semantic-duplication-scan

Semantic Duplication Scan

Overview

Detect business logic that has been reimplemented multiple times across different architectural layers. Unlike linters (which detect syntactic similarity) or domain-review (which catches single-instance layer violations), this skill detects semantic equivalence — the same domain calculation independently appearing in domain services, client adapters, and presentation components with different variable names and structure.

Annotation Prompt Version

promptVersion: 1.0

When this version changes, any register entry with a different promptVersion is treated as stale and re-annotated on the next scan pass that touches that file.

Pre-Filter Rules

Apply before any LLM call. No model invocation at this stage.

Trivial Function Definition

A function is trivial — and must be excluded from the register — if it meets ALL of the following:

• Contains no arithmetic operators: +, -, *, /, %, **
• Contains no boolean logic operators: &&, ||, !, not, and, or
• Contains no branching constructs: if, else, switch, case, ternary (?:), match
• Contains no assignments to variables outside its own scope (no external state mutation)
• Contains no calls to higher-order collection operations: map, filter, reduce, flatMap, forEach, find, some, every, or language equivalents

Trivial patterns (always excluded):

• Getters: read and return a field with no transformation
• Pass-through delegators: call one function with the same arguments, return the result unchanged
• Identity functions: return the input unchanged
• Constructors / initializers that only assign parameters to instance fields

If a file contains only trivial functions, output:

No computation units found to analyze

and do not create or modify the register.

File Exclusion Patterns

Exclude the following from annotation regardless of content:

*.test.*
*.spec.*
__tests__/
*.test-d.*
*.generated.*
*.pb.*
*.d.ts
dist/
build/
.next/
coverage/