Skip to main content
Glama
abdelrhmanahmed255

Frontend-Quality

by abdelrhmanahmed255
OverviewSchemaSourceRelated ServersDiscussionsAdmin
JavaScript
Local

Frontend Quality Auditor

A CLI tool and AI agent skill that audits your frontend projects — finds dead code, unused deps, accessibility issues, and responsive layout bugs. Then it generates a clear report and can auto-fix the safe stuff.

Built this because I got tired of shipping code with leftover console.log statements, 3MB hero images, and packages sitting in package.json that nobody actually imports. This tool catches all of that.

Install

As an AI Agent Skill

Option 1: Using the MCP Server (Recommended for Claude Desktop, Cursor, etc.) You can plug this tool directly into any AI that supports the Model Context Protocol (MCP) without cloning anything! Just add this to your AI's MCP settings file:

{
  "mcpServers": {
    "frontend-auditor": {
      "command": "npx",
      "args": ["-y", "frontend-quality-auditor", "mcp"]
    }
  }
}

Option 2: Manual Clone If your agent doesn't support MCP yet, clone it into your agent's skills folder:

# Gemini / Antigravity
git clone https://github.com/abdelrhmanahmed255/Frontend-Quality.git ~/.gemini/skills/frontend-quality-auditor

# Claude Code
git clone https://github.com/abdelrhmanahmed255/Frontend-Quality.git ~/.claude/skills/frontend-quality-auditor

# Cursor
git clone https://github.com/abdelrhmanahmed255/Frontend-Quality.git ~/.cursor/skills/frontend-quality-auditor

Install the CLI's dependencies once inside the cloned folder:

cd ~/.claude/skills/frontend-quality-auditor   # or the folder you cloned into
npm install

Then just tell your agent something like:

"Audit my project for code quality and responsive issues"

As a CLI

# run directly without installing
npx frontend-quality-auditor audit .

# or install globally
npm install -g frontend-quality-auditor

As a Project Dependency (Recommended)

To run audits in CI or as a regular check before committing:

npm install --save-dev frontend-quality-auditor

Then add scripts to your package.json:

{
  "scripts": {
    "audit": "frontend-quality-auditor audit .",
    "audit:fix": "frontend-quality-auditor fix --safe",
    "audit:ci": "frontend-quality-auditor audit . --strict --format=sarif"
  }
}

Run them using:

npm run audit
npm run audit:fix

Related MCP server: MCP Frontend Analyzer

Usage

# Full audit — prints to terminal + saves a markdown report
frontend-auditor audit

# Same audit but on a specific directory
frontend-auditor audit ./my-app

# Get suggestions without touching any files
frontend-auditor suggest

# Auto-fix safe issues (removes console.log, debugger, etc.)
frontend-auditor fix --safe

# Exit with code 1 when there is any P0 or P1 finding (useful as a CI gate)
frontend-auditor audit --fail-on=P1

The audit generates a markdown report at audit-reports/FRONTEND_AUDIT_REPORT.md with severity ratings, quick wins, and file-level evidence.

Options

Option

What it does

--format=json

Print the report as JSON and save it to audit-reports/FRONTEND_AUDIT_REPORT.json

--format=markdown

Only write the markdown report (no terminal report)

--output=<path>

Write the report file to a custom path

--verbose

Show every finding in the terminal instead of the top 15

--url=<url>

Also run the browser audit against a running app (see below)

--dry-run

With fix --safe: show what would change without touching files

--strict[=<score>]

Exit with code 1 when the health score is below <score> (100 if omitted)

Browser audit

Start your dev server, then pass its URL:

npx playwright install chromium   # one-time browser download
frontend-auditor audit . --url=http://localhost:3000

The page is loaded in headless Chromium at 8 widths (320px to 1440px) and checked for horizontal overflow (with the elements causing it), images without alt, and small touch targets on mobile.

CI

--strict turns the health score into a pass/fail check:

# fail the job if the health score drops below 80
npx frontend-quality-auditor audit . --strict=80

The health score starts at 100 and loses 20 points per P0, 10 per P1, 5 per P2, and 2 per P3 finding.

