roast-my-agents-md

To audit AGENTS.md files for redundant instructions, you can use static analysis and A/B testing to identify dead weight rules, providing evidence-based recommendations for optimization and reducing token costs.

A/B testing for AI instruction tuning empirically proves which rules the AI already follows without explicit instruction, identifying dead weight and delivering a data-backed score for configuration optimization.

To reduce token costs caused by bloat in CLAUDE.md files, perform a static audit to detect anti-patterns and redundancy, then apply data-backed restructuring steps to remove ineffective instructions.

The best way to optimize AI performance through instruction tuning is to conduct A/B tests that empirically prove rule effectiveness, then apply evidence-based recommendations to restructure your configuration.

Yes, you can use static analysis to quickly audit AI configuration files for common anti-patterns, bloat, and redundancy, providing a data-backed score and actionable steps for optimization.

You should restructure your AI configuration files when A/B testing identifies dead weight instructions that the AI already follows without explicit rules, proving they are ineffective and increasing token costs.