prompt-craft

Structure prompts for dual-model AI systems with templates and few-shot design.

3|Updated Feb 15, 2026

Search Tags:#prompt engineering #few-shot learning #structured output #ai prompting #large language models #model adaptation

Ready to run

npx skills add https://github.com/pvliesdonk/agents.md --skill prompt-craft-pvliesdonk

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: prompt-craft
Source: https://github.com/pvliesdonk/agents.md/tree/main/claude-skills/prompt-craft
Command: npx skills add https://github.com/pvliesdonk/agents.md --skill prompt-craft-pvliesdonk

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill helps you craft effective prompts for AI models, ensuring you get the desired output by adapting to different model capabilities and output formats.

Core Features & Use Cases

Prompt Structuring: Provides a clear template for organizing prompt components (Role, Context, Task, Format, Constraints, Examples).
Dual-Model Adaptation: Offers specific strategies for optimizing prompts for both small, local models (e.g., Ollama 4B-8B) and large, cloud-based models (e.g., GPT-5, Claude).
Structured Output Guidance: Details how to request and receive structured data (JSON, etc.) from AI models, including error handling.
Use Case: You need to generate product descriptions using a small local model, but then refine them using a powerful cloud model. This Skill guides you on how to structure your prompts for each stage to maximize quality and efficiency.

Quick Start

Use the prompt-craft skill to create a prompt for generating a JSON output describing a user's sentiment.

Frequently Asked Questions about prompt-craft

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

FAQPage Schema

How do I structure prompts for small local models versus large cloud models?▼

Prompt structure for large language models requires explicit instructions and concise templates for small models (4B-8B), while large models (GPT-5, Claude) handle complex context and nuanced formatting better.

What is the best way to get structured JSON output from an AI model?▼

Structured output prompting requires specifying JSON formatting requirements in the prompt template, using defensive patterns and systematic testing to ensure large language models consistently return valid structured data.

How do I use few-shot learning to improve AI prompting results?▼

Few-shot learning improves AI prompting by providing input-output examples within the prompt template, helping large language models understand desired formatting and task patterns to reduce output errors.

Why does my prompt engineering fail to adapt across different large language models?▼

Prompt engineering fails across large language models when strategies are not adapted to specific capabilities; adjusting prompt length, explicitness, and output formatting requirements ensures consistent performance between different model architectures.

Can I use the same prompt template for generating product descriptions on different model sizes?▼

Using the same prompt template across different model sizes is not recommended; small local models need more explicit constraints and structured examples, whereas large cloud models can infer context with fewer instructions for tasks like product descriptions.