa-efficiency-with-ai-c6f3c10f6d8e?source=rss-d7e6421e2633------2</link>
            <guid isPermaLink="false">https://medium.com/p/c6f3c10f6d8e</guid>
            <category><![CDATA[prompt]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[qa]]></category>
            <category><![CDATA[prompt-engineering]]></category>
            <dc:creator><![CDATA[Katya Pereverzeva]]></dc:creator>
            <pubDate>Thu, 15 May 2025 10:36:16 GMT</pubDate>
            <atom:updated>2025-05-15T10:38:42.354Z</atom:updated>
            <content:encoded><![CDATA[<h3>Prompt Frameworks to Boost QA Efficiency with AI</h3><p><strong>The AI field is rapidly growing and we can’t afford to ignore it.<br></strong> To improve and accelerate our work, many of us have started using chatbots and AI assistants. But the question is: <em>are we using them effectively?</em></p><p>There are multiple prompt techniques and frameworks designed to help us communicate with AI more efficiently. In my experience, many of these approaches share similar structures, but we won’t dive into that comparison just yet :)</p><p>Instead, I want to introduce you to the prompt frameworks I personally use the most as a QA Engineer.</p><h3>1. RTF Framework — Role, Task, Format</h3><p>The first prompt technique I often use is the <strong>RTF framework</strong>, which stands for <strong>Role</strong>, <strong>Task</strong>, and <strong>Format</strong>.</p><p>This framework is about structuring your prompt in three simple parts:</p><ol><li><strong>Role</strong> — Define <em>who</em> the AI should be (e.g, a senior QA engineer, a test automation expert)</li><li><strong>Task</strong> — Clearly state <em>what</em> you want the AI to do</li><li><strong>Format</strong> — Specify <em>how</em> you want the response to be structured (e.g, a list, table, step-by-step guide, or summary)</li></ol><p><strong>In my opinion</strong>, RTF is one of the most effective and lightweight frameworks out there. It offers a clear and concise structure — just enough structure to guide the AI without overcomplicating things.</p><p><strong>RTF framework example: comparing Jira test management plugins</strong></p><p><strong>Prompt using RTF:</strong></p><blockquote><strong>Role:</strong> You are a QA analyst.<br><strong>Task:</strong> Compare the most popular Jira plugins for test management.<br><strong>Format:</strong> Present the comparison in a table with columns: Plugin Name, Key Features, Pros, Cons, Best For.</blockquote><h3>2. RISEN framework — Role, Input, Steps, Expectation, Narrowing</h3><p>The second technique I often use is the <strong>RISEN framework</strong>, which stands for:</p><ul><li><strong>Role</strong> — Define who the AI should be.</li><li><strong>Input</strong> — Describe the context or the information the AI will work with.</li><li><strong>Steps</strong> — Specify if you want a step-by-step process. These are the instructions or directions you give.</li><li><strong>Expectation</strong> — Clarify what kind of output you’re expecting. What do you want at the end? It’s the goal or outcome you’re aiming for with your prompt.</li><li><strong>Narrowing</strong> — Add any constraints or limits to keep the response focused and on-topic.</li></ul><p>I see RISEN as an <strong><em>extended version</em></strong><em> </em>of RTF — it gives the AI more guidance, especially for complex or multi-step tasks. It’s great when you need more control over the structure and depth of the answer.</p><p><strong>RISEN framework example: comparing Jira test management plugins (again, but with more details)</strong></p><p><strong>Prompt using RISEN:</strong></p><blockquote><strong>Role:</strong> You are a QA analyst.</blockquote><blockquote><strong>Input:</strong> You are analyzing test management tools used by QA teams inside Jira.</blockquote><blockquote><strong>Steps:</strong> Go through each plugin and describe its core features, strengths, and limitations.</blockquote><blockquote><strong>Expectation:</strong> Present the comparison in a table format.</blockquote><blockquote><strong>Narrowing:</strong> Only include tools that provide integration with automation testing frameworks or pipelines</blockquote><h3>3. REACT framework — Reasoning + Acting</h3><p>The <strong>REACT framework</strong> stands for <strong>Reasoning and Acting</strong>, and it’s designed to guide the AI through problems in a structured, iterative way. It’s particularly useful for tasks that require:</p><ul><li>Decision-making</li><li>Multi-step problem-solving</li><li>Analysis before action</li></ul><p>The basic structure is:</p><ol><li><strong>Reasoning</strong> — ask the AI to think through the problem step by step, explain what it’s doing, or evaluate options.</li><li><strong>Acting</strong> — then have it take an action or provide the answer based on that reasoning.</li></ol><p>I find REACT especially helpful when working with QA scenarios that involve <strong>diagnostics, root-cause analysis, or choosing between multiple tools or solutions</strong>. It slows the AI down just enough to avoid rushed or surface-level answers.