Author Your First Skill

The learning problem

A professional can understand what an agent skill is and still struggle to make one useful. They need to choose an appropriate workflow, describe when the skill should activate, and determine whether its output meets their needs.

I designed this experience around producing and testing a working skill. The lesson connects unfamiliar technical concepts to a task the learner already recognizes from their job.

My role and constraints

I designed and facilitated the workshop and developed the interactive web lesson, including role-based examples, authoring decisions, practice activities, and explanatory feedback.

The audience brought different roles and levels of technical experience. The lesson needed a common sequence while allowing each person to choose a relevant outcome. The course version also records progress, so the public lesson asks for an email before starting.

Design decisions

Start with the learner’s work

Learners choose a role and a recurring workflow before drafting a skill. This gives the technical decisions a purpose: what job should the skill do, what should it produce, and when should an agent use it?

Sequence decisions before construction

The build and test steps unlock after learners define the skill’s name, output, and activation conditions. The sequence asks them to establish a usable brief before generating instructions.

Practice the distinctions that cause errors

An anatomy exercise asks learners to locate parts of a skill file. A sorting activity then asks them to decide what belongs in metadata, instructions, or supporting resources. Incorrect placements return a hint, giving learners a reason to reconsider their choice.

Make testing part of the learning task

Learners test how their skill responds to real requests and refine it from the result. Later activities ask them to diagnose failures, connecting a symptom to the part of the skill that needs revision.

Inside the experience

The live lesson contains the complete workflow and interactive practice. The repository provides the implementation, including activity logic and feedback.

Evaluation and learning

I facilitated the workshop with professionals from edtech, military education, and community colleges. The intended performance is a reusable agent skill that participants can explain, test, and revise for their own workflow.

The lesson makes learners’ decisions visible through authoring choices, practice responses, and test results. These provide opportunities to identify misunderstandings and give targeted feedback; completing an activity alone does not establish that someone can transfer the skill to a new task.