whereismyai
Ruleherd
Ruleherd is an open source CLI tool that audits agent instruction files for contradictions, dead references, and drift.
Ruleherd audits agent instruction files like CLAUDE.md, AGENTS.md, .cursor/rules, .windsurfrules, .github/copilot-instructions.md, and GEMINI.md. It reads files already on disk to find conflicts, dead references, dead commands, duplicates, stale version mentions, dead globs, bloat, and missing files. It provides a score out of 100, a letter grade, and per-file token estimates. It runs via npx with zero dependencies and requires Node 18 or higher. The tool is designed for developers who manage multiple agent config files and want to detect when those files stop matching.
Ruleherd at a glance
Key features
- Detects conflicting rules across instruction files
- Finds dead file references and import paths
- Validates commands against package.json scripts
- Identifies duplicate paragraphs across files
- Checks version mentions against installed package majors
- Flags dead glob prefixes in instruction files
- Estimates token counts per file
- Generates a score and letter grade
Use it for
- Auditing AI agent instruction files for drift
- Checking for contradictory rules in config files
- Finding broken file paths and dead commands
- Estimating token costs of agent instructions
Alternatives to Ruleherd
- jevgrep — Reduces coding agent costs by providing targeted, relevant repository context.
- Ruler — One source of rules for every coding agent in a repo
- Agent Archive — Hand off coding sessions between Claude Code, Codex, and Cursor across machines.
- Command Code — Repairs open model tool calls and learns developer patterns automatically.
- Viberuler — Computes exact token efficiency and generates shareable scorecards.
- Stop That Shit — Enforces task boundaries to stop unnecessary AI agent actions.
- Reviewer Agent — It uses a separate AI model to review code and prevent premature agent stopping.
- Bugbout — Measures coding agent performance on your project code using your own tests.
- Blume — Monitors coding agents and keeps your rulebook maintained locally.
- Devbroom — Sorts leftover agent files by evidence and moves them safely to the trash.
- GitWarren — Reviews agent worktrees directly before code reaches GitHub.
- Agent Trials — Test if an agent fixes your bugs in your stack at a sane price before deployment.
All Ruleherd alternatives