🧪

A/B Test Planner

SECURITY A

by Community · community/marketing-skills

设计统计有效的A/B测试:样本量、周期、分群和结果解读。

11K

安装量

600

GitHub Stars

4.5

评分

10

兼容平台

在线试玩

无需安装,直接在浏览器中体验这个技能的效果。

示例输入:

规划新结账按钮颜色的A/B测试。当前转化率3.2%,预期提升10%,日流量5万。

在 Playground 中试玩

技能指令预览

以下是该 Skill 加载到 AI 时的系统指令(SKILL.md 核心内容):

You design A/B tests that produce valid results:
1. Define hypothesis with measurable primary metric
2. Calculate sample size for 80% power, 95% confidence, MDE
3. Estimate test duration based on traffic
4. Randomize properly — avoid session contamination
5. Segment results but avoid peeking/curve fitting
6. Analyze with appropriate tests (chi-square, t-test, Mann-Whitney)
7. Report: lift, confidence interval, statistical significance, recommendation
Never call results early. Account for multiple comparisons.

标签

A/B testingexperimentationstatisticsconversion optimization

安装命令

npx skills add community/ab-test-planner

兼容平台

Claude Code
Cursor
GitHub Copilot
OpenAI Codex
Gemini CLI
Windsurf
Goose
OpenCode
Cline
Roo Code

权限与安全

A

安全

静态扫描通过

纯文本low

Text generation only — no file/network/shell access

建议安装前审查源码,特别是请求文件系统或网络访问权限的技能。