What does a QA engineer's workday with Claude look like?

Morning — a sprint test plan from Jira tasks, prioritized by business risk. During the day — test cases from a PR description, JSON with 20 test products, a bug report in 30 seconds from a raw description. Before the release — a prioritized 3-hour regression checklist; in the evening — a 5-7 sentence report for the lead in 2 minutes.

What should go into CLAUDE.md for a QA project?

A CLAUDE.md file in the repo root should hold: the stack, environment URLs (dev, staging, prod), test accounts, business rules (minimum order 500 RUB, maximum 10 items in the cart), critical modules, and known limitations. Claude Code reads it in every session and knows the project context right away.

How do I create my own skill for Claude Code?

Create a file at .claude/skills/name/SKILL.md with frontmatter (name, description, allowed-tools) and the prompt text. For example, /qa-testcases generates test cases as a table and /qa-bugreport formats a bug from a free-form description. Create it once — use it every day instead of a long prompt.

How do I integrate Claude with Jira, Postman, and TestRail?

Through output formats: from swagger documentation Claude generates a JSON collection for Postman import with a pre-request script and tests; bug reports and acceptance criteria come out in Jira markdown; test cases go to CSV with Title, Preconditions, Steps, and Expected Result fields for TestRail import; bug summaries to CSV for Excel.

How much time does Claude save a QA engineer?

Test cases in minutes instead of hours, a bug report in 30 seconds instead of 10 minutes, a manager report in 2 minutes. Plus analyzing project code without programming knowledge and bug reports with the file and line number. Start with one command a day — within a week it becomes part of your workflow.

Claude for QA engineers: the full workflow

Contents

A full day with Claude

Let's go through a typical QA engineer day and show where Claude saves time. This part ties together everything from the earlier parts of the series — from first prompts to working with the repository.

09:00 — Morning: planning

You open Jira, see sprint tasks. Copy them into Claude:

Sprint 15 tasks:
1. SHOP-234: Add filter by clothing size
2. SHOP-235: Fix double-charge bug on card payment
3. SHOP-236: New "Returns" page
4. SHOP-237: Catalog loading optimization

Build a week's test plan.
Prioritize by business risk.
For each task: what to test, how long, which testing types.

A minute later you have a plan: payment bug first (critical), then new page, then filter, optimization at the end.

10:00 — Got a PR: analysis

The developer closed SHOP-234 (size filter). If you have repo access:

Look at the latest commit. What changed?
Which features are affected? What might break?
Generate test cases based on the changes.

If no repo access — use the PR description:

PR description:
- Added size filter (XS, S, M, L, XL, XXL)
- Filter combines with price filter
- When no items of selected size — message "Nothing found"
- Sizes are fetched from API

Write test cases. Include: functional, UI, edge, negative.

11:30 — Testing: need data

You need test products with various sizes:

Generate JSON with 20 products for testing the size filter.
Fields: id, name, price, sizes (array), category, in_stock.

Include:
- Products with all sizes
- Product with one size
- Product with no sizes (empty array)
- Product with price 0
- Out-of-stock product

14:00 — Found a bug: format it

While testing you found: combining size + price filter, if you select size first, then price, then remove size — the list is empty.

Format a bug report for Jira:

Product filter. Select size M → products shown.
Add price filter (100-500) → products filtered.
Remove size filter → list empty, although products in 100-500
without size filter should remain.
Chrome 125, staging.

Copied the result into Jira — bug filed in 30 seconds.

16:00 — Regression

Before release, regression must run:

Release 2.15 contents:
- Size filter (new feature)
- Double-charge fix
- New returns page
- Catalog optimization

System modules: catalog, cart, payment, profile, orders.

Create a prioritized regression checklist.
Total time: 3 hours.

17:30 — Report

Write a testing report for the lead. Short, 5-7 sentences.

Data:
- Tested: SHOP-234, SHOP-235, SHOP-236, SHOP-237
- Test cases run: 67
- Bugs found: 4 (1 major, 3 minor)
- Major: filter combination breaks the list (SHOP-234)
- No blockers
- Regression: clean
- Recommendation: ship after major bug fix

Done. The whole day — with Claude.

