experiment-validation

Validate product feature hypotheses and produce Go/No-Go decisions from structured experiments.

Updated Aug 27, 2026
One-click install
npx skills add https://github.com/wuji-technology/pm-workflow-plugin --skill experiment-validation
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: experiment-validation
Source: https://github.com/wuji-technology/pm-workflow-plugin/tree/main/skills/experiment-validation
Command: npx skills add https://github.com/wuji-technology/pm-workflow-plugin --skill experiment-validation

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Validate product feature hypotheses end-to-end and produce defensible Go/No-Go decisions by structuring experiment planning, data collection, guardrail enforcement, multi-agent inquiry, and report generation into a repeatable workflow.

Core Features & Use Cases

  • Upstream reconciliation: automatically read and reconcile Design and Define artifacts to extract hypotheses, success/kill criteria, constraints, and risks.
  • Experiment planning & tracking: create prioritized experiment plans (L3-L4), confirm statistical and guardrail parameters, track progress, and persist working files for large contexts.
  • Data-led decisioning & inquiry: collect and summarize experimental data, run multi-agent advocate/challenger/observer inquiry sessions, and produce a validated report with traceable success-standard provenance.
  • Integration & persistence: optionally query Feishu project MCP workitems, persist interim _wip files, generate final validate-v{ver}-v{N}-{YYYY-MM-DD}.md, and write learnings to assets.

Quick Start

Start a validation run by specifying the product and feature and confirming resources so the skill can read Design/Define artifacts, collect experiment data, run the inquiry session, and generate the validate report.

Frequently Asked Questions about experiment-validation

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I validate product hypotheses through structured experiments?

Structured experiment validation works by reading upstream Design and Define artifacts to extract hypotheses, creating prioritized L3-L4 experiment plans, enforcing guardrail criteria, and generating a validated Go/No-Go report. This ensures feature-level validation decisions are defensible and traceable.

What's the best way to make Go/No-Go decisions for product features?

Make Go/No-Go decisions by collecting experimental data from A/B tests, prototype tests, or bench tests, then running a multi-agent advocate/challenger/observer inquiry session. This enforces success and kill criteria to produce a validated report with traceable provenance.

How do I enforce guardrail and kill criteria during A/B experiment data collection?

Enforce guardrail and kill criteria during A/B experiment data collection by applying the experiment plan's statistical and guardrail parameters. The workflow tracks progress, collects data, and validates results against these pre-defined constraints before producing a final report.

Can I use Feishu project MCP workitems for experiment design and tracking?

Yes, you can use Feishu project MCP workitems for experiment design and tracking. The validation workflow optionally queries Feishu project MCP to reconcile upstream artifacts and support data collection for L3-L4 verification processes.

Do I need upstream Design and Define artifacts to start hypothesis testing?

Yes, you need upstream Design and Define artifacts to start hypothesis testing. The validation process automatically reads and reconciles these artifacts to extract hypotheses, success criteria, constraints, and risks required for structured experiment planning.

How does multi-agent inquiry work for L3-L4 feature validation?

Multi-agent inquiry for L3-L4 feature validation works by running advocate, challenger, and observer agents to assess collected experimental data. This session challenges assumptions against guardrail metrics to ensure the Go/No-Go decision is robust and defensible.