You already wrote the skill. Let's make it a runbook.

A skill tells an AI how to do something. A runbook adds standards for “done” and a way to check the work. Jetty turns a skill you've already built into a runbook your team can run reliably — in about two minutes.

If you're here, you've probably already done this

You wrote a skill. Maybe in Claude, as a custom SKILL.md. Maybe as a custom GPT. Maybe as a prompt you keep in a Notion doc and paste in every time.

It works for you. You know when to override it. You know what a good output looks like. You know the weird edge cases where it falls over — and you correct them before they become a problem.

The trouble starts when you share it. Someone else runs it. They get back something plausible-looking but wrong. They don't notice, because they don't have the instinct you do. The output ships. You find out later.

You're not alone in this. Four marketers profiled in the MKT1 newsletter — Emily Kramer, Elaine Zelby, Aditya Vempaty, Kamil Rextin — are all running marketing operations off a folder of skills. The ones that work consistently across a team have the same thing in common: a standard for what “done” looks like, and a check.

That's the runbook.

The equation

Skills + standards = runbooks.

A skill is a set of instructions.
A standard is what “good” looks like for this specific work.
A runbook is both, packaged together, with a self-check loop.

The gap between “the AI did something” and “the AI did something I'd ship” is the standard. Most skills are missing it. Most runbooks have it.

How Jetty turns a skill into a runbook

1. Bring your skill.

Paste your SKILL.md. Upload the folder. Share the Notion doc. We don't care what format your skill is in — if it's a set of instructions you've written for an AI, we can work with it.

2. We ask you three questions.

What does good output look like? What's the finished work — file, document, report, response? What would make you proud to ship it?
What would you reject? If the AI handed you something you wouldn't show a client or teammate, what would be wrong with it? Name two or three failure modes.
What would you double-check before trusting it? If a coworker did this task and you only had thirty seconds to spot-check, what would you look at?

Your answers become the standards and the check section of the runbook. We write the first draft. You edit until it reads like something you'd hand a new hire.

3. Run it.

Drop a real input in. Let the runbook work. See what comes back.

The first run usually surfaces one thing you forgot to write down. You add it. The second run is closer. By the third or fourth, the runbook is good enough to hand off.

This is the pattern. You don't get a great runbook by writing it perfectly. You get there by running it and updating it.

What it looks like, specifically

Let's say your skill today is something like:

Brand voice reviewer. You are a brand voice editor. You review marketing content against the attached voice guide. Flag anything that breaks the rules. Suggest replacements. Return a marked-up version.

It works for you. When your marketing manager runs it on a blog draft, she gets results that seem fine — but she doesn't know that you'd reject anything over a certain length, or that you'd flag any sentence starting with “We believe.” So those sneak through.

A runbook version adds:

Standards for done.
Every paragraph has been reviewed.
Banned words from the voice guide are flagged with a suggested replacement.
Any sentence starting with “We believe,” “We think,” or “We feel” is flagged.
The marked-up version is saved as a .md file.
A one-paragraph summary is produced, calling out the top three issues by frequency.
Check before done.
Verify the marked-up file exists and isn't empty.
Verify the summary mentions exactly three issues.
Re-read the output and confirm every banned word has a replacement.
If any check fails, fix and recheck. Three tries max.

Two minutes of writing. Your marketing manager can now run this without you watching — and the output will meet the standard you actually hold.

The three things you're probably wondering

“Do I have to rewrite my skill?”

No. Your skill is the starting material. Jetty adds the standards and check around it. The work you already did is preserved.

“What if I don't know what ‘good’ looks like in words?”

Most people don't, at first. That's fine. Run the skill on a real input. Look at the output. Tell Jetty what you'd change. The things you'd change are the standard. We'll help you put them into words.

“What if my skill is really just a good prompt?”

That works too. A “skill” in this context is just “instructions you've given an AI that you'd want to reuse.” It doesn't matter whether it lives in a SKILL.md, a custom GPT, or a Notion page.

Ready to convert yours?

Convert your first skill →

Bring a skill you've built. Get a runbook back in two minutes.

Book a 20-minute walkthrough →

If you'd rather walk through it with Jon on a call, pick a time.