Why I Open-Sourced 96 Agent Skills

Most so-called proprietary playbooks are not a moat. They're documentation with an ego problem.

CS leaders love talking about secret sauce as if the value sits in the template, the framework, or the internal operating language wrapped around it. Usually it doesn't. Usually the thing being protected is a decent arrangement of public ideas plus a flattering story about why nobody else could possibly replicate it.

That's fantasy.

The value is not in owning the artifact. The value is in knowing when the artifact works, when it breaks, and why.

That distinction matters more than most teams want to admit.

The Playbook Is Not the Edge

When you run CS at scale, standardisation and reality start fighting almost immediately. You build a clean process for one cohort. You roll it out across the function. Within months, your best operators have already bent it around the accounts that matter most. A year later, the official playbook still exists, but the real operating model is living in judgment, exceptions, side-notes, and undocumented adaptations. The company thinks it owns a system. What it actually owns is a gap between what the system says and what good people do to make it work.

Locking the tools down doesn't fix that. It just makes the illusion easier to maintain.

That's why I open-sourced the skills.

If a tool is genuinely useful, it should survive contact with people outside your own environment. If it can't, the problem is not that the market lacks sophistication. The problem is that you've overestimated the tool.

Closed Systems Hide Failure

Open-sourcing forces the work into contact with reality. People don't politely nod at it. They fork it. Break it. Remove parts. Adapt it to messier conditions. Use it in ways you didn't expect. Ignore the parts that only made sense to you. That is not a downside. That is the test.

A closed system gives you agreement. An open one gives you evidence.

And evidence is more valuable.

A CSM telling you the playbook works tells you almost nothing. People say systems work all the time while quietly routing around them. A user showing you the exact point where the logic collapses under a more complex stakeholder structure, stranger renewal motion, or less forgiving operating environment: that's signal. In private, failure gets hidden. In public, failure leaves fingerprints.

Open source is useful because it makes the work answerable to reality.

The Real Edge Is Judgment

That is why the commodity-vs-judgment distinction matters.

A reusable skill that surfaces an upsell signal is useful. But the skill is not the edge. The edge is knowing whether that signal is real, whether it's mistimed, whether the account politics make it irrelevant, whether the customer is structurally unstable, or whether the whole thing is just noise dressed up as insight. The repeatable mechanic can be shared. The judgment sits somewhere else.

That's the part worth caring about.

Open-sourcing the skills didn't weaken the advantage. It exposed where the advantage actually was. Not in the mechanics. In the interpretation. Not in the template. In the decision quality around the template.

Why Open Source Makes the Work Better

Open systems are easier to build on. Closed systems create dependency on local language, local habits, and institutional memory. Open systems make the work legible. They let other people understand the logic faster, question it faster, and improve it faster. That doesn't weaken the work. It hardens it.

It also forces intellectual honesty. If the value disappears the moment other people can see the thing, the value was never very deep. If the work gets stronger when other people pressure-test it, you've learned something useful about where the real leverage sits.

And if someone takes the skills and builds a better version, good. That's not theft. That's evidence that the market found a better answer faster than you did. The correct response isn't defensiveness. It's to learn from it and move again.

What Is Actually Worth Protecting

Too much CS thinking still treats control as if it were defensibility. It isn't. Control can preserve an artifact. It does very little to improve one.

The goal was never to own the playbook. The goal was to produce better judgment, better decisions, and better customer outcomes. If the work can't survive exposure, it probably wasn't that good. If it gets stronger when other people touch it, you've learned where the real value sits.

I've become far less interested in protecting the machinery.

I'm much more interested in understanding what still works after the machinery is no longer hidden.

That's where the edge is.