string-reviewer_skill

Overview

This skill reviews user-facing strings and UI copy across a TypeScript codebase to ensure clarity, brevity, consistency, and actionable guidance. It acts as a Senior UX Writer that flags problematic text, maps issues to voice principles, and proposes precise replacements. Use it to keep CLI and agent strings deterministic, scannable, and repairable without changing code autonomously.

How this skill works

The skill scans inline text, error messages, prompts, and settings labels to detect violations of core voice principles (deterministic clarity, system transparency, front-loaded actionability, and agentic error recovery). For each issue it produces a concise rationale and a suggested replacement following the required list format. It never rewrites strings without explicit user approval and pairs every error critique with one clear corrective action or prompt suggestion.

When to use it

• Before a release to audit CLI messages and Git-tracked UI strings
• When adding new commands, flags, or error cases to ensure consistent voice
• During localization prep to simplify and normalize source strings
• When addressing user confusion reports tied to ambiguous messages
• When updating settings schema labels or descriptions

Best practices

• Always front-load intent: use the [Goal] + [Action] pattern for prompts and buttons
• Keep status and loading states under five words and show telemetry for long tasks
• Use deterministic verbs for system/infrastructure and probabilistic verbs for Gemini outputs
• Pair every error with exactly one immediate recovery command or suggested prompt
• Avoid first-person pronouns; attribute thinking to Gemini and doing to the system

Example use cases

• Audit CLI error messages to add one-click recovery commands and human-readable context
• Review new feature copy to ensure the first three words describe user intent
• Normalize settings labels and descriptions after schema changes
• Convert polite loading text into telemetry-driven micro-state cycles for long-running tasks