ts-search-first

name: ts-search-first description: "Before writing any new function, hook, type, or component, search the codebase for existing implementations to reuse." metadata: short-description: Codebase-first (reuse before creation)

Why this exists

Claude defaults to generating fresh implementations even when existing code is a 90% match. This skill forces a search-first habit to catch reuse opportunities before writing anything.

Procedure

1. Never start by coding. Start by locating what already exists.
2. Search the codebase by:
   • Probable names (camelCase, PascalCase, kebab-case)
   • Types/interfaces/export names
   • Patterns (useXxx, XxxProvider, getXxx, buildXxx, createXxx, updateXxx)
   • Routes/paths (backend) and query keys (frontend)
3. Use fast searches (e.g. ripgrep):
   • rg "use[A-Z]" src
   • rg "function <name>|const <name> =" src
   • rg "<TypeName>" src
   • rg "<route-path>" src
4. If found:
   • Propose direct reuse, or minimal extension
   • Respect local conventions (folders, naming, exports)
5. If not found:
   • Create the smallest possible solution, consistent with existing architecture

Output

• Short report: "Found / not found" + files examined.
• Then the proposal: reuse (preferred) or minimal implementation.

Composability

• Run coding-convention after extending found code to verify naming/file placement.
• Run regression-check after modifying existing code to catch broken consumers.

Gotchas

• When creating a new oRPC contract/schema, also search for reusable utility schemas (pagination, search, sort, userId filters) — not just routes. Missing these leads to duplicated schemas discovered only at review time.
• Search for aliased re-exports — exact name matches miss code that re-exports under a different name.
• In monorepos, search all workspace packages, not just src/. Duplicate utilities across packages are common.