optimize-agent-instructions

Auditing agent instruction files involves checking them against a 12-point quality checklist to enforce Karpathy principles and remove stale content. This process produces consistent, authoritative guidance across CLAUDE.md, GEMINI.md, and Copilot instructions.

Karpathy principles for agent instruction files enforce authoritative, concise guidelines that govern agent behavior. They are applied through a 12-point quality checklist to ensure instructions remain clear and consistent across different platforms.

Yes, the agent instruction audit works with GEMINI.md files by applying platform-specific notes and adding Gemini tool mapping. It ensures consistent, authoritative guidance across GEMINI.md, CLAUDE.md, and Copilot instructions within your repository.

When updating Copilot instructions, the audit process preserves your existing project-specific rules. It removes stale or foreign content while applying a standardized 12-point checklist to maintain your customized agent behavior.

The best way to standardize agent instructions across multiple frameworks is to apply a standardized 12-point checklist and platform-specific notes. This enforces consistent, authoritative guidance across projects while eliminating stale or foreign content.

Removing stale or foreign content from instruction files is necessary to produce authoritative, concise guidelines that accurately govern agent behavior. This ensures your AI agents follow current, relevant project rules instead of outdated instructions.