@clipkit/lint

Protocol-aware soft checks and plain-language summaries for Clipkit Sources.

These are pure functions over a validated Source — no I/O, no rendering. They catch the things that pass schema validation but still surprise you at render time, and they describe a project in words so a human or an agent can sanity-check it without rendering. Shared by @clipkit/cli (explain, validate --explain) and @clipkit/mcp-server (validate_project, describe_project).

import { lintSource, describe } from '@clipkit/lint';

for (const w of lintSource(source)) console.log(`${w.where}: ${w.message}`);
console.log(describe(source));

lintSource(source) → LintWarning[]

Returns warnings for things the runtime will silently drop or clip:

• No top-level duration — the runtime can't tell how long the composition is.
• Element runs past the composition end — it'll be cut off.
• Non-ASCII text / caption words — the runtime text atlas is a fixed ASCII coverage font, so emoji, accents, smart quotes, and CJK are silently dropped.

LintWarning is { where: string; message: string } — where is an element id or (source).

describe(source) → string

A plain-language read-back: dimensions · fps · duration · format, an element breakdown by type, a per-track timeline (paint order low→high), and the lintSource warnings.