AI Skills and Agents in QA Automation

Why take this course

Most QA teams that have “implemented AI” have actually only provided access to AI chat tools. Individual testers interact with AI independently, producing one-off results without repeatability, quality control, or governance. While this approach can improve individual productivity, it does not scale across teams and organizations. [sii.pl]

The next step is moving from conversations to systems:

• Skills – specialized, tested instructions that deliver predictable results for specific tasks.
• AI Agents – configured AI instances operating within defined rules, boundaries, and control mechanisms.
• Workflows – combinations of skills and agents integrated into processes that can be implemented across teams and measured for effectiveness.

This course teaches participants how to design and implement these mechanisms in a QA Automation environment while addressing security, governance, and realistic limitations.

What you’ll learn

After completing this course, participants will be able to:

• Explain what AI skills and AI agents are in the QA context and understand when they are appropriate to use.
• Design a QA skill by defining objectives, inputs, outputs, constraints, and quality criteria.
• Build working skills for bug analysis, test generation, test code reviews, and requirements validation.
• Create AGENTS.md files and system instructions that define AI agent behavior in testing projects.
• Design QA workflows that combine multiple skills into a cohesive process.
• Define agent responsibility boundaries, distinguishing between autonomous activities and tasks requiring human validation.
• Implement governance mechanisms, including quality control, auditing, versioning, and feedback loops.
• Assess team and organizational readiness for introducing AI agents into QA processes.

Certification & Exam

Participants receive a personalized certificate of course completion.

The course does not include an external certification exam.

Optionally, participants may prepare a final project (a designed skill together with an AGENTS.md file) for peer review with the trainer after the course.

Who is this course for

This course is intended for experienced QA professionals and test architects who want to move beyond ad hoc AI usage and start building scalable AI-powered systems:

• Senior Automation QA Engineers / SDETs with extensive automation experience and an interest in designing AI-driven solutions
• Test Architects responsible for testing strategies and tools across the organization
• QA Tech Leads leading technical teams and looking for ways to scale AI adoption
• QA Managers with a technical background who make decisions regarding AI implementation in QA processes

This course is designed for professionals who already use AI when working with code and want to transform individual practices into structured, repeatable systems.

Experience in test automation and confidence working with source code are required.

Recommended prerequisites

• 2–3 years of experience in test automation
• Confidence working with code (Java, TypeScript, Python, depending on the technology stack)
• Practical experience using AI tools in software development contexts (e.g., ChatGPT, Claude, Copilot, Claude Code); occasional chatbot use is not sufficient
• Familiarity with concepts such as prompts, language models, context, and system instructions
• Recommended but not required: completion of the “AI for Automation Testers” course or equivalent practical experience

Topics covered

Day 1: Skills – From Prompt to Repeatable Mechanism

Objective: Understand the concept of AI skills and the difference between prompts and skills. Learn how to design, build, and test QA skills and create working solutions for key testing activities. [sii.pl]

• From Prompt to Skill

What a skill is: definition, structure, and differences between a skill, a prompt, and a template

When a skill is appropriate and when a simple prompt is sufficient

Key components of a skill: Objective, Role, Context, Output format, Constraints, Examples, Quality criteria

Skill lifecycle: Draft -> Testing -> Iteration -> Production -> Maintenance

Demonstration: ad hoc prompt versus structured skill

• Designing QA Skills

Skill design methodology: Task identification -> Input/output analysis -> Constraint definition -> Prototype creation -> Testing

Skills for common QA activities:

Bug analysis (stack trace → root cause + fix proposal)

Test generation (user story → test cases including edge cases)

Test code review (code → issues and recommendations)

Workshop: Design a skill for one selected QA task

Create documentation covering objectives, inputs, outputs, limitations, and usage examples

• Building and Testing Skills

Implementing skills through: Custom instructions, System prompts, Variable-based templates

Skill testing techniques across different input scenarios

Iterative improvement based on results

Skill versioning and change tracking

Workshop: Implement the designed skill

Test it against 3–5 scenarios, including edge cases

Document findings and improve the solution

• Skill Libraries and Knowledge Sharing

Team skill library structure

Metadata and versioning strategy

Skill card format: Purpose, Owner, Inputs and outputs, Known limitations, Change history

Review, approval, and publication process

Tool integrations: Confluence, Git repositories, Slack snippets

Workshop: Finalize skill design, Create a skill card, Present the solution to the group

Conduct peer reviews using defined quality criteria

Day 2: AI Agents – Workflows, AGENTS.md, and Governance

Objective: Learn the concept of AI agents, design AI-driven QA workflows, create AGENTS.md files and system instructions, define responsibility boundaries, and implement governance mechanisms.

AI Agents in the QA Context

Agent definition: Skill + Context + Rules + Boundaries

Differences between skills and agents

Types of agents:

Fully autonomous

Semi-autonomous

Human-in-the-loop

QA agent examples:

Bug triage agent

Code review agent

Test result analysis agent

Realistic expectations regarding agent capabilities

Demonstration: Agent operating in a terminal environment using AGENTS.md

Bug analysis and fix recommendation

• AGENTS.md and System Instructions

Purpose and role of AGENTS.md

Recommended structure: Objective, Role, Available tools, Constraints, Project conventions

Custom instructions vs. system prompts vs. AGENTS.md

Best practices for creating reliable instructions

Common anti-patterns:

Instructions that are too generic

Contradictory guidance

Excessive restrictions

Workshop: Create AGENTS.md for a training repository

Test whether agent behavior aligns with instructions

• Designing Agent-Based Workflows

Workflow as a combination of skills and agents

Example QA workflows:
Bug triage → root cause analysis → fix proposal → regression test creation

New user story → requirements analysis → test case generation → review → test data preparation

Responsibility boundaries: What agents can do independently, What requires human approval

Defining control points and human checkpoints

Workshop: Design a workflow for a real project context

Present and review solutions with peers

• Governance, Security, and Quality Control

AI governance in QA: Ownership, Configuration responsibilities, Monitoring responsibilities

Measuring effectiveness: Precision, Recall, False-positive rate,

Security considerations: Access to code, Data access, Environment permissions

Creating an AI policy before deployment

Auditing and traceability of agent actions

Final workshop: Create an “Agent Card” defining: Purpose, Boundaries, Control mechanisms, Success criteria, Risks

Present and discuss the solution

• Implementation Planning

Group reflection on what is ready for deployment and what requires further development

Roadmap for introducing skills and agents into a team step by step

Resources and communities for continued learning and development.