rulox

One source of truth for all your AI coding agents.

Most teams use multiple AI coding tools — Claude Code, Cursor, Copilot, Windsurf. Each has its own context file format. rulox audits your context file, flags contradictions/staleness/vagueness, and can generate an optimized version for each agent.

Install

npx rulox

Usage

npx rulox
npx rulox --file ./CLAUDE.md
npx rulox --no-optimize
npx rulox --write
npx rulox --json
npx rulox --sync

Sync to all agents at once

npx rulox --sync

Files written when syncing:

CLAUDE.md
.cursorrules
.windsurfrules
.github/copilot-instructions.md
AGENTS.md
CONVENTIONS.md
.clinerules

Scoring

• Length — line count and file size pressure
• Contradictions — conflicting instructions (always/never, use/don't use)
• Staleness — hardcoded paths, version pins, temporal references
• Specificity — concrete commands and examples
• Redundancy — rules the LLM already knows by default
• Format Compliance — agent-specific formatting limits

Anti-patterns

rulox names what it finds. Named anti-patterns help prioritize fixes:

• 👻 Ghost Path — hardcoded file paths that are probably stale
• 🔮 Obvious Oracle — rules the LLM already knows by default
• ⚡ Contradiction Trap — directly conflicting rules
• 📚 Readme Dump — README pasted as context
• 🗄️ Rule Hoarder — many rules, few removals
• 📸 Stale Snapshot — pinned versions or old dates
• 🧱 Wall of Text — extremely long files
• 🔧 Format Mismatch — markdown in plain formats

Example output

─────────────────────────────────────────
  rulox  v0.1.0
─────────────────────────────────────────

  Analyzing: CLAUDE.md (84 lines)

  SCORES
  ──────
  Length            72/100   ⚠
  Contradictions   100/100   ✓
  Staleness         60/100   ✗
  Specificity       85/100   ✓
  Redundancy        40/100   ✗
  Format Compliance 80/100   ✓

  Overall: 71/100

  ISSUES FOUND
  ────────────
  [staleness]    L23  Hardcoded path "src/utils/helpers.ts" — likely stale
  [redundancy]   L12  "don't break existing tests" — LLM default, remove

  OPTIMIZATION
  ────────────
  Calling Anthropic API to generate optimized version...
  Optimized file: 51 lines (↓ 33 lines, ↓ 39%)

  Diff saved to: CLAUDE.md.diff
  Run with --write to apply changes.

─────────────────────────────────────────

Badges