Use when adding new error messages to React, or seeing "unknown error code" warnings.
npx skills add DonggangChen/antigravity-agentic-skills --skill "hypothesis_generation"
Install specific skill from multi-skill repository
# Description
Scientific hypothesis generation, experiment design and test methodology guide.
# SKILL.md
name: hypothesis_generation
router_kit: FullStackKit
description: Scientific hypothesis generation, experiment design and test methodology guide.
metadata:
skillport:
category: research
tags: [architecture, automation, best practices, clean code, coding, collaboration, compliance, debugging, design patterns, development, documentation, efficiency, git, hypothesis generation, optimization, productivity, programming, project management, quality assurance, refactoring, software engineering, standards, testing, utilities, version control, workflow]
π¬ Hypothesis Generation
Scientific hypothesis generation and test methodology guide.
π Hypothesis Structure
Format
IF [independent variable/action]
THEN [dependent variable/outcome]
BECAUSE [mechanism/reasoning]
Example
IF we reduce checkout steps from 5 to 3
THEN conversion rate will increase by 15%
BECAUSE fewer steps reduce friction and drop-off
π― Hypothesis Criteria
| Criteria | Description |
|---|---|
| Specific | Clear and unambiguous |
| Measurable | Measurable outcome |
| Testable | Testable |
| Falsifiable | Falsifiable |
| Relevant | Aligned with business goal |
π§ Hypothesis Types
A/B Test Hypothesis
**Hypothesis:** Changing CTA button from blue to green
will increase click rate by 10%
**Metric:** CTA Click Rate
**Baseline:** 2.5%
**Target:** 2.75%
**Sample Size:** 10,000 users
**Duration:** 2 weeks
Product Hypothesis
**Problem:** Users abandon during onboarding
**Hypothesis:** Adding progress indicator will reduce
abandonment by 20%
**Success Metric:** Onboarding completion rate
π Experiment Design
Test Plan
## Experiment: [Name]
### Hypothesis
[IF-THEN-BECAUSE statement]
### Variables
- Independent: [What we change]
- Dependent: [What we measure]
- Control: [What stays same]
### Metrics
- Primary: [Main KPI]
- Secondary: [Supporting metrics]
- Guardrail: [Safety metrics]
### Design
- Type: A/B / Multivariate
- Split: 50/50
- Duration: [X] weeks
### Analysis Plan
- Statistical test: [t-test, chi-square, etc.]
- Confidence level: 95%
- MDE: [Minimum detectable effect]
π Prioritization (ICE)
| Hypothesis | Impact | Confidence | Ease | Score |
|---|---|---|---|---|
| H1 | 8 | 7 | 6 | 7.0 |
| H2 | 9 | 5 | 4 | 6.0 |
| H3 | 6 | 8 | 9 | 7.7 |
ICE Score = (Impact + Confidence + Ease) / 3
Hypothesis Generation v1.1 - Enhanced
π Workflow
Phase 1: Observation
- [ ] Data: Capture an "Insight" from analytics data or user interviews.
- [ ] Problem: Convert observation into a clear problem statement.
Phase 2: Construction
- [ ] Formula: Use IF [action] THEN [outcome] BECAUSE [reason] template.
- [ ] Variables: Clarify independent (changing) and dependent (measured) variables.
Phase 3: Prioritization
- [ ] ICE Score: Score Impact, Confidence, Ease from 1-10.
- [ ] Risk: What do we lose if the test fails?
Checkpoints
| Phase | Verification |
|---|---|
| 1 | Is the hypothesis falsifiable? (If always true, it's not a hypothesis) |
| 2 | Is the result a measurable metric (Click rate, Retention)? |
| 3 | Is the "Because" part based on a logical user behavior? |
# Supported AI Coding Agents
This skill is compatible with the SKILL.md standard and works with all major AI coding agents:
Learn more about the SKILL.md standard and how to use these skills with your preferred AI coding agent.