Does this skill support modular rules?

Yes, it scans the .claude/rules/ directory to help you split monolithic CLAUDE.md files into path-scoped, topic-specific rule files for better organization.

What is Code-Doc Drift?

Drift occurs when source code evolves but documentation remains static, leading to AI hallucinations. This skill detects semantic mismatches between actual code signatures and documented APIs.

Who developed the best practices used by this optimizer?

The optimization patterns are derived from documentation strategies shared by Boris Cherny (Head of Claude Code) and Thariq Shihipar at Anthropic.

How does it detect the project lifecycle stage?

It uses heuristics including git history, versioning patterns, test coverage, and TODO counts to categorize projects as INIT, ACTIVE, STABLE, or MAINTENANCE.

Why is the token count of CLAUDE.md important?

Large CLAUDE.md files cause 'context rot,' where the AI begins to ignore or misinterpret instructions. Keeping the file around 2.5k tokens ensures high instruction adherence and optimal cache performance.

Claude Documentation Optimizer

Name: Claude Documentation Optimizer
Author: kojott

bykojott

•

学习与文档

Optimizes CLAUDE.md and project documentation ecosystems to improve AI context efficiency and code-doc synchronization.

The Documentation & CLAUDE.md Optimizer is a specialized tool designed to streamline project documentation according to Anthropic's battle-tested patterns for Claude Code. It utilizes a multi-agent architecture to perform deep token analysis, detect anti-patterns like context stuffing, and identify 'context rot' caused by oversized configuration files. By performing semantic synchronization between source code and API documentation, the skill ensures that your docs never drift from the actual implementation, maximizing Claude's reasoning capabilities and instruction adherence.

主要功能

01Parallel subagent analysis for project stage detection and anti-pattern identification

02Deep semantic synchronization to detect drift between source code and API documentation

034 GitHub stars

04Automated token analysis to keep CLAUDE.md within the optimal 2.5k token range

05Hierarchical scanning of documentation levels from enterprise to local project settings

06Comprehensive mapping of the .claude/ ecosystem including modular rules and hooks

使用场景

01Scaffolding standardized documentation ecosystems for new projects based on Anthropic best practices

02Auditing mature codebases to identify stale documentation or undocumented API endpoints

03Refactoring bloated CLAUDE.md files to improve AI instruction adherence and cache performance

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add kojott/claude-docu-optimizer docu-optimizer

For use in Claude.ai and ChatGPT

Download Skill