What It Checks

Category

What it looks for

Code Quality

console.log, debugger, hardcoded localhost URLs, hardcoded API keys and tokens, empty click handlers, monster components (300+ lines)

Imports

Unused imports that are imported but never referenced in the file

Dependencies

Packages in package.json that are never imported anywhere in your code

Assets

Images over 500KB, old formats (PNG/JPG) that should be WebP/AVIF, unreferenced files

Responsive

Horizontal scroll overflow, broken grids, elements bleeding outside viewport

Accessibility

Missing alt tags, unlabeled form fields, icon-only buttons and links without an accessible name, missing <h1> or skipped heading levels, missing <html lang>, small touch targets (<44px)

Every finding gets a severity level:

  • P0 — breaks the page, fix now

  • P1 — major problem, fix before shipping

  • P2 — noticeable issue, fix when you can

  • P3 — minor cleanup, nice to have

GitHub code scanning

--format=sarif writes audit-reports/FRONTEND_AUDIT_REPORT.sarif, which GitHub can show as code scanning alerts and inline annotations on pull requests:

- run: npx frontend-quality-auditor audit . --format=sarif
- uses: github/codeql-action/upload-sarif@v3
  with:
    sarif_file: audit-reports/FRONTEND_AUDIT_REPORT.sarif
    category: frontend-quality

The job needs permissions: security-events: write for the upload. P0/P1 findings become errors, P2 warnings, and P3 notes. Browser findings from --url have no source file, so they are only in the other report formats.

Config

Drop a config/auditor.config.json in your project root to customize thresholds:

{
  "viewports": [375, 768, 1280],
  "audit": {
    "codeQuality": true,
    "dependencies": true,
    "assets": true
  },
  "rules": {
    "maxAssetSizeKB": 500,
    "maxComponentLines": 300,
    "disallowConsole": true,
    "disallowDebugger": true,
    "checkUnusedImports": true,
    "checkUnusedDeps": true
  },
  "reporting": {
    "outputDir": "./audit-reports"
  }
}

Option

Effect

viewports

Widths (px) tested by the browser audit (--url). Defaults to 320–1440.

audit.codeQuality / audit.dependencies / audit.assets

Set to false to skip that whole category.

rules.maxAssetSizeKB

Images above this size are reported as oversized.

rules.maxComponentLines

Files longer than this are reported as monster components.

rules.disallowConsole / rules.disallowDebugger

Set to false to stop reporting (and auto-fixing) console.* or debugger.

rules.checkUnusedImports / rules.checkUnusedDeps

Set to false to skip unused import or unused dependency detection.

reporting.outputDir

Where report files are written, relative to the audited project.

Any option you leave out keeps its default, and if no config file exists the defaults are used.

Supported Frameworks

Auto-detects your setup:

  • Next.js (App Router & Pages Router)

  • React + Vite / CRA

  • Remix

  • Static HTML/CSS

  • Tailwind CSS

  • TypeScript

Project Structure

Frontend-Quality/
├── .github/workflows/
│   └── npm-publish.yml          # Publishes to npm on v* tags
├── bin/
│   └── auditor.mjs              # CLI entrypoint
├── config/
│   └── auditor.config.json      # Default configuration
├── rules/                       # Reference docs for each audit category
│   ├── accessibility.md
│   ├── assets.md
│   ├── code-quality.md
│   ├── dependencies.md
│   ├── performance.md
│   ├── responsive.md
│   └── ux.md
├── src/
│   ├── index.mjs                # Main orchestrator
│   ├── config-loader.mjs        # Reads and merges config
│   ├── analyzers/
│   │   ├── project-detector.mjs # Detects framework, TS, Tailwind
│   │   ├── route-discovery.mjs  # Finds all routes
│   │   ├── import-analyzer.mjs  # Unused import detection
│   │   ├── dependency-analyzer.mjs
│   │   ├── asset-analyzer.mjs
│   │   ├── code-analyzer.mjs    # console.log, debugger, etc.
│   │   ├── browser-analyzer.mjs # DOM checks that run inside the page
│   │   └── run-browser-audit.mjs # Drives Playwright across viewports (--url)
│   ├── reporters/
│   │   ├── terminal.mjs
│   │   ├── markdown.mjs
│   │   └── json.mjs
│   ├── fixes/
│   │   ├── safe-fixes.mjs       # Removes console.log, debugger, unused imports
│   │   └── approval-required.mjs
│   └── utils/
│       └── safe-string-search.mjs # Blanks strings/comments before matching code
├── assets/
│   └── banner.jpg
├── CHANGELOG.md
├── CONTRIBUTING.md
├── SKILL.md                     # Agent skill specification
├── package.json
├── LICENSE
└── README.md

Contributing

Got a bug or feature idea? Open an issue first so we can talk about it, then submit a PR. Keep it simple.

License

MIT © Abdelrhman Ahmed

This server cannot be deployed

One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

Maintenance

ActivityMaintained
ResponsivenessNo issues

Resources

Related MCP Connectors

Related MCP Servers