context-optimizer

To optimize context usage, you can extract coding rules and workflows from CLAUDE.md into separate frontmatter-compliant files under the .claude/ directory. This minimizes token waste by compressing the main file to core essentials while retaining project guidance.

The recommended line count threshold is determined by analyzing your CLAUDE.md and .claude/ directory markdown lines. The context optimizer flags files exceeding recommended limits to prevent unnecessary token consumption during AI sessions.

You extract coding rules from CLAUDE.md by classifying sections into rules or skills, generating an extraction plan with target paths, and creating frontmatter-compliant files like rules/typescript.md or skills/review-fix/SKILL.md using provided templates.

Yes, you can analyze context distribution before refactoring CLAUDE.md by running the analyze-context.sh script. It counts markdown lines, reports ratios between CLAUDE.md and .claude/ files, and flags oversized files exceeding recommended thresholds.

You validate extracted rules and skills files by checking frontmatter compliance, description style, and reference depth using provided templates and checklists. This ensures the newly created files in the .claude/ directory meet formatting standards.

Your CLAUDE.md consumes excessive tokens during AI sessions because it likely contains oversized, monolithic content. Extracting rules and workflows into structured .claude/ files reduces context consumption by loading only essential project guidance.