</p><p><strong>REACT framework example: troubleshooting a 500 error on login</strong></p><p><strong>Prompt using REACT:</strong></p><blockquote>You’re troubleshooting a 500 Internal Server Error that occurs during user login on a web application.<br> <strong>First, reason through possible causes of the issue.<br></strong> <strong>Then, act by writing a step-by-step guide to help isolate the root cause.</strong></blockquote><h3>🧩 When to use RTF, RISEN, and REACT in QA work</h3><p><strong>RTF </strong>is ideal for quick, focused outputs. It gives just enough structure to get a clear result without overcomplicating things.<br><strong>Use it when</strong> generating test cases, writing bug reports, creating acceptance criteria, or summarizing documentation.</p><p><strong>RISEN </strong>is best for more complex prompts where you want detailed, structured responses. It offers more control and context.<br><strong>Use it when</strong> comparing tools, outlining test strategies, generating structured reports, or handling layered QA tasks that need constraints or multiple steps.</p><p><strong>REACT </strong>is great when you want the AI to think through a problem before responding. It helps simulate logical decision-making or problem-solving.<br><strong>Use it when</strong> debugging issues, analyzing test failures, reviewing logs, or deciding between multiple options based on reasoning.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c6f3c10f6d8e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What NOT to Do in Your Test Automation Framework — Part 2]]></title>
            <link>https://medium.com/@katyapereverzeva/what-not-to-do-in-your-test-automation-framework-part-2-92dab78068bc?source=rss-d7e6421e2633------2</link>
            <guid isPermaLink="false">https://medium.com/p/92dab78068bc</guid>
            <category><![CDATA[automation-testing]]></category>
            <category><![CDATA[testing]]></category>
            <category><![CDATA[qa]]></category>
            <dc:creator><![CDATA[Katya Pereverzeva]]></dc:creator>
            <pubDate>Fri, 11 Apr 2025 16:41:34 GMT</pubDate>
            <atom:updated>2025-04-11T16:43:20.657Z</atom:updated>
            <content:encoded><![CDATA[<h3>What NOT to Do in Your Test Automation Framework — Part 2</h3><h3>UI-Based test setup</h3><h4>What is it?</h4><p>UI Setup happens when tests prepare their preconditions by clicking through the UI, instead of setting things up through APIs or the backend.</p><p>For example, to test deleting an item, you first create the item through the UI at the start of your test.</p><h4>⚠️ Why it’s bad</h4><ul><li>Slows down your tests</li><li>Introduces unnecessary dependencies</li><li>Higher flakiness</li><li>Harder to debug</li></ul><h4>Example: anti-pattern in action</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YO6pVVM1lIhZJMtTxHVZng.png" /><figcaption>example</figcaption></figure><h4>🚫 What’s wrong here?</h4><ul><li>We’re not testing “create item”, but we rely on it working 100% for our delete test</li><li>If the creation form changes, this test fails unnecessarily.</li><li>That setup time adds up</li></ul><h4>✅ Better approach: use API or backend setup</h4><p>Do the setup behind the scenes — using a direct API call or DB insert — and start the test in the right state.</p><h4>📦 Refactored example with API setup</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YqsmLmEcF-x682QXI92dNw.png" /><figcaption>example</figcaption></figure><p>helpers/apiUtils.ts</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*mANF-Yd1-IIfhUnYZbcl4A.png" /><figcaption>example</figcaption></figure><h4>✅ Benefits of API setup</h4><ul><li>Test runs in seconds, not minutes.</li><li>You isolate and test only what matters.</li><li>Your test failures are more reliable and easier to trace.</li></ul><h3><strong>Anti-pattern: Overloaded functional tests</strong></h3><h4>🧠 What it is:</h4><p>A test that tries to verify too many things at once, instead of focusing on a single behavior.</p><p>This often happens when a test goes from point A to Z: it navigates through multiple flows, verifies side features, and attempts to check UI behavior or data correctness.</p><h4>⚠️ Why it’s bad:</h4><ul><li>Unclear test purpose —<strong> </strong>It’s hard to tell what’s exactly being tested</li><li>Hard to debug — If it fails, you don’t immediately know what went wrong.</li><li>Flaky — More moving parts = more chances for unrelated failures.</li><li>Unscalable — As features grow, the test becomes harder to maintain.</li></ul><h4>Overloaded test in action</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CGuidgctPA5WWylt8tOAuA.png" /></figure><p>This test is doing way too much:</p><ul><li>Searching for the product</li><li>Navigating to product page</li><li>Verifying toast message</li><li>Checking existing items in card</li><li>Validating price calculation</li></ul><p>The actual goal — verifying that a product can be added to the cart — is lost.</p><h4>📦 Refactored Example</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FhBYsQ_4Fm93WRDTmE7j2Q.png" /><figcaption>example</figcaption></figure><h4>✅ Benefits of the Refactor</h4><ul><li>Focused — Only one concern: did the item get added to the cart?</li><li>Simple and reliable</li><li>Easier to debug — One failure = one problem.</li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=92dab78068bc" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What NOT to Do in Your Test Automation Framework - Part 1]]></title>
