-
🛑 Class Recording Will be Here
8 Lessons-
StartCreate Custom AI Skills with GitHub Copilot & Claude Code Part 1
-
StartCreate Custom AI Skills with GitHub Copilot & Claude Code Part 2
-
StartAGENT SKILLS FOR QA ENGINEERS — PART 1
-
StartSKILL MASTERCLASS — BUILD REUSABLE AI SKILLS FOR QA
-
StartBUILD CUSTOM AUTOMATION SKILLS FOR PLAYWRIGHT & SELENIUM
-
StartI Gave My AI Agent 250+ QA Skills — The Test Output Changed Completely
-
StartAI Agent Skills Basics and Install
-
StartFull Notes
-
Skill Masterclass for QA
Build reusable AI Skills that turn Claude into a testing specialist for your whole team.
36
agent skills
7
STLC phases
22
framework skills
46
sections of notes
∞
lifetime access
Build a Copilot skill that turns a Jira story into a PDF + XLSX test plan, then install the full 36-skill QA Skill Suite. MIT licensed. Taught by Pramod Dutta.
Course Summary
A hands-on masterclass on designing, writing, and shipping AI Skills for software testing. Package your test strategy, frameworks, and conventions into reusable Skills any agent can invoke: bug reports, test plans, Page Object generation, API testing, and more. For QA leads and SDETs who want to scale AI consistently across the entire testing workflow.
What you will actually be able to do
Stop prompting. Start shipping skills.
By the end of this masterclass you will have built a working agent skill that turns a Jira story into a review-ready PDF and XLSX test plan — and installed a 36-skill suite that covers every phase of the STLC. Not slides. Files you commit to your repo.
Outcome 01
Ship a real skill
A working
test-plan-from-jira skill: one Jira key in, a polished PDF and a filable XLSX out. Yours, in git, on day one.Outcome 02
Master the format
SKILL.md front-matter, description-based auto-matching, bundled scripts. The portable format any skill-aware agent reads.
Outcome 03
Cover the whole STLC
36 skills: 14 mapped to the seven STLC phases, plus 11 Playwright and 11 Selenium automation skills. MIT licensed.
Outcome 04
Keep the human gate
Every skill drafts and hands back. Nothing files a ticket, pushes code, or signs off on its own. Governance you can defend.
The method
One Jira key in. Two shareable files out.
This is the pipeline you build in the first session. Everything else in the masterclass is a variation on this shape.
The syllabus
Everything we teach, end to end
Twelve modules. Concepts first so nothing feels like magic, then the build, then the suite, then how to roll it out without losing your review gate.
01What an LLM actually is
Next-token prediction, the six limitations of a bare model, and why each one is a defect class you already know how to think about.
02LLM vs AI Agent
Agent = LLM + memory + tools + a stopping loop. Where MCP fits, and why bundled scripts beat MCP as a starting point.
03Prompt → Skill → Agent
The ladder, the comparison table, and the same Jira job done four ways. When to climb and when to stop climbing.
04What an agent skill is
A folder with SKILL.md plus optional scripts. Front-matter, description matching, and the instruction body.
05Writing a description that fires
The single reason skills do not get picked up. Weak vs strong descriptions, side by side.
06Install & setup
Folder locations and scopes, the three Python packages, the three environment variables, macOS / Linux / Windows commands.
07Secrets done properly
API tokens in env vars, the .gitignore you write before line one, and the conversation to have with security early.
08Build: test-plan-from-jira
The full SKILL.md,
fetch_jira.py line by line, ADF flattening, custom field ids, and the plan.json contract.09Export to PDF + XLSX
reportlab and openpyxl, wrap-text, repeating headers, and the extensions worth building (branding, traceability matrix).
10The 36-skill STLC suite
All 14 STLC skills phase by phase, plus the 11-skill Playwright pack and the 11-skill Selenium pack. Which five to install first.
11n8n & LangFlow builds
The same agent as a four-node n8n flow and a five-component LangFlow graph. Install commands for both, and how to choose.
12Governance & rollout
The human review gate, the never-invent rule, the security checklist, metrics that survive scrutiny, and the anti-patterns that kill adoption.
Plus the written reference. Two lessons in the curriculum are notes, not video: AI Agent Skills Basics and Install (the build, with diagrams) and Full Notes (a 46-section searchable reference covering every command, script, skill, and rule). You keep both.
