finding-duplicate-functions_skill

Overview

This skill helps audit a codebase for semantic duplicates: functions that do the same thing but have different names or implementations. It combines a classical extraction step with LLM-powered categorization and per-category duplicate detection. The goal is to surface consolidation opportunities and produce a prioritized report for human review.

How this skill works

First, the skill extracts a catalog of functions from source files with context using a shell extractor. Next, an LLM categorizes functions by intent into domains, then splits the catalog into category files. A stronger LLM inspects each category to identify function pairs or groups that implement the same intent and produces confidence-tagged duplicate sets. Finally, a report generator consolidates findings for human review and safe remediation.

When to use it

• Codebase has grown organically with multiple contributors (human or LLM).
• You suspect utility or validation functions were reimplemented in different places.
• Preparing for a large refactor and you want to minimize redundant code before changes.
• After running syntactic duplicate tools (e.g., jscpd) and wanting semantic-level detection.
• When auditing LLM-generated code where new functions are often created rather than reused.

Best practices

• Restrict extraction to exported functions and public methods to reduce noise.
• Run categorization before duplicate detection—compare only within intent categories.
• Only analyze categories with 3+ functions for efficiency and signal quality.
• Require tests for the chosen survivor function before deleting duplicates.
• Treat the LLM outputs as suggestions; perform human review and integration testing before changes.

Example use cases

• Find multiple implementations of date and string formatting utilities spread across a repo.
• Detect repeated API response shaping functions implemented differently per endpoint.
• Locate similar validation and error formatting logic in helper libraries.
• Consolidate path and filesystem helpers that have diverged over time.
• Prioritize cleanup before a library refactor or release to reduce maintenance burden.