Pricing a Skill: What 70/30 Splits Tell You About the Market

Every marketplace eventually has to answer an uncomfortable question: who actually made the sale — the person who built the thing, or the platform that put it in front of a buyer? The answer never gets stated that plainly, but it gets encoded, permanently, in the revenue split. And in 2026, curated skill marketplaces have converged on a number that's worth sitting with: creators keep 70%, platforms take 30%.

That number isn't arbitrary, and it isn't new — it's close to what Apple charged for years on the App Store, what many SaaS marketplaces charge resellers, and what a wide swath of creator platforms have settled on across completely different categories. When a genuinely new market lands on an old, familiar split that fast, it's worth asking what that convergence is actually telling you, because it's rarely a coincidence and rarely arbitrary.

The Split Is a Statement About Who's Replaceable

A revenue split is, structurally, a bargaining outcome dressed up as a policy. The platform sets it, but the number it lands on is constrained by what creators will tolerate before they take their skills elsewhere — or stop making them. A 70/30 split, where the creator keeps the majority, is the platform quietly admitting that the marketplace needs creators more than any specific creator needs the marketplace. That's a genuinely different power dynamic than, say, early app stores in categories with no alternative distribution, where splits sometimes ran the other direction.

The fact that skill marketplaces landed near 70/30 rather than 50/50 or worse tells you creators in this category have real outside options. A skill built to the SKILL.md convention, as covered earlier in this series, is portable across more than 20 different agents and tools. A creator who's unhappy with one marketplace's terms doesn't have to rebuild their product to sell somewhere else, or nowhere at all — they can just point buyers at their own GitHub repo, or list on a competing marketplace, with minimal rework. That portability is doing real work on the pricing power side of this equation, even though it's usually discussed purely as a technical convenience.

What the 30% Is Actually Paying For

It's tempting to read the platform's cut as pure rent-seeking — a toll for doing nothing but hosting a listing page. That's not quite right, and it's worth being specific about what 30% is buying, because it clarifies what a marketplace has to keep delivering to justify the number.

Discovery is the obvious one: a marketplace with real traffic puts a skill in front of buyers who would never have found a solo creator's repo, which is the subject of its own piece later in this series. But discovery alone doesn't obviously justify 30% — plenty of distribution channels charge far less. The second, less obvious thing the cut pays for is trust infrastructure: payment processing that a buyer trusts enough to hand over a card number, refund handling when a skill doesn't do what it claimed, and — increasingly, given how badly this category needs it — some layer of verification that the skill isn't going to run a malicious script against the buyer's codebase. That verification layer is expensive to build and expensive to staff, and it's the part of the 30% that's hardest to replicate by just listing on GitHub instead.

This is why the split isn't purely extractive, even though it can look that way from the creator's side. A marketplace that takes 30% and does nothing with it beyond hosting a page is vulnerable to exactly the disintermediation described above — creators leave for cheaper distribution the moment their skill has enough of a following to not need the marketplace's discovery anymore. A marketplace that takes 30% and spends real effort on curation, verification, and buyer trust is offering something a solo GitHub listing genuinely can't replicate, which is what justifies the number holding rather than eroding over time.

Where the Split Breaks Down

70/30 works cleanly for a single, one-time skill sale. It gets messier the moment a skill's value comes from ongoing maintenance rather than a one-shot purchase — a skill that needs to keep working as the underlying agents change their tool-calling behavior, or that depends on an external API that shifts its contract periodically. A flat percentage of a one-time sale doesn't obviously compensate a creator for six months of unpaid bug fixes after the fact, and marketplaces in this category haven't fully solved that yet. Some are experimenting with subscription-style skill access instead of one-time purchase precisely because it aligns the split with ongoing effort rather than a single transaction.

There's also a quality-adjustment problem the flat split doesn't address. A skill that took a weekend to write and a skill that took three months of testing against edge cases both get the same 70% if they're priced the same, which creates a real incentive to ship fast and shallow rather than slow and reliable — an incentive that cuts directly against the trust and discovery problems the rest of this series covers. A marketplace that wants durably good skills, not just a high volume of listings, eventually has to find a way to reward depth that a uniform revenue percentage doesn't naturally provide on its own.

The Comparison to the MCP Ecosystem Is Instructive

It's worth noting what the adjacent MCP server ecosystem looks like by contrast: mostly free, mostly distributed without any marketplace layer at all, and correspondingly under-verified — the 2026 audit finding the majority of MCP servers vulnerable to path traversal and code injection is, at least in part, a symptom of there being no curated marketplace collecting a cut in exchange for vetting anything. Nobody's incentivized to fund verification when nobody's paying for distribution. The 70/30 split, whatever its flaws, at least creates a funding mechanism for the trust work this category badly needs — which is a point in its favor even for creators who'd rather keep the extra 30% themselves.

What the Number Will Do Next

Splits in mature creator economies tend to move in one of two directions over time: toward the creator, as platforms compete for supply in a crowded market, or toward the platform, as a marketplace accumulates enough network effects that creators lose their outside options. Which way skill marketplaces go depends almost entirely on whether skill discovery — genuinely finding the right skill for a specific problem, as opposed to just listing thousands of them — turns out to be a solved problem or a permanently hard one. If discovery stays hard, the marketplaces that solve it well accumulate real leverage and the split could tighten toward the platform over time. If discovery gets commoditized by better search and better agent-native recommendation, creators' outside options stay strong and 70/30 — or something even more creator-favorable — is likely to hold.

Either way, the number itself is worth watching as a leading indicator, not just an accounting detail. A revenue split is a market's honest answer to the question of who actually creates the value — and right now, in the skill economy, that answer still points mostly toward the people writing the skills.

Part of the "The Skill Economy" series on aiskill.market.