Give your AI agent the tools to clean up after itself — split PRs, catch smells, find duplicates, and refactor oversized files before you even review.
“Claude went crazy on this one. Can someone just review?”
“It's fine, the AI wrote it. Just approve.”
“I'm not reading all that. Looks good to me.”
Your agent calls these automatically after building a feature. Works with TypeScript, JavaScript, Python, Go, Rust, and more.
Analyzes your diff, maps dependencies between files, and suggests how to break it into focused, reviewable PRs.
Finds the junk AI loves to generate — deep nesting, 300-line functions, hallucinated imports, missing tests.
AST-based structural matching that catches copy-pasted code even when variable names differ.
Shows every declaration with line counts, dependency info, and imports. Your AI agent reads the JSON to plan refactors.
Moves symbols between files — copies exact source text, removes from the original, rewrites all imports across the codebase.
Your agent runs these commands, reads the output, and fixes what it finds — all before you open a PR.
Run npx skills add once, and Claude Code, Cursor, Codex, or any AI agent knows exactly when to call VibeRequest — automatically splitting PRs, fixing smells, deduplicating code, and refactoring oversized files.
Your agent builds the feature, then cleans up its own mess. You just review clean PRs.

viberequest smell and fix what it findsviberequest dedup and consolidate duplicatesviberequest inspect + extract to refactor oversized filesviberequest split and suggest a PR plannpm i -g viberequest
npx skills add peterkwitny/viberequest
It runs the tools, fixes what it finds, and splits the PR.
I vibe coded an entire feature in 30 minutes. It took 4 days to get the PR reviewed. Now I VibeRequest it first.
Vibe code all you want. Your agent cleans it up before the PR.
Open source. Runs locally. Works with any AI coding agent.