slop-cleaner

AI Slop Cleaner

Remove patterns that LLMs typically introduce: unnecessary comments, over-abstraction, verbose code, dead code, and "AI verbal tics." The writer and the reviewer MUST be different passes — never self-approve a cleanup.

Rules

1. Scope first: identify which files to clean (args, or git diff --name-only HEAD~3)
2. Read before editing: always read the full file before making changes
3. One category at a time: go through the checklist in order, not all at once
4. Preserve behavior: cleaning must NOT change functionality — only remove noise
5. Run tests after: if tests exist, run them after cleanup to confirm nothing broke
6. Report: list every change made with before/after snippets

Slop Patterns Checklist

1. Unnecessary comments

• Comment restates the function/variable name: // Get the user above getUser()
• Section headers in small files: // --- Imports ---, // Constructor
• Obvious explanations: // Increment counter above counter++
• JSDoc/docstrings that only restate the signature with no added insight
• // TODO without context or assignee (either add context or remove)

2. Over-abstraction

• Single-use interfaces/types that could be inline
• Wrapper functions that only call one thing with no added logic
• Abstract factory / Strategy / Builder for something with one implementation
• Config objects for values that will never change
• Event emitters with a single listener

3. Redundant type assertions

• as Type where the type is already inferred by the compiler
• ! non-null assertions where the value is guaranteed non-null
• Explicit return types that match the inferred type and add no clarity
• String(x) or Number(x) where x is already that type

4. Excessive logging

• console.log / logger.debug on every function entry/exit
• Logging the same data that is already in the request/response
• Logging inside tight loops (performance hazard)
• Debug logs left from development (not guarded by env/level)