ContextKit

Score, generate, and manage AI coding configs.

CLI + web tools for Claude Code, Cursor, GitHub Copilot, and Gemini CLI. Score your CLAUDE.md from the terminal or generate production-ready configs in seconds.

Try the CLI · Web Generator · Web Analyzer · Website

---

CLI

Score your CLAUDE.md from the terminal. Zero dependencies, instant results.

npx contextkit score

─────────────────────────────────────────
  CLAUDE.md Score  8/10 (Good)
  ./CLAUDE.md
─────────────────────────────────────────

  Solid config with room for improvement.

  Categories

  Structure      ████████████████ 2/2
  Architecture   ████████████████ 2/2
  Conventions    ████████████████ 2/2
  Testing        ████████████     1.5/2
  Guardrails     ████████         1/2

  Improvements

  ! Add security rules (XSS, injection, etc.)

  ✓ Good structure with clear sections
  ✓ Tech stack and file structure documented
  ✓ Clear conventions with specific rules
  ✓ Testing framework and commands documented
─────────────────────────────────────────

Usage

# Score CLAUDE.md in current directory (auto-detects config files)
npx contextkit score

# Score a specific file
npx contextkit score path/to/CLAUDE.md

# Score a Cursor config
npx contextkit score .cursorrules

# Pipe content from stdin
cat CLAUDE.md | npx contextkit score --stdin

Auto-detection: If no file is specified, the CLI looks for CLAUDE.md, .claude/CLAUDE.md, .cursorrules, .cursor/rules, AGENTS.md, codex.md, or GEMINI.md in the current directory.

Exit codes: Returns 0 for scores >= 5/10, 1 for lower scores. Use in CI to enforce config quality.

CI example

# .github/workflows/lint-config.yml
name: Lint AI Config
on: [pull_request]
jobs:
  score:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - run: npx contextkit score

The problem

Every Claude Code project needs a CLAUDE.md. Every Cursor project needs .cursorrules. Gemini CLI needs GEMINI.md. GitHub Copilot needs AGENTS.md.

Writing these from scratch means:

• Guessing the right structure (there's no standard)
• Missing conventions your AI tool actually supports
• Copy-pasting from old projects and hoping it still works
• Spending 20 minutes on something that should take 30 seconds

And even if you already have a config file — how do you know if it's actually good?

ContextKit solves both: score existing configs from the terminal or web, or generate new ones automatically.

Web tools

Generator

Feature	Description
5-step wizard	Language, Framework, Project type, Conventions, Export
12+ languages	TypeScript, Python, Go, Rust, Java, C#, Ruby, PHP, Swift, Kotlin
16+ frameworks	React, Next.js, Vue, Nuxt, Svelte, Astro, Express, FastAPI, Django, Flask, Rails, Laravel, Spring, Gin
4 export formats	CLAUDE.md, .cursorrules, AGENTS.md, GEMINI.md
Shareable configs	URL-encoded state — share your exact configuration
No account needed	Runs in your browser, nothing uploaded

Analyzer

Feature	Description
Score out of 10	5 categories: Structure, Architecture, Conventions, Testing, Guardrails
Actionable feedback	Specific suggestions for each weak category
README badges	Show your CLAUDE.md score in your repo
Shareable scores	"My CLAUDE.md scored 8/10" — one-click share

Quick start

From the terminal

# Install globally (optional)
npm install -g contextkit

# Or run directly
npx contextkit score

From the web

1. Generate a config — 5-step wizard, 30 seconds
2. Score your config — paste and get a score

Scoring categories

Category	What it checks	Max
Structure	Headings, sections, markdown formatting, length	2
Architecture	Tech stack, project type, file structure, overview	2
Conventions	Coding rules, do's and don'ts, naming, imports	2
Testing	Test framework, commands, strategy, file locations	2
Guardrails	Security rules, scope limits, read-first policy	2

Privacy

Zero data collection. The CLI reads files locally. The web tools run as client-side JavaScript. Nothing is sent to any server.

Contributing

Issues and PRs welcome. Especially helpful:

• New language/framework templates
• Convention suggestions
• Analyzer category refinements
• CI integration examples

License

MIT — see LICENSE for details.

---

Built by Nova Labs — developer tools for AI-assisted coding.