AISlop CLI Scores AI-Generated Code Smells 0-100
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- aislop is a new MIT-licensed CLI that detects code 'smells' left by AI coding agents including Claude Code, Cursor, Codex, and OpenCode, flagging patterns like narrative comments above self-explanatory code, swallowed exceptions,
as anycasts, hallucinated imports, duplicated helpers, dead code, todo stubs, and oversized functions - The tool ships with 50+ rules across 8 language targets — TypeScript, JavaScript, Expo/React Native, Python, Go, Rust, Ruby, and PHP — and scores every change 0-100 in sub-second time
- The scanner is deterministic, with no LLM in the runtime path, so the same code in produces the same score out, the documentation states; repos mostly written in unsupported languages get
score: nullrather than a misleading number - aislop fix auto-fixes mechanical issues like formatters, unused imports, and dead code, with a
--safemode restricted to reversible changes; the aggressive-fflag handles dependencies and unused files - aislop agent creates a local git worktree, runs safe fixes, and hands remaining issues to an installed coding agent (Codex, Claude Code, or OpenCode) for repair — with no OpenAI or Anthropic API key required, since it uses the provider CLIs' existing local auth
- The tool supports CI integration with JSON and SARIF 2.1.0 output for GitHub code scanning, plus a quality-gate mode that blocks edits if the score regresses below a captured baseline
- Hook installation is available for Claude Code, Cursor, Gemini CLI, and pi, enabling per-edit scanning and immediate feedback flows after every agent edit
Why it matters: For teams shipping AI-assisted code, aislop provides a free, sub-second static-analysis layer that catches the structural decay AI agents leave behind — swallowed exceptions, narrative cruft, hallucinated imports — even when tests and linters pass. Because it's deterministic and LLM-free at runtime, it sidesteps the cost and latency of AI-on-AI review while still plugging into Codex, Claude Code, or OpenCode for the judgement calls via `aislop agent`.
Ask SkimNews
