Is this skill specifically for Claude prompts?

No, the patterns used are model-agnostic and effective across all major Large Language Models, including Claude, GPT-4, and Gemini.

How does the three-step flow work?

It first performs an analysis to detect issues, applies optimizations based on severity (P1-P3), and finally performs a balance adjustment to ensure quality and precision.

What kind of research is this skill based on?

It is based on peer-reviewed research and industry consensus from 2025-2026, including studies from ArXiv and ICLR regarding attention mechanisms and task decomposition.

Will this skill change the intent of my original prompt?

The skill follows a 'Non-Destructive' philosophy, meaning it aims to preserve your original intent while minimizing constraint creep and improving clarity.

Prompt Engineering Optimizer

Name: Prompt Engineering Optimizer
Author: sniper-fly

bysniper-fly

0•

Productivity & Workflow

Optimizes LLM prompts using evidence-based research patterns and a rigorous three-step refinement process to maximize AI performance.

The Prompt Optimization skill provides a systematic framework for analyzing and improving prompts across various AI models including Claude, GPT, and Gemini. Based on 2025-2026 research, it utilizes eight specific pattern benchmarks (BP-001 through BP-008) to identify critical issues like negative instructions, vague phrasing, and missing structures. The skill follows a disciplined three-step workflow—Analysis, Optimization, and Balance—to ensure your instructions are high-precision and context-rich while strictly preserving your original intent.

Key Features

01Research-backed pattern detection (BP-001 to BP-008) for critical prompt failures.

02Three-step workflow: Initial Analysis, Targeted Optimization, and Final Balance Adjustment.

03Automatic task decomposition for complex, multi-objective instructions.

040 GitHub stars

05Model-agnostic optimization effective across Claude, GPT, and Gemini.

06Hallucination reduction through uncertainty permission and explicit output formatting.

Use Cases

01Refining complex developer instructions into clear, structured, and evaluable tasks.

02Debugging inconsistent prompt outputs by identifying structural weaknesses and vague instructions.

03Standardizing prompt engineering best practices across a development team's workflow.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add sniper-fly/souma-recette prompt-optimization

For use in Claude.ai and ChatGPT

Key Features

01Research-backed pattern detection (BP-001 to BP-008) for critical prompt failures.

02Three-step workflow: Initial Analysis, Targeted Optimization, and Final Balance Adjustment.

03Automatic task decomposition for complex, multi-objective instructions.

040 GitHub stars

05Model-agnostic optimization effective across Claude, GPT, and Gemini.

06Hallucination reduction through uncertainty permission and explicit output formatting.

Use Cases

01Refining complex developer instructions into clear, structured, and evaluable tasks.

02Debugging inconsistent prompt outputs by identifying structural weaknesses and vague instructions.

03Standardizing prompt engineering best practices across a development team's workflow.