A good skill has three qualities: a clear trigger, a defined output, and a repeatable process. If you can describe exactly when the skill runs, what it produces, and the steps it follows every time, it’s a good candidate. If the task requires improvisation, judgement calls that change each time, or outcomes you can’t describe in advance, it’s too vague for a skill and should stay a human task.
The Three-Part Test
Before building any skill, run it through three questions. First: when does this run? A good skill has a clear trigger — “every Monday morning,” “whenever a new student enrolls,” “when I need a lesson plan for a new topic.” If the trigger is vague — “whenever I feel like it” or “sometimes when things are slow” — the task isn’t well-defined enough for a skill.
Second: what does it produce? A good skill creates a specific, predictable output — a 200-word community post, a structured lesson plan with five sections, an email draft with subject line and body. If the output varies wildly each time or you can’t describe what “good” looks like in advance, the task needs human judgement rather than a skill.
Third: can you write the steps? A good skill follows the same process every time. “Read the topic, check the audience profile, write an opening hook, develop three teaching points, add a practice activity, close with a takeaway.” If the steps change depending on the situation, or you’d handle it differently each time, it’s not ready for automation.
Good Skill vs. Bad Skill Examples
Good skills: “Create a community discussion post about [topic]” — clear trigger, defined output, repeatable process. “Draft a welcome email for [student name] who enrolled in [course]” — specific input, predictable format, consistent steps. “Generate five quiz questions from [lesson content]” — well-scoped, clear deliverable, mechanical process.
Bad skill candidates: “Decide whether to launch a new course” — requires strategic judgement and context that changes. “Comfort a struggling student” — requires empathy and reading emotional cues. “Figure out what’s wrong with my marketing” — too open-ended, no clear output format. These are human tasks that benefit from AI assistance but shouldn’t be fully delegated.
What This Means for Educators
As a coach or course creator, your best skill candidates are the tasks that feel mechanical even though they require intelligence. Writing weekly community posts, creating lesson outlines, drafting student communications — these all follow patterns you could teach to an assistant. If you could write a standard operating procedure for the task, you can build a skill for it.
The Bottom Line
Good skills are specific, predictable, and repeatable. When in doubt, ask yourself: “Could I train a competent assistant to do this task by giving them a one-page instruction sheet?” If yes, it’s a great skill candidate. If no, keep it as a human task where AI assists rather than leads.