Growth that feeds itself, not your ad budget.
Describe your product and how people use it. Your assistant ranks five growth loop types for it, designs the strongest one step by step, estimates the loop coefficient, and lays out what to build in the next 90 days.
- ViralUsers’ content is shared outside and brings new users back.
- UsageUsers create, share, others consume and engage.
- CollaborationUsers invite colleagues to work together.
- User-generatedDiscover others’ work, make and share your own.
- ReferralInvite others for a reward, then invite again.
What it is
A loop design method
One instruction file, based on growth loops research by Ognjen Bošković. It takes a capable assistant through five steps: define product value, evaluate loop fit, design the mechanics, calculate the loop coefficient, and build the highest-leverage loop first.
The goal is compounding, product-native acquisition that reduces reliance on paid channels.
Who it is for
- Founders who want growth that outlasts ad spend
- Growth and product managers designing referral or invite features
- PLG teams looking for the most natural loop in their product
- Strategists analysing how competitors grow
Works well for
- Designing growth mechanisms for a product
- Building viral or referral traction
- Reducing reliance on paid acquisition
- Analysing competitor growth strategies
- Making a product more product-led
Not designed for
- Measuring your real loop rates; it estimates until you supply data
- Growth forecasts you can bank on
- Building or instrumenting the feature itself
- Paid media planning
What you provide, what you receive
You provide
- Product and its primary user action
- Target users and how they behave
- Existing sharing or collaboration features
- Current growth channels and metrics
- Constraints and opportunities
You receive
- All five loop types ranked for your product
- One primary loop with an implementation plan
- Secondary loops to layer later
- Metrics and a measurement framework
- A 30-60-90 day roadmap and loop coefficient estimates
Where it runs
Any AI assistant that can read a Skill file, including Claude Code and Codex. Download the free plaintext package from Finch, then install it with your assistant’s Skill method or ask it to read SKILL.md.
No extra API key, script or service. You still need access to an assistant model, which may have its own subscription or usage cost. The Skill does no web research of its own.
Limits and licence
Loop coefficients are only as good as the rates behind them. Until you measure invites and conversions, treat projections as scenarios, not forecasts.
MIT licence, Copyright (c) 2026 Pawel Huryn. Source: phuryn/pm-skills at commit 2a12295a7717, path pm-go-to-market/skills/growth-loops. The package adds a README, NOTICE, provenance note and a cover; the instructions are unchanged.
How it was checked
One Claude Code run on a fictional local B2B service produced all six parts the Skill promises. It ranked all five loop types, said plainly that viral and user-generated loops fit a physical local service poorly, designed a referral loop, and showed in three labelled scenarios that referral alone could not reach the company’s 100-office goal. The package downloaded back from Finch matched the audited files byte for byte. Native client installation, repeat runs and real growth results were not tested.