Course outcomes are one of those things everyone agrees matter and almost nobody likes writing. Done well, they drive everything else — what you teach, how you assess, what students actually walk away with. Done badly, they're vague filler that ends up in a syllabus nobody reads.
I've helped a lot of people write them. And I've had a version of this prompt living in various notebooks and documents for a couple of years. It uses Bloom's taxonomy to pick the right action verbs for the course level, avoids the usual offenders ("understand," "know," "be familiar with" — none of those are measurable), and pushes for outcomes specific to the course rather than generic ones that could apply to anything.
I finally built it properly as a Claude Code skill and a TempLight template so I'm not rewriting it every time someone asks.
What it does
Ask it for course outcomes. It'll ask you for:
- Type of institution (community college, university, trade school, etc.)
- Course level (developmental through graduate)
- Course name
- Course description — a rough draft is fine
- Any additional context (optional)
From there it generates 3–5 outcomes with example assessment types for each. It'll also note if it had to make inferences because the description was thin.
It works for any discipline. I'm in higher ed so that's where my head is, but the structure holds for K-12, corporate training, or anywhere else outcomes need to be written. After the first pass, ask it to revise individual outcomes, shift Bloom's levels up or down, or rewrite for a different audience — student-facing language reads differently than what goes in an accreditation report.
Two ways to use it
If you use Claude Code, download the skill file and drop it into ~/.claude/skills/. After that, just ask Claude Code to help you write course outcomes and it'll handle the rest.
If you use Claude.ai, ChatGPT, or any other AI tool, use the TempLight version instead. It's a fill-in-the-blanks prompt — drop in your course details and paste it into whatever you're using.
Open TempLight PromptBoth produce the same output. The skill just integrates tighter into the Claude Code workflow.