SiteDev Utilities & Vibe-Coding Tooling · Codebase hygiene
Open benchmark that ranks AI coding models by how much code their SWE-bench patches add, counting net syntax units instead of lines, with an upload-your-own-results chart.
Why it’s a gem
The only leaderboard that punishes models for bloating your codebase.
Decompiles Sierra SCI adventure games, abstract-interprets the scripts to find unwinnable states, then derives, verifies and compiles guard patches.
Codemod CLI that strips unused variables, imports and parameters from source files while preserving surrounding formatting and comments.
Markdown convention plus linter for AI-written project docs, catching stale versions and fields duplicated across files before agents read them back as context.
Converts components between MUI, Chakra, Ant Design, Mantine, EUI, Bootstrap and Quasar using 219 hand-verified mappings, running entirely in the browser.
Opinionated Oxlint plugin with rules that reject AI-slop TypeScript patterns like chained type assertions, unknown parameters and ad hoc typeof narrowing.
Agent skill that proves accidental complexity before deleting it, tracking consumers, contracts, and verification so simplifications stay safe.