naming · timschoch — Claude Code skill: code naming rules

What naming does

The skill keeps one rule list, synced from the hub. Before naming anything, the agent runs node .agents/skills/naming/scripts/config.mjs and treats the output as the live rules: the script prints references/naming.json merged with your project overrides. Some of it a machine can check — the naming gate runs on push and in CI — and the rest is left to judgement.

Twelve rules sit inside. A word comes from GLOSSARY.md first, then from existing types, tables and routes; the repo's word outranks the word in the request. If no word exists yet, term-check runs first and the term goes into the glossary through domain-modeling. Prefixes I and T are out, along with Hungarian notation, abbreviations like opts and err, and single letters. Names are assembled from ordered parts, largest to smallest; an action opens with a verb from prefixes, and an event reads as a fact in the past. enum is not used — a string-literal union or an as const object takes its place.

Overrides go in .skilly/naming.json, same shape as the base file with only the keys you change. An existing repo convention beats the skill's rules, and you set it in artifacts.file.case and artifacts.folder.case. Once a meaning shifts, the rename has to land everywhere in one change and the glossary has to follow, or code and docs end up with two words for one thing.