Reusable AI capabilities

Build it once. Sell it to every client.

A skill is a bounded, versioned, priced capability. Publish brand-voice-check on a Tuesday. By month three it has 11,200 calls and a 95% pass rate — without another hour of labour.

$0.01 floor price · no cut on marketplace revenue · 14-day free trial

agentskills.io: brand-voice-check · 11,200 lifetime calls · 95.3% pass rate · $0.04/call · Author: Northgate Agency

3,400

brand-voice-check calls in month 1 (12 workspaces)

$136

Credits earned, month 1 — $136

100%

Eval gate threshold — all test cases must pass

10ms

Routing on a highway vs new path

"Build it once. Sell it to every client. That is the economic logic of skills on ONE."

Anthony O'Connell · Founder · ONE

From Tuesday morning to passive revenue

Six commands. Forty minutes. Earning on every call.

1

Scaffold the skill

oneie skill new brand-voice-check. The CLI generates a markdown template with every supported field. Write your instruction body and two eval test cases. Ten minutes.

oneie skill new · skills/*.md · 10 minutes

2

Write the evals

At minimum: one test case that expects a pass, one that expects the specific failure you are guarding against. oneie skill eval runs test cases against the live model — not a mock.

evals: block · real model calls · 100% pass required

3

Publish to the marketplace

oneie skill publish. Validates frontmatter, runs evals, lists on agentskills.io. Price is set in the frontmatter. The platform takes no cut between authoring workspace and importer.

oneie skill publish · platform floor $0.01/call · you set the price

4

Earn on every call

Any workspace that imports your skill pays the published price per call. The substrate attributes credit to your workspace. The analytics dashboard shows per-skill call count, pass rate, and cumulative revenue.

L4 economic loop · direct credit attribution · dashboard visible

What skills give your agency

IP that earns without your time.

Build it once. Sell it to every client.

A skill is a bounded, versioned, priced capability. Write brand-voice-check on a Tuesday. Publish at $0.04/call. By month three, it has 11,200 calls and 95.3% pass rate. Every call earns without another hour of labour.

Eval gate — 100% pass required

Every skill must pass its test cases before publishing. Not 4 of 5. All of them. A skill that fires incorrectly in production corrupts the substrate's routing with positive pheromone. The gate is non-negotiable.

Semantic versioning + dependency boundary

Skills use semver. Importing workspaces stay on the pinned version until they explicitly refresh. A major version bump that breaks its own evals cannot publish. 40 clients on the same skill — update safely.

Marketplace at agentskills.io

Published skills list publicly on agentskills.io. Importers see lifetime call count and pass rate. 50,000 calls at 98.7% earns more trust than 200 calls with no evals. The marketplace converts on track record.

Pheromone attribution

After 50 calls, the path between calling agent and skill becomes a highway. Routing drops from ~1,500ms to under 10ms. Every call pays the authoring workspace. The L4 economic loop tracks per-skill revenue.

Six CLI verbs, full lifecycle

new · validate · eval · emit · publish · import. That is the complete workflow from scaffold to marketplace. oneie skill eval runs test cases against the live model. oneie skill publish blocks on any eval failure.

Agencies earning on skills

"brand-voice-check was live by 10:30. By 17:00 it had caught four off-brand drafts without a human touching them. By month three it had 8,200 calls across 12 workspaces. We never wrote it again."

Sarah O'Callaghan

Practice Manager · North Dublin Dental

"I published qualify-lead-dental. By month two, three other agencies in my vertical had imported it. I earn on every lead their agents qualify. That's what reusable IP looks like."

Marcus J.

Director · Northgate Marketing, Manchester

"The eval gate is the point. A skill that fires incorrectly in production is worse than no skill. We write the test cases first, then the body. Every skill we publish has a 95%+ live pass rate."

Brad S.

Founder · PT Corp

Skills vs. prompts stored in a CRM

FeatureONE SkillsPrompts in HubSpot / Google Docsn8n / Make automations
Versioned, priced capability per skill
Eval gate (100% test pass required)
Marketplace with revenue attribution
Semver dependency boundary
Highway formation per skill (sub-10ms routing)
Skills compose (call a skill from a skill)
Platform takes a cut of marketplace revenueLower = better. Direct attribution: authoring workspace earns the full published price.n/a

Publish and earn on every plan

Annual billing pre-selected. You keep the full published price.

Starter

Import skills · run evals

$500/mo
5M credits/mo
Import from marketplace
Run evals locally
5 custom skills
Email support
Most popular

Agency

Publish + earn · private skills

$5,000/mo
60M credits/mo
Unlimited custom skills
Publish to agentskills.io
Private skill whitelisting
Per-skill revenue dashboard
Priority support

Scale

Cross-client skill learning

$50,000
750M credits/mo
Cross-client skill performance aggregation
Skill compliance reporting
Dedicated success manager
SLA 99.99%

Questions about skills

The first agency to own qualify-lead-dental owns it.

Publish the skill. Earn on every call from every agency that imports it. The moat is the accumulated signal, not the code.

$0.01 floor · no cut on marketplace revenue · 14-day free trial