Here's my problem: how do I know if a skill or prompt is any good (e.g. GStack's /office-hours)?
How do I compare similar skills (e.g. different "deep research" skills)?
Spotting broken software is (relatively) easy — it crashes, prints errors. Broken skills don't. Perfectly polished, confident-sounding skills routinely mislead me and waste my time, to the point where I wish I weren't using an LLM at all.
AI skills are software — and they should come with regression tests.
LLM teams have tons of prompt regression tests. LLM-wrapper SaaS companies have tons of prompt regression tests. But when it comes to open-source skills, SKILL.md reads reasonable, yet ships with zero tests (e.g. GStack's /office-hours has none at the time of writing).
Garry Tan, if you hear me — please consider shipping regression tests for your /office-hours, /plan-ceo-review, /plan-eng-review, and so on.
Regression tests should:
1. Prove the skill works correctly
2. Demonstrate correct and incorrect usage
3. Prove the skill's value
4. Come with a scoring rubric to allow skill benchmarking
5. The last one is the most valuable, because it lets you benchmark similar skills against each other.
So I started doing this myself.
Here's a work-in-progress example: plan-cmo-review, a skill to complement GStack since GStack is missing a marketing review at the time of writing. I'm not a marketing guy; the point of sharing this skill is to outline its regression setup.
Briefly, here's how my exploration progressed:
- I used GStack on a couple of products and realized the resulting design_document.md was leading me to failure, mainly marketing-wise.
- I dug into the skill's failures manually with Claude Opus 4.8's help and ended up finding the correct solution.
- I asked Claude to build a plan-cmo-review skill, ran it, and it arrived at a flawed solution (similar to GStack's output).
- I gave Claude the correct (manual) solution to analyze and add as a regression fixture with a scoring rubric.
- Claude ran the (blind) regression — it failed. We iterated several times and found the key problem: Claude was trusting my prompts implicitly as the ultimate truth. Claude believed GStack knew what it was doing. GStack believed I knew what I was doing. But I was doing product/startup research — and by definition, "research" is what you do when you don't know what you're doing. That trust chain is what broke the skills.
- We fixed the trust problem and the regression test passed. We added a few more. They passed.
- I had Claude run the regressions multiple times — cracks appeared. Claude iterated the skill. Now they pass.
- This methodology is still flawed. I'd like to try running different LLMs, cross-model judging, and a lot more regression tests.
Skill github.com/remakeai/plan-cmo-review . Notes at iliaov.substack.com .
1 comments