onerules

Stop your AI from writing slop.

One command generates anti-slop coding rules for 12 AI tools.
Detects your stack, libraries, and tooling. Works offline. Under 2 seconds.

简体中文

The Problem

AI coding tools generate slop: unnecessary abstractions, wrapper classes that wrap nothing, useMemo on everything, try/catch around code that can't throw, AbstractFactoryProviderManager, and 500-line files that should be 50.

Meanwhile, your CLAUDE.md says "follow best practices" and "write clean code" — rules that AI completely ignores because they're too vague.

How Sloppy Are Your Rules?

Run onerules doctor on your existing rules to find out:

Then fix them in one command:

Quick Start

# Install
npm i -g @blackforge/onerules

# Score your existing rules
onerules doctor

# Generate anti-slop rules for all 12 tools
onerules

# Or use the interactive wizard
onerules init

No API keys. No config. Works offline.

What Makes This Different

These aren't generic "follow best practices" rules. Every rule targets a specific AI slop pattern:

# Instead of vague rules AI ignores:
- "Follow best practices"
- "Write clean code"
- "Add proper error handling"

# You get rules that actually change AI behavior:
- DO NOT over-abstract. No AbstractFactoryProviderManager. If the class
  name needs 3+ words, it's doing too much or too little.
- DO NOT add try/catch around code that cannot throw.
- DO NOT create interfaces for a single implementation. Interface
  `IUserService` with one class `UserService` is pointless indirection.
- DO NOT install axios. The Fetch API is built-in and cached by Next.js.

Stack-Aware Rules

onerules doesn't just detect your framework — it detects your libraries and generates rules specific to each one:

  onerules inspect

  Languages:    typescript
  Framework:    nextjs
  Libraries:    tailwindcss, prisma-client, zod, tanstack-react-query,
                zustand, stripe, next-auth, radix-ui-react-slot
  Package mgr:  pnpm
  Test fwk:     vitest
  Linter:       biome

  Rules that will be generated:
    ✓ base anti-slop rules
    ✓ typescript language rules
    ✓ nextjs framework rules
    ✓ tailwindcss library rules
    ✓ prisma-client library rules
    ✓ zod library rules
    ✓ tanstack-react-query library rules
    ✓ zustand library rules
    ✓ stripe library rules
    ✓ next-auth library rules
    ✓ radix-ui-react-slot library rules
    ✓ vitest testing conventions
    ✓ biome linting

Two Next.js projects with different libraries get different rules.

12 AI Tools, One Command

Features

• Anti-slop rules — every rule targets a specific AI failure mode, not generic advice
• Stack-aware — detects your libraries and generates Prisma, Zod, Tailwind, Stripe, NextAuth-specific rules (25 libraries total)
• onerules doctor — scores your existing rules 0-100 and suggests specific fixes
• onerules init — interactive wizard with tool selection and stack preview
• onerules inspect — shows everything detected and which rules will apply
• onerules monorepo — per-workspace rules in monorepos
• --merge mode — smart merge: adds missing rules to existing files without overwriting project-specific content
• --strict mode — extra aggressive rules (max function length, no default exports)
• --minimal mode — core anti-slop only, skip framework/library details
• .onerulesrc — custom rules that survive onerules update
• 12 AI tools, 23 frameworks, 25 libraries, 6 languages
• No LLM required — deterministic, fast (<2s), works completely offline

Commands

onerules                         # Generate for all 12 tools
onerules doctor                  # Score existing rules 0-100
onerules init                    # Interactive setup wizard
onerules inspect                 # Show detected stack + rule categories
onerules update                  # Re-detect and regenerate all files
onerules monorepo                # Generate per-workspace in monorepos
onerules diff                    # Preview what would be generated
onerules -t claude,cursor        # Generate for specific tools only
onerules --merge                 # Add missing rules to existing files
onerules --strict                # Extra aggressive rules
onerules --minimal               # Core anti-slop only
onerules --force                 # Overwrite existing files

Already Have a CLAUDE.md?

Don't replace it — merge into it:

onerules --merge

This scans your existing rule files, detects which anti-slop rules you're missing, and appends only the new ones. Your project-specific context stays intact.

  Merged into 1 existing file:
    + CLAUDE.md                    23 rules added (44 already present)

Custom Rules (.onerulesrc)

Add a .onerulesrc file to your project root. Your custom rules merge into all generated files and survive onerules update.

{
  "projectContext": "This is a fintech app handling real money.",
  "doNot": [
    "DO NOT use floating point for money. Use integers (cents).",
    "DO NOT use any ORM. Write raw SQL queries."
  ],
  "security": ["All endpoints require authentication."]
}

Supported Stacks

Languages: TypeScript, JavaScript, Python, Go, Rust, Ruby

Frameworks: Next.js, React, Vue, Nuxt, Svelte, SvelteKit, Angular, Astro, Remix, Express, Fastify, Hono, FastAPI, Django, Flask, Rails, Gin, Fiber, Actix, Axum, Tauri, Electron, React Native

Libraries (with specific rules): Prisma, Drizzle, Zod, tRPC, Tailwind CSS, React Query, Zustand, Jotai, Redux Toolkit, Radix UI, Framer Motion, Stripe, NextAuth, Lucia, Mongoose, SQLAlchemy, Pydantic, Celery, Redis, Socket.io, GraphQL, Three.js, Playwright, Cypress

Tooling: pnpm, yarn, bun, npm, uv, poetry, pip, cargo, bundler | Vitest, Jest, Playwright, Cypress, pytest, RSpec | ESLint, Biome, Ruff, RuboCop | Prettier, dprint, Black

Why This Works

Generic rules like "follow best practices" and "write clean code" are useless — AI tools are already trained on those. They ignore them.

onerules generates rules that target specific AI failure modes: the AbstractFactoryProviderManager class nobody asked for, the useMemo on a function that runs once, the try/catch around a + b.

See the full before vs after comparison for side-by-side examples across Next.js, React, FastAPI, and testing.

FAQ

How do I know if my existing rules are any good? Run onerules doctor. It scores your rules 0-100 and tells you exactly what's weak.

Does this replace my existing CLAUDE.md? Not by default. Use --force to overwrite. Or run onerules doctor first to see if your rules need replacing.

Does this call any AI APIs? No. Fully deterministic. Works offline. No API keys needed.

How is this different from awesome-cursorrules or Karpathy's CLAUDE.md? Those are single files for one tool. onerules generates rules for 12 tools simultaneously, tuned to your specific framework AND libraries, with anti-slop rules that target AI failure modes specifically.

Does this work with monorepos? Yes. Run onerules monorepo to generate per-workspace rules with workspace-specific detection.

Contributing

We welcome contributions! The easiest ways:

1. Improve rules — make the anti-slop rules sharper for any framework or library
2. Add a framework — create a fragment in src/templates/fragments/
3. Add a library — add rules in src/templates/libraries/index.ts
4. Add a tool — create a generator in src/generate/

See CONTRIBUTING.md for the full guide.

License

MIT