| name | localize-strings |
| description | Run the build-add-in localization workflow when UI strings are added or changed, including extracting strings, generating translation prompts, updating TSV language files, embedding localized output, and validating results. |
Localize Strings Workflow Skill
Use this skill when the user asks to localize new or changed UI strings in this repository.
Purpose
Apply the documented localization process in docs/localize.md and related localize/ files.
Repository Rules
- Only localize UI strings.
- Ensure UI strings are wrapped in
loc("...") before running localization tooling.
- If a language translation is missing, English is the fallback.
- Keep changes minimal and aligned with the existing project style.
Required Workflow
- Run string extraction:
- Perform translation directly:
- Use
localize/translate-prompt.txt as the translation prompt.
- Generate translated TSV output.
- Apply results to
localize/strings.tsv.
- Embed localized language resources:
- Validate output quality:
- Spot-check changed strings in
src/core/localize/strings.json and generated artifacts.
- Confirm no obvious formatting or placeholder regressions.
- Report exactly what changed and any errors or issues encountered.
Operating Behavior
- Execute translation directly as part of the workflow; only ask the user for input if translation fails or output is ambiguous.
- Automatically run localization commands using the platform shell:
- Windows: run in
cmd (cmd /c "npm run extract-loc" and cmd /c "npm run embed-loc").
- macOS: run in
zsh (zsh -lc "npm run extract-loc" and zsh -lc "npm run embed-loc").
- Preserve TSV formatting exactly (tabs, column order, escaping).
- Do not introduce new dependencies.
- Prefer targeted edits; avoid unrelated refactors.
Expected Deliverables
- Updated localization source files (typically
localize/strings.tsv, plus embedded outputs).
- Short summary including:
- Commands run.
- Files changed.
- Any manual follow-up required.
- Any strings that remain intentionally untranslated (for example, approved brand/product names).
Quick References
- Process guide:
docs/localize.md
- Prompt template:
localize/translate-prompt.txt
- Localization table:
localize/strings.tsv
- Embed command source flow:
npm run embed-loc
