Content
85%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A well-crafted, highly actionable skill body: clear task routing, copy-paste commands, explicit validation with expected outputs, and sensible global-vs-locale rule layering. Its two weaknesses are the dead references to missing `locales/*.md` files and repeated restatement of the locale-guideline-precedence rule.
Suggestions
Ship the 19 `locales/*.md` guideline files referenced in the Locale-Specific Guidelines table (or gate the table on their presence), since the translation workflow depends on reading them and every link is currently broken.
Consolidate the locale-guideline-precedence instruction into a single authoritative statement (e.g., in Locale-Specific Guidelines) instead of repeating it in Translating, Updating an Existing Translation, and the General Translation Rules intro.
State explicitly what to do when the completeness check still reports missing/TODO entries after translating (e.g., re-run, fix, re-validate) to close the error-recovery loop in the Validation section.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and assumes competence — no explanation of what i18n is, commands given directly, tables instead of prose. It falls short of 5 because the instruction to read the locale-specific guideline and let it override global rules is repeated in four places (Translating, Updating an Existing Translation, General Translation Rules intro, and Locale-Specific Guidelines), which could be consolidated into one authoritative statement. | 4 / 5 |
Actionability | Fully executable guidance throughout: exact `mkdir -p` path, `breeze ui check-translation-completeness` commands with precise flags (`--add-missing`, `--remove-unused`), `prek run --from-ref main --hook-stage pre-commit`, exact config edit formats for three files, and a sample of the scaffolded JSON output. Placeholders like `<locale>` are legitimate parameters, not pseudocode. | 5 / 5 |
Workflow Clarity | Clear routing decision up front (locale exists vs. not), sequenced steps for both branches, and a dedicated Validation section with explicit expected outcomes ("0 missing, 0 TODOs, 0 unused, and 100% coverage") plus a lint/format pass. The destructive `--remove-unused` step is gated by a prior completeness check, so the batch/destructive validation cap does not apply; the re-runnable completeness check provides the feedback loop. | 5 / 5 |
Progressive Disclosure | The locale table is well-signaled and references are one level deep, but none of the 19 referenced `locales/*.md` guideline files (nor any bundle directories) exist in the skill bundle, so every link in the Locale-Specific Guidelines table is dead. Not 2, since the SKILL.md body itself is well-structured and the intended split is clear; not 4, because navigation to the referenced material is broken as shipped. | 3 / 5 |
Total | 17 / 20 Passed |