Menu
当需要设计或执行A/B测试实验时使用。实验验证指挥官，调度experiment-design/execution。关键词：A/B测试、实验设计、统计显著性、实验执行、效果验证、AB测试、对照实验。
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Use when managing Sprint cycles or tracking agile execution. Agile execution commander orchestrating agile-sprint-planning, agile-daily-sync, and agile-review sub-skills. agile-review has merged retrospective-auto auto-retrospective capabilities. Keywords: agile execution, Sprint planning, daily standup, Sprint review, agile management, Sprint retrospective, iteration retrospective, agile development, retrospective report, auto-retrospective.
Use when planning a Sprint. Sprint Planning automation, transforming Product Backlog into Sprint Backlog, including Sprint Goal suggestions, Story auto-selection, workload estimation, and capacity matching validation, outputting a complete Sprint plan. Keywords: Sprint planning, Sprint plan, iteration planning, Story selection, capacity matching, scheduling, what to do this iteration.
Use when you need to consolidate competitor tracking data into a complete, deliverable monitoring report. Competitor Dynamic Monitoring Report auto-generation, including competitor dynamics summary, feature change tracking, market strategy changes, threat assessment, and response recommendations. Keywords: competitor monitoring report, competitor dynamics, feature tracking, threat assessment, competitor response, competitor report, what are competitors doing.
Use when you need to track competitor dynamics and develop response strategies. Competitor Dynamic Tracking & Response, monitors competitor feature changes, evaluates dynamic changes in own advantages, generates response strategies, and tracks effectiveness. Keywords: competitor tracking, competitor analysis, competitor monitoring, feature changes, competitive analysis, competitor changes, competitor dynamics, competitor changed, competitor made a move.
Use when you need to diagnose product health. Automated product health diagnosis that collects multi-dimensional data and performs comprehensive scoring, trend prediction, and bottleneck identification, outputting a health report. Keywords: health score, product diagnosis, multi-dimensional scoring, health check, product health, health rating, product checkup, status check.
Use when planning iteration cycles or adjusting product priorities. Iteration decision commander orchestrating iteration-backlog-grooming and iteration-retrospective sub-skills. Keywords: iteration decision, Backlog optimization, priority adjustment, iteration retrospective, iteration planning, requirement restructuring, RICE scoring, iteration management. This orchestrator dispatches 2 sub-skills: Backlog grooming (no cross-module dependencies) and iteration retrospective (depends on pm-08 output).
| name | experiment-orchestrator |
| description | 当需要设计或执行A/B测试实验时使用。实验验证指挥官，调度experiment-design/execution。关键词：A/B测试、实验设计、统计显著性、实验执行、效果验证、AB测试、对照实验。 |
