agent-md-refactor

To refactor large agent instruction markdown, extract universal essentials into a minimal root file, move categorized guidelines into linked topic files, and prune redundant, vague, or outdated instructions.

Progressive disclosure for agent instructions separates root-only essentials from topic-specific guidance, generating a linked file hierarchy that prevents AI agents from being overwhelmed by oversized markdown documents.

Handling contradictions in agent instruction markdown requires identifying conflicting instructions and prompting the user to resolve precedence before proceeding with the file restructuring.

You should restructure AI agent documentation when markdown files become too long, inconsistent, or hard to maintain, causing agents to unreliably follow tangled configuration guidance.

The best way to clean up bloated agent configuration markdown is performing a pruning pass that flags non-actionable, default, or redundant instructions for deletion while reorganizing the remaining content into self-contained linked topics.