LLM-generated codebases have a nasty habit of reimplementing the same utility function six different ways instead of reusing what exists. This skill catches those semantic duplicates that traditional copy-paste detectors miss. It extracts your functions, uses Haiku to group them by domain, then unleashes Opus on each category to find functions that do the same thing despite different implementations. The two-tier model approach is smart: cheap categorization, expensive analysis only where it matters. Focus on your utils folders and validation code first, those accumulate duplicates fastest. Just make sure the survivor function has tests before you consolidate, because the LLM can miss edge cases that only one implementation handled.
npx skills add https://github.com/obra/superpowers-lab --skill finding-duplicate-functions