How do you fix component sprawl without adding another design system?
Start with ownership, not another component. Selective Intelligence generates a project index of directories, functions, hooks, components, UI primitives, duplicates, and competing exports; then it applies reuse → extend → extract → consolidate before creating anything new.
Signals that this is the problem
- Several buttons or fields perform the same job with slightly different props and styling.
- Cards, forms, dialogs, or layout divs have multiplied across routes.
- Developers cannot identify the canonical primitive or module owner.
- A new screen looks locally polished but does not match the rest of the product.
What Selective Intelligence does
- Refresh the repository’s generated directory, symbol, function, hook, component, and primitive index.
- Map every overlapping implementation to its callers, states, routes, and responsibility.
- Reuse the canonical owner, extend it, extract real shared behavior, or consolidate duplicates.
- Create a new abstraction only when its responsibility is genuinely different.
- Render and use the affected journeys at realistic sizes and states before calling the UI verified.
What proves improvement
- The intended routes use one defensible canonical owner for each shared responsibility.
- Duplicate behavior and bypassed primitives are removed or explicitly justified.
- The rendered experience remains consistent across normal, empty, loading, error, and mobile states.
Boundary: Reducing file count is not the goal. Unrelated responsibilities stay separate, and a generic abstraction is rejected when it would hide meaningful product differences.
Try this exact task
Install the public plugin, then paste this into ChatGPT or Codex with the work you want fixed:
Selective Intelligence: audit this interface for duplicate buttons, cards, fields, forms, and layout systems. Reuse or consolidate the right owners, then verify the affected user journeys.
The exact words Selective Intelligence, an unmistakable request for a named responsibility, or any correction, dissatisfaction, or failure feedback directly activates the canonical skill. The skill never widens permission to publish, spend, delete, deploy, or share.