A few months ago, a talented developer on a team I work with had a skill that worked, and didn't share it.
"It works. But I don't feel like it's mature enough."
What stopped him wasn't the skill. It was the fear that what he'd built would look not good enough, in front of everyone.
Today, several of his skills run across the whole team, in development and during on-call shifts. They save time and catch problems early.
He didn't change. The place where he's allowed to get things wrong did.
You have to let your team get things wrong and fail, and enjoy it. A team Harness is one way to do that.

The organization's Harness isn't enough on its own
Many organizations build a shared Harness: skills, custom sub-agents, hooks, and more. It's an important foundation. That's how knowledge moves between teams, and the next person doesn't start from scratch.
But a lot of what gets built inside teams never makes it there. From working with teams, I see two reasons.
The first is technical: uploading involves a lot of friction, review depends on a few people, and there are unnecessary steps every time. One team lead told me he doesn't want to open a ticket and run a full CI pipeline for a four-line change.
The second is talked about less: putting something in the shared space means exposing it to everyone. And maybe being embarrassed.
A convenient Harness solves the first. It doesn't solve the second.
A team Harness, alongside the organization's Harness
That's what a team Harness is for. A folder in the team's repo, or a separate repo if you share a repo with many teams.
Are you crazy? Duplication! More maintenance!
Honestly? True. That's why it doesn't replace the organization's Harness. It sits alongside it, in parallel.
For that developer, it looked like this: the team set up a split in the repo between things that are mature and things that are still experimental. Half-jokingly, they called it "baked" and "unbaked." Suddenly, uploading something unfinished became a normal thing, not a confession.
Uploading to the experimental side is simple. You don't skip evals on the way to production, but when you're only checking whether something works and is useful to someone other than me, you don't need the whole wrapper. And when his skill grew, the team itself asked to add a check before merging. It's not all or nothing.
What proves itself moves up to the organization. And as confidence builds, that jump gets easy.
Don't stop at a feeling: measure
- Contribution: how many things were added to the team Harness this month, and by how many different people.
- Real usage: how many of them run for people who didn't build them.
- Then connect it to the metrics that matter to the team. For example, Cycle Time: how long it takes from the moment you start working on something until it reaches the customer. If the skills really catch problems early, fewer things come back for fixes, and that's where it should show.
Questions for your morning coffee
Next week, over your morning coffee, ask:
- Does my team feel comfortable taking part in the organization's Harness?
- If not, why? A technical reason? Or something deeper?
- Which skill works for someone on the team, and nobody else knows about it?
And for you? Where do your team's best skills live, in the shared space or on someone's laptop?