ru-score: A Numerical Quality Gate for Russian Text Agents

Where this fits

A developer asks an agent to review a Russian landing page. The agent returns a list of editorial findings, but the developer actually needs one number to decide whether the copy is shippable. ru-score provides that number: 0.0–10.0 across typography, clean language, grammar, structure, and precision for the reader, without editing the file. The skill is meant for the moment you need a quality gate, not a proofreading session.

Why agents benefit

• Returns a scalar score, so agents can enforce thresholds in release pipelines.
• Uses a fixed rubric stored in ru-text references, making scores consistent across runs.
• Disallowed tools (Write, Edit, Bash) mean it cannot modify text or trigger side effects.
• Trigger phrases like "оцени текст" and "балл за текст" make it easy for Russian-speaking users to invoke.
• Finds references through Read/Glob/Grep only, so it works in forked or sandboxed contexts.

Practical scenarios

Pre-publish gate for Russian UI copy

Before shipping an onboarding flow, the developer runs ru-score on final strings. Any string below 7.5 is flagged for human review; no file is changed silently.

Compare drafts before A/B testing

Two email variants are scored with ru-score. The higher-scoring variant enters the test, and its score is stored as a covariate for later analysis.

Batch content triage

An agent scores a folder of Russian articles and emits a table of numbers. The developer sorts by the lowest scores and prioritizes rewrites without reading every report.

Add it to your agent workflow

Install the skill with:

npx skills add talkstream/ru-text --skill ru-score

Then invoke it naturally:

User: Оцени текст: Наш сервис помогает командам быстрее публиковать контент.
Agent: 8.6

Because the skill is user-invocable, you can bind it to a slash command or call it with text passed as $ARGUMENTS.

Compare before adopting

Compare quality signals: Does the five-dimension rubric match your editorial standards? Check maintenance freshness by looking at recent commits and issue responses. Consider alternatives: a tool that lists findings is better for rewriting, while ru-score is better for yes/no decisions. Since it never edits files, pair it with an editing skill if you also need corrections. You can compare similar skills in the OpenAgentSkill directory: https://www.openagentskill.com/skills/talkstream-ru-score.

Why it is worth tracking

With 224 stars from talkstream, ru-score shows community momentum for Russian text quality scoring. Track it if you routinely evaluate Russian-language agent output or want a repeatable quality metric. Re-evaluate it when the underlying ru-text rubric changes or when your content standards shift; a scoring skill is only useful while its scale is aligned with what you ship.