growth-engineering — Claude Code skill: what it does, how to install
growth-engineering is a Claude Code skill with architecture patterns for feature flags, A/B testing, PLG funnels, and referral tracking.
MIT 133 installs ★ 114 updated 2026-10-25
npx skills add travisjneuman/.claude --skill growth-engineering -a claude-codeWhat growth-engineering does
This skill supplies your coding assistant with concrete TypeScript patterns for feature toggling, experimentation schemas, and product-led growth metrics. It lays down specific guidelines for flags: retiring experiment toggles within two weeks, avoiding nested flag evaluations, and conducting monthly cleanups.
Experiment setup includes type definitions for audience allocation (weights summing to 100), exposure tracking via analytics events, and statistical guardrails like defining sample sizes in advance. It also covers referral funnel stages and benchmark targets for core metrics like trial-to-paid and viral coefficient.
Running the provided flag evaluation code requires the posthog-node library and an active POSTHOG_API_KEY environment variable.
What it helps with
- implement type-safe feature flag wrappers in TypeScript
- track experiment exposure events for A/B testing
- structure referral steps and PLG activation funnels
- evaluate experimentation tools like Growthbook, Statsig, and PostHog
Example prompts for the agent
- Create a typed feature flag utility using PostHog in TypeScript
- Define an experiment configuration interface with weighted variants
- Write a tracker function for referral funnel lifecycle states
How to install growth-engineering
1Skills CLI (Vercel)
npx skills add travisjneuman/.claude --skill growth-engineering -a claude-code2Manually: put the files in place
~/.claude/skills/growth-engineering/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/growth-engineering/
1Skills CLI (Vercel)
npx skills add travisjneuman/.claude --skill growth-engineering -a codex2Manually: put the files in place
.agents/skills/growth-engineering/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add travisjneuman/.claude --skill growth-engineering -a cursor2Manually: put the files in place
.cursor/skills/growth-engineering/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install travisjneuman/.claude skills/growth-engineering2Manually: put the files in place
.github/skills/growth-engineering/SKILL.md1Skills CLI (Vercel)
npx skills add travisjneuman/.claude --skill growth-engineering -a gemini-cli2Platform CLI
gemini skills install https://github.com/travisjneuman/.claude --path skills/growth-engineeringManually: put the files in place
.devin/skills/growth-engineering/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add travisjneuman/.claude --skill growth-engineering -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about growth-engineering
How to install growth-engineering in Claude Code?
Run npx skills add travisjneuman/.claude --skill growth-engineering -a claude-code in your terminal.
What does the growth-engineering skill provide?
It gives the AI assistant reference designs for feature flags, split test tracking, PLG metrics, and comparisons across experimentation platforms.
Which platforms are covered for feature toggles and experiments?
The skill outlines use cases for PostHog, LaunchDarkly, Statsig, Growthbook, and Optimizely.
Source text
SKILL.md · MIT
---
name: growth-engineering
description: >-
A/B testing infrastructure, feature flags (LaunchDarkly, Unleash), experimentation platforms, PLG patterns, and funnel optimization. Use when building experimentation systems, implementing feature toggles, or optimizing conversion funnels.
---
# Growth Engineering Skill
Infrastructure and patterns for product-led growth, experimentation, and conversion optimization.
