Cal AI skills: ten steps from a consumer app idea to scale
A ten-skill set for building, monetizing, designing and distributing a consumer app — from the one obvious outcome to influencer operations, paywall conversi…
On this page
Cal AI skills: ten steps from a consumer app idea to scale
TL;DR
- The repository is a set of ten skills in a default order, taking a consumer app from idea to scale: one obvious outcome, distribution, influencer marketing, conversion, pricing, design trust, App Store demand and operating speed.
- Each skill is standalone; the sequence is a default rather than a requirement, so you load only the skill the current decision needs.
- The README's operating rules are the useful part: treat the numbers in the source posts as directional benchmarks, not universal laws, and validate against your own app, market, cohort and acquisition source.
- It is a synthesis of public posts by Jake Castillo about Cal AI and consumer-app growth — not an official Cal AI or Jake Castillo publication.
What it is and who it is for
Cal AI Skills is a practical skill set for building, monetizing, designing and distributing consumer apps. The README states its provenance plainly: the collection is based on the public posts supplied in the request by Jake Castillo (@jakecastilloooo) about Cal AI, consumer-app growth, influencer programs, paywalls, onboarding, Apple Search Ads and operating speed, and it is a synthesis rather than an official publication.
It is for the person who can build the app and cannot get it to grow: the same app, the same store listing, no acquisition. Each of the ten skills answers one decision in that chain rather than offering a general playbook.
The ten skills
| # | Skill | The decision it settles |
|---|---|---|
| 1 | viral-consumer-app | reduce the product to one obvious outcome and a demonstrable magic moment |
| 2 | consumer-app-distribution | pick a niche and an acquisition mix with distribution built in |
| 3 | influencer-niche-fit | decide whether influencer marketing fits before spending |
| 4 | influencer-20-second-test | qualify creators on views, comments and relationship quality |
| 5 | influencer-operations | turn creator partnerships into a repeatable operating system |
| 6 | conversion-funnel-paywall | build and diagnose onboarding and paywall conversion |
| 7 | trial-offer-architecture | choose access model, trial length and offer from activation and retention |
| 8 | app-design-trust | install a small, enforceable design system that signals trust |
| 9 | apple-search-ads | capture high-intent App Store demand and connect it to the funnel |
| 10 | operator-speed-systems | decide quickly, own outcomes and automate bottlenecks |
The order is a dependency chain, and that is the argument for reading it as one: a paywall decision made before the niche decision is a paywall built for the wrong audience.
The operating rules
The README's shared rules are the part that keeps the set from becoming a list of tips:
- Treat the numbers in the source posts as directional benchmarks, not universal laws — validate against the app, the market, the cohort and the acquisition source.
- Optimize for retained profitable subscribers, not downloads, views or trial-to-paid conversion in isolation.
- Test one structural variable at a time when learning matters.
- Keep the acquisition source attached to every cohort, and use revenue, payback, renewal and LTV for scale decisions.
- Prefer small, fast, reversible bets, and never let one experiment threaten the business.
- Preserve platform, legal, privacy, disclosure and accessibility requirements — a native-feeling marketing claim never justifies a deceptive one.
The second rule is the one most consumer-app advice breaks: a download number is a vanity metric when the cohort behind it cannot renew.
How to use it
Load only the skill for the current decision. The default path is the sequence above, and the README is explicit that it is a default and not a requirement — if the open question is the paywall, load conversion-funnel-paywall and skip the influencer set.
Each skill is standalone and sits in the repository as its own skill directory, so an agent can load one without the other nine. The source URLs and the topics they informed are mapped in references/source-map.md, which is where the attribution lives.
When to use it, when not to
Use it when you are about to spend money on acquisition, when a paywall is converting worse than expected and you do not know whether it is the offer or the onboarding, or when you need the question "which growth decision comes first" answered in order.
Do not use it as a benchmark source: the numbers are directional by the README's own rule, and they come from posts rather than from a study you can re-run. And do not use it as an official Cal AI publication — the README says it is a synthesis.
Sources
- Cal AI Skills — repository — the ten-skill sequence, the shared operating rules, attribution and synthesis note quoted above.
- README.md — skill sequence, operating rules, attribution.
Project page: Cal AI Skills
FAQ
Is this an official Cal AI publication? No. The README calls it a synthesis of public posts by Jake Castillo, not an official Cal AI or Jake Castillo publication.
Do I have to use the skills in order? No. The sequence is a default; the README says to load only the skill the current decision needs.
Can I rely on the numbers? Only as directional benchmarks. The operating rules say to validate each number against your app, market, cohort and acquisition source.
Which metric should the growth decisions optimize? Retained profitable subscribers, rather than downloads, views or trial-to-paid conversion in isolation.