Deep-dive review: aislop's 6-engine AI code smell scanner

[LAIguapi]

Last week I picked up a project where the code compiled and tests passed, but something felt off—comments longer than the code, empty catch blocks, unused imports everywhere. I tried aislop to scan for these "AI smell" patterns from source code.

Key findings:

• 6 parallel engines (AI patterns, format, lint, code quality, security, spell-check) using Promise.allSettled for concurrent scanning
• 40+ rules across 7 languages (TypeScript, Python, Go, Rust, Ruby, Java, C++) with deterministic scoring—no LLM dependency
• Narrative comment detection (270+ lines) distinguishes AI-generated explanations from legitimate docs (Go doc comments, Ruby YARD tags, Rust doc comments)
• Logarithmic decay scoring: individual rule penalties follow weight × log10(lines + 1) to prevent any single file from dominating
• Language-aware analysis: different comment syntax handling per language (e.g., Python # vs Java // vs Ruby #)

Installation hiccup: npm install failed in my environment (network restrictions), and the published tarball doesn't bundle dependencies. If your network is fine, npx aislop scan should work in one line.

Compared to alternatives: slop-scan (223★ TS), vibecheck (14★ TS), sloppylint (82★ Python), ai-gen-code-search (31★ Python)—aislop is the only one that doesn't require an LLM API key or GPU, making it the most practical option for CI/CD integration.

Full deep-dive review (Chinese): AI写的代码有没有味道？我试了个专门闻味的工具

Great work on this tool @heavykenny — the rule design, especially the narrative comment detection module, shows serious attention to detail. 👏