Skills drift silently.
Until they don't.
Metaskill is the layer that decides which skill runs, in what order, and whether the result is trustworthy enough to ship.
Ungoverned skills fail quietly.
- 01
Skills pile up with no owner, no version, no review date.
- 02
Two skills quietly do the same job — until they disagree, in front of a client.
- 03
Nothing traces an AI output back to which methods actually shaped it.
Five parts. One governance layer.
Not a feature grid. Each part answers a different governance question — routing, ownership, quality, traceability, and naming — and they only make sense together.
Routes each request to the right skills, in a fixed precedence order — context, then structure, then brand and register, then phrasing — so a later step never overwrites an earlier decision.
One pass. Every output governed.
request → Conductor → skills (precedence) → Pyramid gate → Ledger → shipped + orchestration note
- STEP 01Request arrives
A prompt or task enters the layer. Metaskill treats it as input to be governed, not fulfilled blindly.
- STEP 02Conductor selects skills
The Conductor identifies which skills apply and runs them in precedence — context, structure, brand, phrasing.
- STEP 03Pyramid gates the result
The output is measured against a five-level quality gate. Anything under threshold does not ship.
- STEP 04Ledger records everything
Which skills shaped it, which gate level it passed, who owns it, when it's due for review. Attached to the output.
Two profiles. Same governance need.
Running many skills, at speed.
You have a growing library of prompts, workflows and agents. Duplication is quiet, ownership is fuzzy, and nobody can tell you which methods shaped last week's client deliverable. Metaskill removes the guesswork.
Solo, but shipping serious work.
You need enterprise-grade traceability without an enterprise-size team behind you. Every output leaves an audit trail your future clients — and your future self — can rely on.
The thinking, in long form.
What a Metaskill is, and why AI skills need governance
A working definition of a skill, the three ways ungoverned skills fail, and the three things a governance layer must do — routing, precedence, and a quality gate — before any of it ships. Written for teams and operators already running AI at real stakes.
Governance before it's needed
is worth more than after.
Metaskill is opening access to a small first cohort. Leave an email and we'll be in touch.