            <link>https://medium.com/@katyapereverzeva/what-not-to-do-in-your-test-automation-framework-part-1-7d767193688a?source=rss-d7e6421e2633------2</link>
            <guid isPermaLink="false">https://medium.com/p/7d767193688a</guid>
            <category><![CDATA[qa]]></category>
            <category><![CDATA[coding]]></category>
            <category><![CDATA[test-automation]]></category>
            <category><![CDATA[testing]]></category>
            <dc:creator><![CDATA[Katya Pereverzeva]]></dc:creator>
            <pubDate>Tue, 08 Apr 2025 18:04:12 GMT</pubDate>
            <atom:updated>2025-04-08T19:49:31.977Z</atom:updated>
            <content:encoded><![CDATA[<h3><strong>Monolithic Page Objects (God POM)</strong></h3><h4>What is it?</h4><p>The most popular design pattern for UI test automation frameworks is the Page Object Model (POM). In this pattern, each page or component that you interact with in your tests should have its own class, known as a Page Object.</p><p>A God POM is a Page Object class that tries to do everything. Instead of representing a single page or component, it contains logic for the entire page, plus all modals, popups, sidebars, and nested components — all bundled into one giant class.</p><h4>⚠️ Why it’s bad</h4><ul><li>Reduces modularity.</li><li>Hard to reuse or test individual components (like a modal).</li><li>Violates separation of concerns.</li><li>Not scalable for growing test suites.</li></ul><h4>Example: God POM in action</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6CWrYfj3lS3BRiP5VTc2RQ.png" /></figure><h4>🚫 What’s wrong here?</h4><p>Just because the login popup opens from the home page doesn’t mean it belongs inside the HomePage class.</p><p><strong>Issues:</strong></p><ul><li>All popup selectors and logic are hardcoded inside HomePage</li><li>You can’t reuse the LoginPopup in other contexts (e.g., product page, checkout)</li></ul><h4>✅ Better approach: split components</h4><p>Refactor popups and shared UI elements into separate, reusable component classes.</p><h4>🔑 Refactored example</h4><p>pages/HomePage.ts</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*GToxfg8bhEtZNMc5lxTdMg.png" /></figure><p>components/LoginPopup.ts</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fqRQdJHlm00t5tBsWgAnPg.png" /></figure><h4>🧪 Example test using the refactored version</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*2QU5zV8eR8F4_LM91fxOiQ.png" /></figure><h4>✅ Benefits of this refactor</h4><ul><li>Reusable popup component across multiple pages</li><li>Separation of concerns: HomePage handles navigation, LoginPopup handles its own logic</li><li>Easier to maintain</li><li>More scalable for larger projects with shared components</li></ul><h3>Page Objects with Logic</h3><h4>What is it?</h4><p>A Page Object should represent a UI structure — just the elements and interactions. It should not contain logic related to test decisions, user flow branching, or business rules.</p><p>However, a common anti-pattern is including test flow logic or business decision-making directly into Page Object methods. This makes the class responsible for more than it should be, turning a simple UI wrapper into a logic-heavy, tightly coupled component.</p><h4>⚠️ Why It’s Bad</h4><ul><li>Violates Single Responsibility Principle — Page Objects are no longer just UI abstractions.</li><li>Less flexible — Test scenarios become tied to specific logic baked into the Page Object.</li><li>Scattered responsibility — If business logic changes, Page Objects (not just tests) need updates.</li></ul><h4>Example: Anti-Pattern in action</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7Xfp9YVqXjl1GppPVTG2aQ.png" /></figure><h4>🚫 Why this is a problem</h4><ul><li>The Page Object is now responsible for deciding where the user should land.</li><li>You can’t just call login() without triggering redirect logic.</li><li>You’ve created hidden behavior — calling one method results in two responsibilities.</li></ul><h4>✅ Better approach: keep logic in tests</h4><p>Split concerns: Page Objects should only know how to interact with the UI. Let your tests define what should happen after that.</p><h4>🔑 Refactored example</h4><p>LoginPage.ts</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*GOrgM_FivluiHcP41TEIpg.png" /></figure><p>login.spec.ts</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Gjb0LMs4saeCFAgPmqzMEQ.png" /></figure><h4><strong>✅ Benefits of this refactor</strong></h4><ul><li>Modular: Each part of the test framework does only what it should.</li><li>Reusable: You can use login() anywhere — in tests that care about routing or ones that don’t.</li><li>Easier to maintain: Changes in business rules don’t require digging into Page Object classes.</li><li>Clear test intent: Tests describe the expected behavior clearly and explicitly.</li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7d767193688a" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>