What problem does it solve? AI rule files like CLAUDE.md, AGENTS.md, and .cursorrules accumulate redundant, conflicting, stale, and bloated instructions over time, which wastes context window tokens and causes AI assistants to ignore or misinterpret your team's conventions. ## Core Features & Use Cases - Full-Project Discovery: Finds every AI configuration file across the project (CLAUDE.md, .cursorrules, AGENTS.md, copilot-instructions.md, and more) and builds a token-impact manifest. - 14 Automated Checks: Detects redundancy, in-file repetition, conflicts, staleness, example bloat, fragmentation, inferable-from-code rules, missing emphasis, oversized CLAUDE.md files, heavy imports, and rationale bloat. - Interactive Cleanup: Presents findings for explicit approval, proposes a concrete changeset, applies optimizations, and re-validates with before/after token counts. - Scoped Mode: Audits only files touched by an upgrade run, leaving untouched tuning undisturbed. - Use Case: After months of edits, your CLAUDE.md has grown to 4,500 tokens and contradicts your .cursorrules on export style. Run the optimizer to consolidate rules into a single source of truth and cut context usage per session. ## Quick Start Say "optimize AI rules" or "run ar-optimizer" to audit this project's AI rule files and get cleanup recommendations.
