codebase-scanlisted

# Codebase Scan ## Overview Search before building. Every new implementation must be preceded by a thorough scan for existing code that already solves the problem, partially solves it, or establishes a pattern to follow. Skipping this step is the single most common cause of duplicate code, inconsistent patterns, and wasted effort. A 5-minute scan prevents hours of rework. ## When NOT to Use Skip when the task explicitly names all files to modify and no discovery is needed. ## Time Budget - **Focused scan** (known area): 1-2 minutes - **Broad exploration** (unfamiliar codebase): 3-5 minutes ## The Iron Law ``` NO NEW CODE WITHOUT SCANNING FOR EXISTING ALTERNATIVES FIRST ``` If you haven't run all 4 search steps, you cannot claim "nothing exists." Duplication is the most common implementation failure — a 5-minute scan prevents hours of rework. Without scanning first, agents frequently duplicate existing functionality, creating maintenance burden and inconsistency. The scan also reveals naming conventions, architectural patterns, and reusable utilities that improve implementation quality. Duplication is the single most common form of generative debt. ## The Process ### Step 1: Search by Name Search for the exact name or close variations of what you plan to create. ``` Grep: class FeatureName, def feature_name, FeatureNameService Glob: **/*feature_name* ``` If found, stop. Use or extend the existing implementation. ### Step 2: Search by Concept Search for synony