Optimized Agent Instruction Linting
Maintaining clear and consistent instructions for AI agents is critical for reliable development workflows. The AGENTS.md Linter provides a rigorous analysis of your configuration files (AGENTS.md, CLAUDE.md, .cursorrules) to ensure your AI isn't being slowed down by conflicting commands or stale context.
What it does
This skill scans your repository's instruction files against your actual codebase to identify breaking issues. It goes beyond simple text checking by cross-referencing your project's package.json, Makefile, and source files to verify that the commands your agents are told to run actually exist.
- Detects conflicting rules that cause agent "hallucination" or refusal.
- Flags stale file paths and obsolete terminal commands.
- Identifies overly broad or unsafe instructions that could compromise code quality.
- Suggests consolidation strategies when multiple competing rule files are found.
Why use this skill
Unlike basic prompting, this tool uses specific rule IDs (AML001-AML007) and manifest scanning to provide high-precision feedback. It prevents the common "instruction bloat" that leads to context window waste and ensures your Cursor, Claude, or Copilot agents remain surgical and efficient.