chore(i18n): add locale contributor workflow

[Tzeenchist]

Summary

• document the locale contribution workflow in the repo
• add a scaffold script for new locales
• add a helper to report unchanged translation strings

Validation

• node scripts/create-locale.mjs
• node scripts/check-locale-translation.mjs fr
• node scripts/check-unused-i18n-keys.mjs
• ./scripts/check-i18n.sh

---

Note

Low Risk
Low risk: adds documentation and developer tooling scripts only; no runtime i18n behavior or CI enforcement changes.

Overview
Adds repo documentation for contributing new locales (docs/i18n.md) and a design note describing the two-PR plan for upstreaming Russian (docs/designs/russian-localization-upstream.md).

Introduces two Node scripts: scripts/create-locale.mjs to scaffold a new locale directory by copying English JSON namespaces, and scripts/check-locale-translation.mjs to report strings that are still identical to English (optionally failing via --fail-on-identical).

Exposes both utilities via package.json scripts (locale:new, locale:check) to standardize the workflow.

^{Written by Cursor Bugbot for commit 6eb3f77. This will update automatically on new commits. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

[cursor]

Hardcoded namespace list diverges from dynamic discovery

Medium Severity

check-locale-translation.mjs hardcodes namespaceFiles while create-locale.mjs dynamically discovers namespace files via fs.readdirSync. If a new namespace JSON file is added to the English locale directory, the scaffold script will copy it but the translation checker will silently skip it, leaving untranslated strings undetected. Since these two scripts form a paired workflow introduced together, they need to agree on which files to process.

Additional Locations (1)

• scripts/create-locale.mjs#L71-L75

 

[Juanlucasbg]

Aggressive review summary — PR #3297

342-line tooling addition: locale contributor workflow. New files: docs/i18n.md (contribution guide), scripts/check-locale-translation.mjs (validator), scripts/create-locale.mjs (scaffolder), updated package.json scripts, plus docs/designs/russian-localization-upstream.md (a previous-design reference). Verdict: mixed — recommend close in sunset window unless maintainer wants archival completeness.

What it does

Builds a workflow for contributors to add new locales: a scaffolder generates a new locale from the source-language file, and a validator checks that every key in the source has a translation in every locale (or is left as the source-language fallback). Adds pnpm i18n:create and pnpm i18n:check scripts.

Findings

• Structural — PASS (within the change's scope): Standard i18n contributor tooling. The create-locale.mjs scaffolder presumably copies the English source file with // TODO markers; check-locale-translation.mjs walks both files and reports key drift. Common pattern.
• Sunsetting context — concerning: Per the README banner and Sunset project routes to an export-only page (Vibe Kanban) #3387 sunset of project routes, the project is winding down. New locale contributions don't reach end users post-sunset. The most likely outcome of merging this PR is locale-pack PRs piling up as language packs that never ship — note that feat: add Russian language support #3212 (Russian), Add Hebrew translations and localization  #3288 (Hebrew), Add Portuguese language support to Vibe-Kanban #3096 (Portuguese), feat(i18n): add Turkish language support #2785 (Turkish), Feature/add turkish language support #2434 (Turkish — different author), When vibe coding breaks: structured debugging plugin for AI agents #3289 / Add Hebrew translations and localization  #3288 etc. are already queued. Adding contributor workflow tooling encourages more.
• Adversarial — N/A: Pure tooling, no runtime impact.
• Conventions — PASS: Documents are well-structured.

Maintainer decision points

1. Close: Sunset window says no new contributor flows.
2. Merge but reject downstream locale PRs: Tooling lands, but the maintainer doesn't accept new languages. Risks frustration.
3. Merge and accept all queued locale PRs: Burns through the queue at once. Highest goodwill but greatest scope creep against the sunset stance.

NITs (assuming merge)

• docs/designs/russian-localization-upstream.md: Reference doc for a prior design discussion. Useful archival but should be flagged as historical (date + status) so future readers don't think it's a current spec.
• Tests: Validator scripts have no tests of their own. A small fixtures-and-assertion test would prevent regressions when the validator's logic evolves.

Verdict

Recommend close in the sunset window. If the maintainer prefers to keep contributor tooling for forks / archival, merge — but pair with a public statement that no new locale PRs will be accepted upstream.

— Reviewed by automated single-pass review (sunset-window contributor-tooling triage; full 4-tool battery skipped — the policy decision dominates the technical review).