Why should I avoid negative instructions in CLAUDE.md?

Models often focus on key nouns; saying 'Never use X' still primes the model with the concept of 'X'. Using positive framing like 'Use Y instead' provides a clear, actionable alternative.

How do 'Reason' blocks improve Claude's performance?

Claude generalizes better when it understands the 'why' behind an instruction, which helps it apply your rules more consistently in novel or edge-case scenarios.

What makes this skill specific to Claude 4.5?

It emphasizes direct action language and structures tasks to leverage Claude 4.5's specific strengths, such as executing multiple tool calls in parallel and more concise communication.

Can this skill help with tool selection issues?

Yes, it encourages the use of structured tool-selection tables that map specific operations to the correct Claude Code tools, reducing errors in file operations and execution.

Claude Code Prompt Optimizer

Name: Claude Code Prompt Optimizer
Author: BbgnsurfTech

byBbgnsurfTech

•

개발자 도구

Refines CLAUDE.md files and Skill configurations using Anthropic's best practices to improve model instruction following and precision.

소개

The Claude Code Prompt Optimizer is a specialized skill designed to enhance the performance of the Claude Code CLI by auditing and improving system instructions. It leverages Anthropic's official prompt engineering principles, such as transforming negative prohibitions into positive action patterns, adding contextual motivation to rules, and implementing multi-shot examples. This skill is particularly useful for developers who want to reduce model hedging, ensure strict adherence to project-specific coding standards, and take full advantage of Claude 4.5's capabilities like parallel tool execution and direct action language.

주요 기능

Structures CLAUDE.md files using optimized Markdown headings and front-loaded priorities.
Provides multi-shot examples for complex behaviors like error handling and commit formats.
Converts 'Don't' prohibitions into clear 'Do' instructions with positive framing.
3 GitHub stars
Injects 'Reason:' blocks into guidelines to improve model generalization and compliance.
Optimizes Skill metadata for improved discovery and tool-call accuracy within the CLI.

사용 사례

Creating high-performance custom Skills with precise tool restrictions and descriptions.
Refactoring existing CLAUDE.md files that contain conflicting or vague instructions.
Onboarding new projects by setting up structured, project-specific memory for the AI.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add bbgnsurftech/claude-skills-collection prompt-optimization-claude-45

For use in Claude.ai and ChatGPT

Download Skill

GitHub

소개