AI Skills Workshop

A hands-on workshop for building reusable AI-powered development and testing workflows.

Learning objectives

01

Understand how agent skills turn repeated instructions into reusable, predictable workflows.

02

Use agentic quality checks to validate a feature against its acceptance criteria.

03

Create, refine, and integrate a reusable skill for writing E2E tests.

Hands-on workshop

Working with agent skills.

Instead of prompting and re-solving the same problems, agent skills package instructions into reusable workflows, and you can build quality checks right into them. This improves the consistency and predictability of an agent’s output. It is especially powerful for teams doing agentic development, but these quality-checking skills work just as well on their own.

  1. Create a feature user story

    Open the scrum board and create a user story for a feature you would like to add. Include clear acceptance criteria so the agent can determine when the feature is complete and validate the expected behavior. Get creative! You can use an already defined feature, but it’s more fun to come up with something new. 😊

    Open scrum board

  2. Develop and validate the feature

    Start the develop-feature skill to take your feature through the full development workflow, from implementation to agentic checks. As one of its steps, the workflow uses the test-user-story skill to validate the completed feature against the user story.

    Review and iterate

    Inspect the generated report and supporting evidence. If the executed cases do not give you enough confidence, ask the agent to run additional cases or target specific scenarios. Continue until the evidence demonstrates that the user story works as intended.

  3. Ask the agent for an E2E test

    Ask your coding agent to write an E2E test for the feature. Do not give it additional workflow instructions yet.

    Write an E2E test for the feature.
    Reflect

    What did you like about the test?

    What would you change?

    Write down your observations before continuing.

  4. Create a write-e2e-test skill

    Create a .agents/skills/write-e2e-test directory with a SKILL.md file. In the skill metadata, set the name to write-e2e-test so it exactly matches the skill’s directory name.

    SKILL.md Example metadata 
    ---
name: write-e2e-test
description: Creates and validates an E2E test for a completed feature.
---

    Define the workflow the agent should follow when implementing an E2E test. The questions below can help shape the workflow around your testing preferences.

    • Do you want to use a Page Object Model?
    • Should the agent write immediately, or propose an approach and ask questions first?
    • How will the agent verify that the generated test works?
    • Should it use Gherkin, explicit test steps, or another format?
    • How should the test’s initial state and data be prepared?
    • How should the skill report failures?

  5. Run, refine, and integrate

    Use your new skill on its own to write the E2E test. Did the result match what you had in mind? Adjust the skill and run it again until the workflow fits your use case. Once you are satisfied, add write-e2e-test as the final step of the develop-feature skill.

    Goal

    Finish with a repeatable skill that produces a working E2E test in the style and structure you expect.