deslop_skill

Overview

This skill removes AI-generated code slop from the current branch after code is written. It scans diffs against main and cleans up unnecessary comments, abnormal defensive checks, and inconsistent style introduced by AI assistants. The goal is a minimal, human-consistent codebase without changing intended behavior.

How this skill works

The tool compares the current branch to main and inspects changed files for patterns of AI-generated noise: verbose or redundant comments, atypical try/catch or guard blocks, casts to any and other type-workarounds, and style mismatches. It edits files to remove or consolidate these elements while preserving logic, then stages changes for review and produces a short summary of edits. It focuses on idiomatic style for the repository and avoids altering trusted validation or core logic unless clearly redundant.

When to use it

• After an AI assistant or copilot produced a substantial code diff in a branch.
• Before opening a pull request to ensure changes match repository style and conventions.
• When code contains excessive comments, obvious defensive scaffolding, or 'any' casts.
• During cleanup sprints where AI-assisted churn needs human sanitization.
• When you want a concise diff that reviewers can read quickly.

Best practices

• Run the skill on a branch with tests passing to prevent unintended behavior changes.
• Review each automated edit to verify no edge-case logic was removed.
• Keep the repo's style guide or linter active so edits follow repository conventions.
• Favor removing only noise and not refactoring unrelated code in the same commit.
• Include the short automated summary in the PR description for reviewer context.

Example use cases

• An AI generated a new feature but left verbose explanatory comments and extra try/catch blocks—use the skill to strip those out.
• A batch of TypeScript files contains multiple 'any' casts introduced to silence the type checker—use the skill to restore proper typing or remove unnecessary casts.
• A refactor added repetitive guard clauses around already-validated inputs—use the skill to remove redundant checks.
• Before sending code to reviewers, run the skill to produce a focused diff and a one-line change summary.
• During a code sweep after AI-assisted pair programming to align style and remove machine-added artifacts.