Refactor research codebases with behavior-preserving, risk-tiered cleanup.
npx skills add https://github.com/UnaryLab/ai-for-research --skill code-clean
Please help me install this Agent Skill. Skill: code-clean Source: https://github.com/UnaryLab/ai-for-research/tree/main/skills/code-clean Command: npx skills add https://github.com/UnaryLab/ai-for-research --skill code-clean
Clean up an existing codebase safely by removing dead code, debug cruft, and style issues while preserving behavior and outputs.
Run behavior-preserving cleanup for the codebase on your branch by first capturing a baseline (tests or a golden outputs snapshot) and then applying safe formatting/lint fixes followed by evidence-backed dead-code removal.
High-intent search queries and answers about installing and using this skill.
To clean up research code without changing outputs, apply behavior-preserving refactoring using an explicit safety net like tests or golden snapshots, verifying after each batch and reverting any changes that alter numerical results.
The safest way to remove dead code and debug prints is evidence-based deletion using tool detection and reference checks, while keeping a deletion log and avoiding removal of entry points, public APIs, or dynamically dispatched code.
Yes, you need a pre-change safety net of tests or golden characterization snapshots before refactoring experiment code to capture a baseline, apply incremental verified edits, and prevent regressions during cleanup.
To extract hardcoded paths and hyperparameters into config for handoff, perform risk-tiered refactoring that treats any changed outputs as regressions to revert, ensuring identical behavior while tidying the codebase structure.
You should avoid automated linting and dead code removal when you lack a pre-change safety net of tests or golden snapshots, as conservative deletion requires baseline verification to prevent behavior regressions.
CANTAGESTUDIOIdentify missing tests and establish a minimal safety net before refactoring.
ys0714Remove build artifacts, dependencies, and secrets before packaging Node.js TypeScript projects for distribution.
MinhDuyDEVSimplify verified code by removing duplication and dead code without altering behavior.
jonnymuirStandardize project history files to record only final confirmed decisions.
