ux-writing

数据/ML仪表盘的语音、语气与内容指南。涵盖微文案、错误信息、成功状态及数据呈现语言。在文案、消息、内容、标签、错误信息关键词上自动触发。

SKILL.md

--- frontmatter

name: ux-writing
description: Voice, tone, and content guidelines for data/ML dashboards. Covers microcopy, error messages, success states, and data presentation language. Auto-invokes on copy, messaging, content, labels, error messages keywords.
allowed-tools: Read

UX Writing (Core)

Voice

•Professional yet Accessible - Technical precision without jargon
•Educational - Explain complex concepts clearly
•Evidence-Based - Data-driven with context
•Progress-Oriented - Focus on forward momentum

Tone by Context

Context	Tone	Example
Executive	Strategic, narrative	"ML models require quality data for production performance"
Analyst	Precise, data-rich	"Retention: discovery → silver dataset"
Error	Calm, solution-focused	"Unable to load metrics. Check connection and refresh."
Success	Confirming, brief	"Pipeline completed successfully."

Microcopy Patterns

Buttons

•Primary: "View Dashboard" (not "Click to View")
•Secondary: "Learn More" (not "Read Documentation")
•Destructive: "Delete Dataset" (not just "Delete")

Errors

Structure: Problem + Solution

•"Unable to connect. Check URL and try again."
•"File exceeds 10MB. Compress or split."

Empty States

Structure: Explanation + Action

•"No datasets yet. Upload your first dataset to get started."

Loading

Be specific: "Loading dashboard..." not "Loading..."

Data Presentation

Numbers

•Use commas: 1,234,567
•One decimal for %: 45.2%
•Context: "1,234 records (+12% from last month)"

Trends

•Positive: "+12.5%" or "↑ 450 records"
•Negative: "-8.2%" or "↓ 120 records"

Accessibility

Alt Text

Describe: "Bar chart showing training data distribution" Not: "Image of chart"

Links

Specific: "View dataset documentation" Not: "Click here"

Preferred Terms

Use	Avoid
Dataset	Data set
Pipeline	ETL process
Real-time	Realtime
Training data	Train data

Version: 2.0.0