More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    vibe code cleanup

    1

    Systematically audit and refactor AI-generated codebases through phased, behavior-preserving cleanup steps.

    $9.99

    /mo

    Subscription · cancel anytime

    Secure checkout via Stripe

    • Always the latest version
    • Delivered live to your agent via MCP
    • Cancel anytime, access ends at period end
    Security scanned
    vibe code cleanup

    vibe code cleanup

    Example session with this skill installed

    My AI-built React app is a mess and I can't add new features without breaking things. Can you audit the code and clean it up?

    • Read your context and instructions
    • Compiled the vibe code cleanup
    • Generated the document

    I have completed the reconnaissance and initial audit of the repository. The project is a Next.js application with several instances of redundant state logic and inconsistent styling patterns typical of AI-generated code.

    Audit Report: AI-Generated Smells

    LocationSeveritySmell TypeRecommended Fix
    src/components/UserDashboard.tsxCriticalDuplicate State SyncConsolidate 3 useEffect hooks into a single state transition.
    package.jsonMajorHallucinated DepsRemove react-query-v5-adapter (non-existent package).
    src/lib/api.tsMajorHardcoded SecretsMove OPENAI_API_KEY from code to .env.local.
    src/hooks/useAuth.tsMinorVerbose LogicRefactor 45-line switch statement into a lookup object.
    styles/globals.cssMinorCSS RedundancyRemove 12 unused utility classes generated by repetitive prompting.

    Executive Summary: The codebase is functional but fragile. The primary issues are duplicate logic in the dashboard and security risks involving hardcoded keys. Fixing these will reduce the build error count from 14 to 0.

    Proposed Cleanup Phases:

    1. Phase 1: Security & Hygiene: Move secrets to environment variables and remove dead dependencies.
    2. Phase 2: Logic Consolidation: Refactor the dashboard state management to prevent race conditions.
    3. Phase 3: Type Safety: Resolve the 14 TypeScript errors in the auth and API layers.

    Next steps

    • Approve the three cleanup phases listed above.
    • Confirm if the react-query-v5-adapter was intended for a specific internal tool.
    • Provide the .env.example values for the hardcoded secrets.

    UserDashboard.tsx

    PDF · document

    Generated

    Example file from a real run - the skill writes it into your workspace.

    $9.99/mo

    Subscription · always the latest version

    • Secure checkout via Stripe
    • Cancel anytime
    • Delivered live via MCP

    About this skill

    Inheriting a codebase built entirely by LLMs often means dealing with a "vibe-coded" mess of inconsistent patterns, dead code, and fragile logic. This skill provides a systematic framework for developers to audit, plan, and execute a behavior-preserving cleanup of AI-generated projects. It moves beyond simple linting by identifying specific "AI smells"—like hallucinated library usage or redundant state logic—and fixing them in verifiable phases.

    What it does

    • Codebase Reconnaissance maps the stack, dependencies, and file scale to establish a technical baseline.
    • AI-Smell Auditing scans for specific patterns typical of AI generation, ranking them by severity from minor to critical.
    • Phased Refactoring breaks down the cleanup into hygiene, typing, structure, and state management phases.
    • Automated Verification runs builds and typechecks after every phase to ensure no regressions occur.
    • Final Guardrails provides a before/after metrics report and recommendations for CI/CD and linting rules.

    How it works

    1. Baseline: Run the existing project to record current type errors, lint warnings, and build status.
    2. Audit: Generate a severity-ranked report identifying specific AI-generated technical debt.
    3. Plan: Review the proposed 3-6 phase cleanup plan and approve the scope.
    4. Execute: The agent performs one phase at a time, verifying the build between each session.
    5. Finalize: Receive a comparative report showing improved metrics and future maintenance guidelines.

    Frameworks & tools

    Works with any standard web or backend stack including React, Next.js, Node.js, and Python. It utilizes existing project tooling like TypeScript, ESLint, and Vitest/Jest for verification gates.

    Why this beats prompting it yourself

    Generic cleanup prompts often lead to massive, breaking changes that are impossible to debug. This skill enforces a strict "one phase, one verification" rule and uses a specialized catalog of AI-specific code smells that general-purpose prompts usually miss.

    Use cases

    • MVP Handover: Cleaning up a prototype built with AI before hiring a full-time engineering team.
    • Technical Debt Triage: Fixing a project that has become too buggy to add new features safely.
    • Security Hardening: Identifying and removing hardcoded secrets or unsafe patterns left by rapid AI prompting.

    Known limitations

    Does not perform framework migrations or architectural rewrites. Requires a functional build or test command to establish a baseline.

    Requires existing build/test scripts to establish a baseline. Does not support full framework migrations or repo restructuring without explicit user requests.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 11 days ago

    Permissions

    Read Files
    Write Files
    Listed11 days ago

    What's inside

    Frequently Asked Questions

    Unlike generic cleanup prompts that often break dependencies, this skill uses a phased approach that verifies the build after every single change to ensure your application remains functional throughout the entire refactoring process.

    This skill is designed for any environment where an agent has filesystem access, and it leverages your project's existing TypeScript, ESLint, or Vitest configurations to establish its verification gates.

    The skill identifies specific 'AI smells' such as hallucinated API methods, redundant logic loops, inconsistent variable naming conventions, and fragile state management patterns typical of rapid LLM prompting.

    You receive the full phased auditing framework, the specialized catalog of AI-specific code smells, and the automated verification sequence used to manage the refactoring sessions.

    The skill requires that your project has a functional build or test command; it cannot refactor a codebase that is currently in a non-runnable state or lacks a basic build script.

    No, the skill focuses on 'behavior-preserving' cleanup, meaning it improves code quality, readability, and type safety without altering the existing features or end-user functionality of your app.

    New in Code Quality & Review