| metadata | {"module":"产品度量运营","sub-module":"实验验证","type":"orchestrator","version":"8.0","domain_tags":["通用"],"trigger_examples":["设计一个A/B测试","验证一下方案效果","跑一下对照实验","分析实验结果"]} |
实验是学习的最快方式
每一个实验都是一次有控制的探索，目标不是证明假设正确，而是以最快速度获得可靠的学习。实验的价值在于学习速度，而非实验数量。
遵循 orchestrator-protocol.md 编排协议。
pipeline: experiment-orchestrator
version: 8.0
post_pipeline:
- action: stage-summary
output: output/phase-reports/pm-metrics-ops/experiment-orchestrator.md
stages:
- id: phase-1
name: "实验设计"
depends_on: []
skills: [experiment-design]
gate:
condition: "实验设计经人类审核确认"
fail_action: "阻止实验上线，修改后重新审核"
- id: phase-2
name: "实验执行"
depends_on: [phase-1]
skills: [experiment-execution]
gate:
condition: "样本量充足且统计检验完成、实验报告经人类审核确认"
fail_action: "延长实验周期或扩大流量"
Skill: experiment-design
输入:
hypothesis: 用户提供（假设陈述）
available_traffic: 用户提供（可用流量）
metrics_system: metrics-system → metrics.json（可选）
historical_data: analysis-funnel/analysis-retention（可选）
输出: output/pm-metrics-ops/experiment-design/
验证: 假设已结构化（If-Then-Because-For）；主指标与假设直接对应；护栏指标覆盖留存、收入、技术三个维度；样本量计算参数有据可依
模式: 🤖→👤
Skill: experiment-execution
输入:
experiment_design: output/pm-metrics-ops/experiment-design/experiment_design.json
experiment_data: 用户提供
termination_conditions: output/pm-metrics-ops/experiment-design/experiment_design.json
product_background: 用户提供（可选）
输出: output/pm-metrics-ops/experiment-execution/
验证: 实验分组流量分配正确；护栏指标未触发告警；实验数据采集完整；统计显著性计算正确；统计结论与数据一致；行动建议与结论一致；护栏指标全覆盖；异质性效应已分析（至少3个分群维度）
模式: 🤖→👤
所有子Skill执行完成后，必须生成阶段总结文档，写入 output/phase-reports/pm-metrics-ops/experiment-orchestrator.md，包含以下6项结构（均不可为空）：
| 参数 | 值 |
|---|---|
| 子Skill输出路径 | output/pm-metrics-ops/ |
| 总结输出路径 | output/phase-reports/pm-metrics-ops/experiment-orchestrator.md |
| 审批记录路径 | output/approvals/{orchestrator-name}/{stage-id}.approval.json |
下游衔接: primary: decision-orchestrator（实验完成，将实验结论转化为决策行动） alternatives: - target: release-orchestrator reason: 实验结果显著，建议全量发布 condition: 实验结果统计显著（p<0.05）且业务意义达标时 - target: analysis-orchestrator reason: 实验结果需更深入的数据分析 condition: 实验结果存在异常或需多维下钻时 special_cases: []
| 卡口 | 条件 | 未通过处理 |
|---|---|---|
| 实验方案人类已审核 | 实验设计经人类审核确认 | 阻止实验上线，修改后重新审核 |
| 统计显著性已判断 | experiment-result输出文件已生成且非空 | 延长实验周期或扩大流量 |
| 实验报告已审核 | 实验报告经人类审核确认 | 补充分析或修改结论 |
| 阶段总结已生成 | output/phase-reports/pm-metrics-ops/experiment-orchestrator.md 已生成且6项结构均非空 | 补充缺失结构项后重新生成 |
| 决策点 | 触发条件 | 决策内容 |
|---|---|---|
| 实验方案审核 | 实验设计完成 | 审核假设合理性、指标选择、分流方案 |
| 全量/终止决策 | 实验结果分析完成 | 决定全量发布、终止实验或延长周期 |
| 实验报告确认 | 实验报告生成完成 | 确认报告结论和行动建议 |
| 条件 | Action |
|---|---|
| 样本量达到100% | 立即触发结果分析 |
| 统计显著（p < 0.05）且稳定 | 考虑提前终止 |
| 护栏指标显著下降 | 触发告警，考虑终止 |
| 新奇效应显著 | 延长实验周期 |
| 实验组持续负向 | 考虑提前终止 |
| 异常类型 | 处理策略 |
|---|---|
| 实验设计人类审核未通过 | 阻止实验上线，返回设计阶段修改，不进入执行阶段 |
| 护栏指标突破阈值 | 立即暂停实验执行，触发告警，提交人类决策是否终止实验 |
| 实验数据采集异常 | 标记数据异常，暂停统计检验，提示人类检查数据管道 |
| 实验报告人类审核未通过 | 返回执行阶段补充分析，不传递到下游 |
| 多实验流量冲突 | 按优先级排队，低优先级实验暂停，标注"流量冲突" |
| 阶段总结生成失败 | 基于已完成的子Skill输出生成部分总结，缺失项标注"数据缺失"，不阻塞编排完成 |