CLAUDE.md for a QA project

If you use Claude Code with a repo (part 4), create a CLAUDE.md file in the project root. Claude reads it every time and gets the context.

QA-flavored example:

# E-Shop — clothing online store

## Stack
- Backend: .NET 8 Web API
- Frontend: Angular 19
- Database: PostgreSQL 16
- Payments: Stripe
- Hosting: Azure

## Environments
- Dev: https://dev.eshop.example.com
- Staging: https://staging.eshop.example.com
- Prod: https://eshop.example.com

## Test accounts
- Customer: [email protected] / BuyerTest123!
- Admin: [email protected] / AdminTest123!
- Manager: [email protected] / ManagerTest123!

## Business rules
- Minimum order: 500
- Free shipping above 3000
- Max 10 items in cart
- 14-day return window
- Payment: card (Stripe), cash on delivery

## Critical modules (test thoroughly)
- Order placement + payment
- Registration / login
- Cart (totals, discounts)
- Catalog (filters, search)

## Known limitations
- Safari < 16: incorrect date display
- IE not supported
- Photo upload: max 5 MB, jpg/png

Now every time you ask Claude Code about the project, it already knows the context.

Custom skills

Skills are saved commands for Claude Code. You type /qa-testcases instead of a long prompt. How skills work under the hood is covered in a separate article.

How to create one

In the project folder create:

.claude/skills/qa-testcases/SKILL.md

File content:

---
name: qa-testcases
description: Generate test cases from requirements or code
allowed-tools: Read, Grep, Glob
---

Generate test cases based on the given description
or file.

Include: positive, negative, edge, security.

Format: table with columns:
ID | Type | Title | Steps | Expected result | Priority

Now type /qa-testcases payment form — and Claude generates test cases.

Useful skills for QA

/qa-bugreport — formats a bug from a description:

---
name: qa-bugreport
description: Bug report from a free-form description
allowed-tools: Read, Grep, Glob
---

Format a bug report from the description.
Format: title, severity, steps, expected/actual,
environment, workaround.

/qa-checklist — testing checklist:

---
name: qa-checklist
description: Testing checklist for a feature
allowed-tools: Read, Grep, Glob
---

Create a testing checklist.
Categories: functionality, validation, UI/UX, security,
performance, compatibility.
Format: checkboxes.

Make them once — use every day.

Tool integration

Claude + Postman

Here's the swagger API documentation (or paste JSON):

Generate a Postman collection:
- For each endpoint: valid request + 3 invalid
- Pre-request script for auth
- Tests to verify responses (status code, required fields)

Format: JSON for Postman import.

Claude + Jira

Here are testing results (list of test cases with outcomes).

Create bug reports for each failed test case.
Format: Jira markdown.
Include: title, severity, steps, expected/actual.

Here's a feature description. Write acceptance criteria for a Jira ticket.
Format: list of checkboxes (Jira checkbox format).

Claude + TestRail

Here are test cases (table).

Reformat for TestRail import:
- Section: [section name]
- Title, Preconditions, Steps, Expected Result
- Priority: 1-Critical, 2-High, 3-Medium, 4-Low
- Type: Functional / Regression / Smoke

Format: CSV for import.

Claude + Excel/CSV

Here's bug-tracker data for 3 months (paste or describe).

Create a summary:
- Bug count by severity (table)
- Count by module
- Average fix time
- Weekly trend

Format: CSV so I can paste into Excel and chart.

What you've achieved

If you went through all 5 parts of the series, you can now:

Generate test cases in minutes, not hours
Format bug reports in 30 seconds instead of 10 minutes
Write test plans, regressions, and checklists — not from scratch
Analyze project code without knowing how to code
Find bugs others miss — because you see the code
Write bug reports with file and line numbers — developers love it
Prep manager reports in 2 minutes

You didn't become a programmer. You became a QA who understands the system from inside and uses AI as a tool.

Series complete. You went from your first browser prompt to the full Claude Code workflow. Start small — one command per day. A week later you won't be able to work without it.