Create API Governance Program — Development skill for Claude Code

create-api-governance-program

A [Claude Code](https://claude.com/claude-code) skill for designing an API governance program that an organization will actually adopt.

Part of Emmanuel Paraskakis's API method. Where `design-api` takes requirements to a spec and `design-api-review` checks a spec against standards, those govern *one* API. This one governs *the estate*.

What it does

Scores an organization across five dimensions, then designs **one level of movement** — not the end state.

Dimension	The question it answers
Standards	Is there a written ruleset, owned and versioned?
Design review	Does anything check a design before it's built?
Automation	Can a machine catch violations without a meeting?
Lifecycle	Are versioning, deprecation and discovery governed?
Metrics & feedback	Is governance measured, and do the standards change as a result?

Levels run **1 Ad hoc → 2 Documented → 3 Reviewed → 4 Automated → 5 Measured**.

The output is a markdown scorecard, a one-increment plan with named owners, a rollout starting from a volunteer pilot, and a "deliberately not doing yet" table where each deferral carries a revisit trigger.

Why one increment

Governance programs characteristically fail by arriving fully formed — a council, a 40-page style guide, a mandatory review board, a portal — before anyone has felt the problem they solve. Teams route around it and governance becomes a tax. Moving one dimension one level, proving it, then moving again is slower on paper and faster in practice.

Layout

SKILL.md                      the method
references/maturity-model.md  5x5 matrix, evidence signals, scoring rules,
                              and five common estate patterns
evals/evals.json              3 test cases, 31 assertions

Installing

Symlink it where Claude Code looks for skills:

ln -s "$PWD" ~/.claude/skills/create-api-governance-program

Then invoke with `/create-api-governance-