---
## Feature Flag Systems
### Implementation Pattern
```typescript
// lib/feature-flags.ts
import { PostHog } from 'posthog-node';
const posthog = new PostHog(process.env.POSTHOG_API_KEY!);
interface FeatureFlags {
'new-onboarding-flow': boolean;
'pricing-experiment': 'control' | 'variant-a' | 'variant-b';
'ai-suggestions': boolean;
}
export async function getFlag<K extends keyof FeatureFlags>(
key: K,
userId: string,
): Promise<FeatureFlags[K]> {
const value = await posthog.getFeatureFlag(key, userId);
return value as FeatureFlags[K];
}
// Usage in component
const showNewOnboarding = await getFlag('new-onboarding-flow', user.id);
```
### Feature Flag Best Practices
- Short-lived flags: Remove after experiment concludes (< 2 weeks)
- Long-lived flags: Ops toggles for gradual rollouts, kill switches
- Never nest feature flags (creates exponential complexity)
- Clean up stale flags monthly
- Log flag evaluations for debugging
---
## A/B Testing Infrastructure
### Experiment Design
```typescript
// lib/experiments.ts
interface Experiment {
id: string;
name: string;
variants: {
id: string;
weight: number; // 0-100, must sum to 100
}[];
targetAudience: {
percentage: number; // % of users included
filters?: Record<string, unknown>;
};
primaryMetric: string;
secondaryMetrics: string[];
minimumSampleSize: number;
startDate: Date;
endDate?: Date;
}
// Track experiment exposure
function trackExposure(experimentId: string, variantId: string, userId: string) {
analytics.capture({
event: '$experiment_started',
distinctId: userId,
properties: {
$experiment_id: experimentId,
$variant_id: variantId,
},
});
}
```
### Statistical Significance
- Minimum sample size: Calculate before starting (use Evan Miller calculator)
- Don't peek: Set duration upfront, don't stop early on promising results
- Sequential testing: Use if you must check early (adjusts p-values)
- Minimum detectable effect: Define what improvement matters (e.g., 5% lift)
---
## Product-Led Growth Patterns
### Activation Metrics
| Stage | Metric | Example |
|-------|--------|---------|
| Sign up | Registration complete | User creates account |
| Setup | Profile complete | Fills required fields |
| Aha moment | Core value experienced | Creates first project |
| Habit | Repeated engagement | 3 sessions in first week |
| Revenue | Conversion to paid | Subscribes to plan |
### Viral Loops
```typescript
// Referral system pattern
interface Referral {
referrerId: string;
referredEmail: string;
status: 'pending' | 'signed_up' | 'activated' | 'converted';
rewardGranted: boolean;
}
// Track referral funnel
function trackReferralStep(referralId: string, step: Referral['status']) {
analytics.capture({
event: 'referral_step',
properties: { referralId, step },
});
}
```
### Conversion Optimization
- Reduce friction: Minimize form fields, enable social login
- Social proof: Show user counts, testimonials, logos
- Urgency: Trial countdown, limited-time offers (use sparingly)
- Value demonstration: Interactive demos, free tier with clear upgrade path
- Personalization: Onboarding flow based on use case selection
---
## Growth Metrics
| Metric | Formula | Target |
|--------|---------|--------|
| Activation rate | Activated / Signed up | > 40% |
| Trial-to-paid | Paid / Trial started | > 15% |
| Net revenue retention | (Start MRR + Expansion - Contraction - Churn) / Start MRR | > 110% |
| Viral coefficient | Invites sent * Conversion rate | > 0.5 |
| Time to value | Median time from signup to aha moment | < 5 min |
| DAU/MAU ratio | Daily active / Monthly active | > 20% |
---
## Experimentation Platforms
| Platform | Type | Best For |
|----------|------|----------|
| PostHog | Self-hosted/cloud | Full-stack, open source |
| LaunchDarkly | Cloud | Feature flags at scale |
| Statsig | Cloud | Auto-stats, warehouse-native |
| Growthbook | Self-hosted/cloud | Open source, Bayesian stats |
| Optimizely | Cloud | Enterprise, multi-channel |
---
## Related Resources
- `~/.claude/skills/product-analytics/SKILL.md` - Analytics and tracking
- `agent-product-analytics-specialist` skill - Analytics specialist
- `~/.claude/skills/authentication-patterns/SKILL.md` - Auth for PLG
---
_Measure everything. Experiment constantly. Remove what doesn't work._
© travisjneuman · MIT · travisjneuman/.claude