CodeSmellGuard — AI Code Quality Gate

Catch architecture smells and bloated files before AI-generated code ships

Problem

A developer inherits a project written almost entirely by AI after a colleague leaves; a single source file exceeds a thousand lines with business logic and API layers tangled together. Tech leads are starting to ask: if developers fully accept AI output without reviewing design, who guards code quality? Many teams are already there, but lack a lightweight automated gate.

Solution

A CLI tool or Git hook that scans files generated or modified by AI before commit, detecting oversized files, mixed responsibilities, overly long functions, and missing abstraction layers. It outputs configurable warnings and fix suggestions, and only checks the current diff so it doesn't slow down full CI.

Users

Small teams and indie developers using Cursor, Copilot, Codex, and similar tools, especially tech leads inheriting AI-generated codebases

MVP

Support only JavaScript/TypeScript/Python; use git diff to check only changed files; detect file length, function length, mixed responsibilities, and cyclomatic complexity with AST plus simple rules; output colored terminal report with exit code; provide a .codesmell.yml config file

Revenue

Open-source core rule engine; subscription per seat for advanced rule packs and team dashboard ($5-8/seat/month), or one-time desktop license

Why now

AI coding tools make codebases bloat and architecture rot faster than ever, yet existing linters focus on syntax and style, not AI-specific structural issues; indie developers can move fast into this gap