Use stable naming conventions that match repo/tooling expectations and avoid identifier collisions. **Rules** 1. **Keep script names consistent across packages**: if CI/tooling expects specific npm/pnpm script keys, use the same keys in every package (don’t rename per-package). 2. **Ensure package/example names are globally unique**: every published/package identifier (e.g., `package.json#name`) must be unique across the repo to prevent collisions. 3. **Name compatibility boundaries explicitly**: when an experimental change affects behavior, reflect compatibility intent in naming (e.g., expose a `*-core` for the breaking surface and a separate `*` wrapper that preserves older behavior), instead of changing semantics under the same name. **Example (package.json)** ```json { "name": "@tanstack/lit-query", "type": "module", "scripts": { "test:types": "tsc --noEmit", "typecheck": "pnpm run test:types", "build": "pnpm run build:deps && pnpm run build:esm && pnpm run build:cjs" } } ``` And if you need a compatibility wrapper vs breaking core: - `@scope/lib-core` (breaking) - `@scope/lib` (wrapper preserving old behavior) This prevents CI from skipping packages, avoids naming collisions, and makes breaking/compatibility intent obvious from identifiers.