Why does the skill flag 'think step by step' as an issue?

Modern models like Claude 3.5 and 4.x are often fine-tuned for internal reasoning; explicit 'think step by step' instructions can be redundant and waste tokens compared to high-level reasoning directives.

What is the 'lost-in-the-middle' effect this skill checks for?

It refers to the tendency of LLMs to lose track of information placed in the middle of a long prompt. This skill ensures critical constraints are placed at the beginning or end for maximum impact.

How does this differ from the enhance-agent-prompts skill?

This skill focuses on the internal quality and logic of general prompts and system messages, whereas enhance-agent-prompts focuses specifically on agent configuration files, tools, and YAML frontmatter.

Does it support automated fixes for my prompt files?

Yes, by using the --fix flag, the skill can automatically correct high-certainty issues such as removing excessive capitalization or converting negative-only constraints into actionable positive instructions.

Enhance Prompts

Name: Enhance Prompts
Author: agent-sh

byagent-sh

•

692

•

開発者ツール

Optimizes AI prompt engineering through automated analysis of structure, clarity, and Claude-specific best practices.

The enhance-prompts skill is a specialized diagnostic utility for developers and prompt engineers looking to maximize the reliability of their AI instructions. It systematically analyzes prompt files for structural integrity, clarity, and adherence to modern LLM research, such as the use of XML tags and optimal few-shot example counts. By identifying common anti-patterns like vague qualifiers or buried critical constraints, it helps ensure that models—particularly Claude—respond with higher precision and fewer hallucinations.

主な機能

01Structural analysis using Claude-optimized XML tag patterns

02Detection of vague qualifiers and negative-only constraints

03692 GitHub stars

04Few-shot example validation and count optimization

05Automated fixes for clarity and aggressive formatting issues

06Context window optimization to prevent 'lost-in-the-middle' errors

ユースケース

01Auditing complex prompt templates for instruction hierarchy and clarity

02Refining production-grade system prompts for AI agents

03Standardizing prompt engineering patterns across development teams

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add agent-sh/agentsys enhance-prompts

For use in Claude.ai and ChatGPT