Every other skill in this gallery asks whether a screen is well designed. This one asks whether the product is finished — you built the login, so where is the password reset, the account deletion, the empty state when filters return nothing.
Sixty-six pattern files across four sections: Sidekick (mobile app), Control Room (in-product web app), Storefront (marketing site), and Choreography (cross-screen flows). Nothing loads until it’s needed — the hub reasons from pattern names and one-liners alone, and only opens a reference file for what is actually being built right now.
What earns it a place is the tier system, and specifically the reason given for it. Every pattern is Fundamental, Common, or Conditional, fixing the starting assumption before any product-specific reasoning runs — because without it, “use judgment” has no anchor and drifts toward whichever way is less work. That is the most honest sentence about AI checklists anyone has written down.
The tiering itself is subtle. It explains why payment outranks cart: most products that charge money need some payment flow (Common), but only a subset need a multi-item cart on top of it (Conditional) — and it warns against asking someone to build a capability just to satisfy a pattern. Stage counts too: a pre-launch waitlist page isn’t missing a pricing page, and a two-person startup isn’t missing a careers page.
And it won’t nag. Once the person has confirmed an exclusion, don’t raise it again this session. One line, and it’s the difference between a check you keep installed and one you delete after the second run.
Watch for: the SKILL.md is stronger than the individual pattern files, some
of which run to ordinary checklist items. The judgment framing is the product;
the sixty-six references are the raw material it reasons over.
New and unproven — the repository is days old at the time of writing. The architecture and the reasoning are well past the quality bar; the mileage isn’t there yet.