optimize-claude-md

Optimize your CLAUDE.md by analyzing real transcripts to identify recurring failure modes, then generating a concise, evidence-backed replacement file that anchors rules to those concrete examples.

Your CLAUDE.md likely contains guesswork instead of evidence-based rules. Inferring recurring failure modes from real transcripts and grounding rules in those findings removes unaligned or redundant content.

The best way to generate an evidence-based CLAUDE.md is to process historical transcripts, extract recurring failure patterns, and output a concise replacement configuration strictly grounded in those transcript findings.

Yes, you can use transcript data to create a lean CLAUDE.md. Analyzing transcripts maintains strict discovery filters to keep content non-redundant and removes rules not grounded in observed evidence.

To stop introducing ungrounded content, replace your CLAUDE.md with a version optimized exclusively from real transcript evidence, ensuring all rules align with observed memory and failure modes.

You should avoid transcript-based CLAUDE.md optimization if your repository lacks sufficient historical conversation transcripts to identify recurring failure modes and extract reliable, evidence-backed rules.