Coverage
Every phase of the STLC, covered
The suite folders are numbered in testing order, so the skills hand off the way real work flows — and closure feeds the next cycle.
| Group | Count | What is inside |
|---|---|---|
| STLC skills | 14 | Requirement analyzer, test plan generator, scenario + API designers, case writer, data generator, script generator, regression selector, execution tracker, bug reporter, triage assistant, RCA analyzer, closure reporter, coverage analyzer |
| Playwright pack | 11 | Test generator, POM builder, locator fixer, fixture designer, API tester, network mocker, flaky debugger, trace analyzer, visual regression, a11y auditor, CI configurator |
| Selenium pack | 11 | Test generator, framework scaffolder, POM builder, locator strategist, wait fixer, flaky debugger, driver manager, data-driven designer, cross-browser runner, grid configurator, report integrator |
Start here
The five skills we install first
Thirty-six folders is intimidating, so we do not install all of them. These five cover a full ticket-to-closure loop and give the biggest return in week one.
1jira-requirement-analyzer
The only skill that saves you work by telling you not to work. Judges whether a story is actually ready to test before you plan around it.
2test-plan-generator
The one you build in session one, packaged. Scope, out-of-scope, scenarios, and cases typed positive / negative / edge / boundary.
3test-case-writer
Scenarios are cheap; detailed cases are expensive. This is where your hours actually go, and this is the skill that takes them back.
4bug-reporter
The highest-frequency writing task in QA and the one most often done badly at sprint end. Consistent, reproducible, flags what it does not know.
5pw-locator-fixer or se-wait-fixer
Pick by stack. Both attack the number-one cause of “works locally, fails in CI”.
Why these and not the flashy ones. They map to the four things you do every single sprint — read a ticket, plan it, write cases, file a bug — plus the one thing that quietly eats your afternoons.
The format
A skill is just three files
No installer, no platform, no lock-in. Once you can picture this, you can build a skill for any QA job you do more than twice.
The shift
Before and after
| The task | Before | After this masterclass |
|---|---|---|
| Reading a new story | You start planning and discover halfway through that the AC are untestable | jira-requirement-analyzer gives you a testability verdict and the questions to ask the author |
| Writing a test plan | Half a day, and the negative cases get thin by 5 p.m. | Minutes to draft, then you spend your time reviewing instead of typing |
| Detailed test cases | The expensive, unglamorous hours | Drafted against every AC, with a rule that forces a negative case per criterion |
| Test data | You forget the empty string. Again. | Valid, invalid, boundary, unicode, and edge dates — asked for by name |
| A flaky test | Rerun it and hope | Flaky scored apart from broken, with a real formula and an HTML report |
| Filing a bug | Rushed at sprint end; developer replies “cannot reproduce” | A consistent, reproducible format every time — bounce-backs drop |
| Root cause | “Fixed it” | Structured RCA that also asks why your tests did not catch it |
| Closure report | Written the night before the review | Drafted from your execution log, with the coverage gap list attached |
| Sharing your standards | A wiki page nobody reads | A committed skill the newest joiner inherits on clone |
The honest framing. This does not remove QA work — it moves it. You spend far less time writing and far more time reviewing, judging risk, and deciding what not to test. That is the part of the job that was always the valuable part.
Fit check
Who this is for — and who it is not
This is for you if…
- You are a manual QA engineer and you want the writing work off your plate without giving up judgement. Fourteen of the 36 skills need no code at all.
- You are an automation engineer or SDET and locator repair, wait fixing, and flaky diagnosis are eating your week.
- You are a QA lead who needs a rollout that keeps a review gate and survives a security conversation.
- You have tried prompting an AI for test cases, got something mediocre, and suspected the problem was the approach rather than the model.
- You want files you own in your own repo, not another SaaS subscription.
This is not for you if…
- You want AI to sign off on releases. Every skill here stops at a human review gate on purpose, and we are not going to help you remove it.
- You are looking for a no-terminal, click-only tool. You will run a handful of commands.
- You want a course on prompt tricks. This is about durable, versioned instruction files.
- You need a fully unattended pipeline today. We cover the n8n route, but the focus is developer-in-the-loop work.
