agent-instructions-improver

To audit AI agent instruction files, scan project roots and monorepos for configuration files like CLAUDE.md and .cursorrules, evaluate them against quality rubrics, generate a quality report, and apply targeted updates after user approval.

Agent instruction files like CLAUDE.md and .cursorrules provide project context to AI agents. Quality assessment ensures this context is accurate and up-to-date by scoring the files against documented templates and guidelines, preventing outdated or inconsistent AI behavior.

Yes, you can scan for instruction files across monorepos. The tool discovers agent instruction files located in various project roots, evaluates them against quality criteria, and outputs a comprehensive quality report covering the entire codebase structure.

The best way to update AI agent instructions is through targeted, minimal improvements. After auditing the files and generating a quality report, proposed changes are reviewed and only applied after explicit user approval to maintain template consistency.

No, the instruction file audit does not automatically modify your codebase. It evaluates files against quality rubrics and proposes minimal improvements, but only updates the instruction files after you explicitly review and approve the suggested changes.