claude-context-improver

Audit Claude Code context files and propose optimization diffs.

16|1|Updated Jan 2, 2026
One-click install
npx skills add https://github.com/bahayonghang/my-ai-cli-toolkit --skill claude-context-improver
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: claude-context-improver
Source: https://github.com/bahayonghang/my-ai-cli-toolkit/tree/main/skills/developer-tools-integrations/claude-context-improver
Command: npx skills add https://github.com/bahayonghang/my-ai-cli-toolkit --skill claude-context-improver

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill addresses the common issue of bloated or conflicting context files (CLAUDE.md, .claude/rules, code_map.md) that degrade AI performance and increase token costs.

Core Features & Use Cases

  • Context Auditing: Scores your repository's context layer against Claude 5 engineering principles.
  • Conflict Resolution: Identifies and helps remove contradictory instructions across different configuration layers.
  • Optimization: Provides actionable diffs to slim down guidance files, ensuring they remain under the 200-line target for maximum model adherence.
  • Use Case: Use this when you notice Claude Code is ignoring instructions or when you want to ensure your repository's guidance is concise, current, and conflict-free.

Quick Start

Use the claude-context-improver skill to audit the current repository and generate an optimization report.

Frequently Asked Questions about claude-context-improver

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I audit and optimize Claude Code context files for better instruction adherence?

Claude Code ignores instructions when context files like CLAUDE.md become bloated or contain conflicting guidance across different configuration layers. Auditing these files against context-engineering standards identifies and removes contradictory instructions to restore model adherence.

What is the maximum recommended length for a CLAUDE.md context file?

The maximum recommended length for a CLAUDE.md file is 200 lines. Keeping repository guidance under this target ensures maximum model adherence, reduces token costs, and prevents context bloat that degrades AI performance.

How do I remove conflicting instructions across Claude Code configuration layers?

To remove conflicting instructions, audit your CLAUDE.md, .claude/rules, and code_map.md files. The audit detects contradictory guidance across these configuration layers and proposes targeted diffs to resolve conflicts and ensure progressive disclosure.

Do I need git to audit my repository's Claude Code context?

Yes, you need git and standard shell utilities to audit your Claude Code context. These tools verify the file structure of your context layers and propose targeted diffs for context optimization.

What's the best way to slim down bloated code_map.md files for AI performance?

The best way to slim down bloated code_map.md files is to audit them against Claude 5 context-engineering principles. This generates actionable diffs to optimize repository guidance, ensuring it remains concise and progressively disclosed for better AI performance.