RulesForge

Analyze any codebase and generate optimized AI coding rules for Claude Code, Cursor, Copilot, and more.

Your AI coding assistant is only as good as its context. RulesForge scans your project — frameworks, conventions, patterns, dependencies — and generates tailored rules files that make AI understand your codebase.

One command. Four output formats. Zero configuration.

Quick Start

npx rulesforge

That's it. RulesForge scans your project and generates:

| File | Tool | Purpose | |------|------|---------| | CLAUDE.md | Claude Code | Project context, conventions, and rules | | .cursorrules | Cursor | AI code generation guidelines | | AGENTS.md | Multi-agent tools | Agent task patterns and file ownership | | .github/copilot-instructions.md | GitHub Copilot | Code completion context |

What It Detects

Frameworks: React, Next.js, Remix, Astro, Vue, Nuxt, Svelte, Express, Fastify, NestJS, Django, Flask, FastAPI, and more.

Conventions: naming patterns (camelCase/snake_case), import style (relative/absolute/aliases), file naming, semicolons, quote style.

Tooling: package manager, test framework, linter, formatter.

Structure: source directories, test directories, CI/CD, documentation, entry points.

Why Not Just Use /init?

Claude Code's built-in /init generates a basic CLAUDE.md. RulesForge goes deeper:

| Feature | /init | RulesForge | |---------|---------|------------| | Framework detection | Basic | 15+ frameworks with version detection | | Convention analysis | No | Naming, imports, quotes, semicolons | | Cross-tool output | CLAUDE.md only | CLAUDE.md + .cursorrules + AGENTS.md + Copilot | | Framework-specific rules | Generic | React hooks, Next.js App Router, Express middleware, etc. | | Security rules | No | Stack-specific security best practices | | Quality score | No | Rate your existing rules coverage |

Options

Usage: rulesforge [path] [options]

Arguments:
  path                    Project directory to scan (default: ".")

Options:
  --format <formats>      Output formats (comma-separated: claude-md,cursorrules,agents-md,copilot)
  --dry-run               Preview generated content without writing files
  --diff                  Show what would change in existing files
  --output <dir>          Output directory (default: project root)
  --force                 Overwrite existing files without asking
  --quiet                 Minimal output
  --json                  Output scan results as JSON
  -v, --version           Show version
  -h, --help              Show help

Examples

Generate only CLAUDE.md:

npx rulesforge --format claude-md

Preview without writing:

npx rulesforge --dry-run

Scan a specific project:

npx rulesforge /path/to/project

Get scan results as JSON (for scripting):

npx rulesforge --json

Supported Frameworks

| Framework | Detection Method | Specific Rules | |-----------|-----------------|----------------| | React | package.json + JSX files | Hooks, components, keys, state | | Next.js | next.config.* | App Router, server components, metadata | | Express | package.json | Middleware, error handling, routing | | TypeScript | tsconfig.json | Strict types, interfaces, generics | | Python | .py files | Type hints, docstrings, imports | | And more... | Auto-detected | Framework-specific best practices |

How It Works

1. Scan — Walks your project tree, reads config files, samples source code
2. Analyze — Detects frameworks, infers conventions, maps structure
3. Generate — Applies framework-specific rules + convention enforcement + security rules
4. Write — Creates rules files in your project root

No AI. No API calls. No data leaves your machine. Pure static analysis.

Contributing

Found a framework we should support? Have better rules for your stack? PRs welcome!

git clone https://github.com/Wittlesus/rulesforge
cd rulesforge
npm install
npm test

License

MIT

Built by @WSDevGuy | GitHub