- You are not willing to review generated output. Then this will make you faster at being wrong.
| Prerequisite | Level needed | Notes |
|---|---|---|
| A terminal | Comfortable, not expert | You will run five or six commands |
| Python 3.9+ | Can run a script | Only the two bundled scripts use it; you never write Python from scratch |
| Jira access | Read a ticket, create an API token | The whole worked example is Jira-driven |
| GitHub Copilot or Claude Code | Installed and signed in | Either works — the skill format is portable |
| Playwright or Selenium | Optional | Only needed for the framework packs |
| A paid API key | Not needed | Your existing Copilot or Claude subscription is enough |
Your instructor
Taught by Pramod Dutta
Founder of The Testing Academy, a YouTube channel with 95K+ subscribers where Pramod teaches software testing and test automation. This masterclass is the same material he uses with live cohorts — built from real QA work, not from a documentation page.
Everything in this course is shown end to end: the failing command, the fix, the skill that did not match and the description edit that made it match. You watch the debugging, not just the finished demo.
Questions
Frequently asked
Do I need to know how to code?
No, for most of it. Fourteen of the 36 skills are instruction files only — plain English, no code. The worked build uses two Python scripts that are given to you complete; you copy them and run them. If you can run a command in a terminal you can finish this masterclass.
Does this work with GitHub Copilot, or only Claude Code?
Both. A skill is a folder with a SKILL.md, and the format is portable — that is the whole point. We show the Copilot path and the Claude Code path, plus how to convert a suite skill into a Copilot
.prompt.md prompt file.Do I need a paid API key?
No. Your existing GitHub Copilot or Claude subscription is enough. There is no separate model bill, no cloud account, and nothing to provision. Everything runs locally against your own Jira.
Is the 36-skill suite really free?
Yes — MIT licensed, one download. The masterclass is where you learn how the format works, how to write a description that actually fires, how to debug the Jira call, and how to govern the output. The files are free; knowing how to build and fix your own is the part you are here for.
Will this write my tests for me and let me stop reviewing?
No, and we deliberately do not help you build that. Every skill in the suite stops at a human review gate: it drafts, states its assumptions, and hands back. If your goal is unreviewed AI sign-off, this is the wrong course.
My Jira is self-hosted / Server, not Cloud. Does it work?
The concepts and the whole suite work unchanged. The
fetch_jira.py script targets the Jira Cloud REST v3 API, so on Server or Data Center you adjust the endpoint and auth. We cover how to find your own field ids, which is the part that actually differs.Is my ticket data safe?
That is your decision to make deliberately, and we show you exactly what to check. The fetch script requests six fields, not the whole issue, precisely so you know what leaves your machine. Section 07 covers env-var secrets, the .gitignore you write first, and the conversation to have with your security team.
What about Playwright vs Selenium?
Both are covered with a dedicated 11-skill pack each. Playwright pack is TypeScript; Selenium pack is Java + TestNG. Pick the one your team actually runs — you do not need both.
How long will this take me?
One afternoon gets you the working skill and the five-skill install, which is most of the practical value. The course includes a 4-week path if you want to take a full story through the whole STLC and then write a skill of your own.
Do I get written notes or just video?
Both. Two lessons in the curriculum are written references: AI Agent Skills Basics and Install (the build, with diagrams) and Full Notes — a 46-section reference with every command, the complete SKILL.md, both scripts explained line by line, all 36 skills, troubleshooting tables, a glossary, and self-check questions.
Is there a refund?
Yes — the 30-day money-back guarantee shown on this page applies. If you work through it and it does not help, ask and you get your money back.
Skill Masterclass for QA
Build one skill this week. Install thirty-six next week.
You leave with a working
test-plan-from-jira skill in your repo, the full 36-skill STLC suite installed, a 46-section written reference, and a governance model you can defend in a review. Lifetime access and a 30-day money-back guarantee.✓ 36 agent skills, MIT licensed✓ Full STLC coverage✓ Playwright + Selenium packs✓ 46-section written notes✓ Lifetime access✓ 30-day money back
Enroll in Course →Instant access · Start with the notes while the recordings download
Course Curriculum
Pramod Dutta
Founder of The Testing Academy, a YouTube Channel with 95K subscribers where Pramod teaches about Software testing & Test Automation. With overall 10+ years of experience in Software Testing & Test Automation, he has mentored 10,000+ students in Software Testing, API Testing, Test Automation.
Pramod Dutta is working as SDET Manager at Tekion | Ex BrowserStack Employee & BrowserStack Champion & Certified Scrum Master. Pramod has a vast range of experience handling from Manual Testing, Mobile & Web Automation, Desktop & cloud services like AWS, GCP extra.
By joining this Automation Testing Course , you’ll have the opportunity to take control of your life, work in an exciting industry with infinite possibilities and live the life you want.
