prompting

Guide writing effective system prompts for LLM agents.

862|67|Updated Mar 2, 2026

Search Tags:#prompt engineering #llm #system prompt #prompt design #ai agent #llm configuration

Ready to run

npx skills add https://github.com/saffron-health/libretto --skill prompting-saffron-health

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: prompting
Source: https://github.com/saffron-health/libretto/tree/main/.agents/skills/prompting
Command: npx skills add https://github.com/saffron-health/libretto --skill prompting-saffron-health

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill helps users create and refine system prompts for AI agents, ensuring clear direction, effective constraints, and optimal performance.

Core Features & Use Cases

Prompt Design Guidance: Provides principles for writing effective system prompts, focusing on clarity and token efficiency.
Structure and Formatting: Offers best practices for organizing prompts using markdown and examples.
Model-Specific Tuning: Includes references for tailoring prompts to different LLM architectures (Claude, GPT, Gemini, Codex).
Use Case: You are building a new AI assistant and need to write its core system prompt. Use this Skill to understand how to structure the prompt for maximum effectiveness and to avoid common pitfalls.

Quick Start

Use the prompting skill to learn how to structure a system prompt for an AI agent.

Frequently Asked Questions about prompting

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

FAQPage Schema

How do I write an effective system prompt for an LLM agent?▼

Writing an effective system prompt requires structuring directions clearly using markdown, imposing strict constraints, and iterating on prompt design to optimize token efficiency and agent performance.

What is the best way to format system prompts for AI agents?▼

The best way to format system prompts is using markdown structures and providing examples, which helps organize directions clearly and ensures the LLM understands constraints and expected behaviors.

Can I use the same system prompt for different LLM architectures like Claude and GPT?▼

Using the same system prompt across different LLM architectures is not optimal; prompts should be tailored and tuned specifically for models like Claude, GPT, Gemini, and Codex to achieve maximum effectiveness.

Why does my LLM configuration produce inconsistent agent outputs?▼

Inconsistent LLM agent outputs often result from poorly designed system prompts lacking clear constraints, insufficient formatting, or a lack of model-specific tuning and iteration strategies.

How do I impose constraints on an AI agent in a system prompt?▼

To impose constraints on an AI agent, you must design the system prompt with explicit rules, clear boundaries, and structured formatting that guide the LLM's behavior and limit unwanted actions.

When do I need to optimize token efficiency in prompt design?▼

You need to optimize token efficiency in prompt design when building AI applications or agent configurations, as concise prompts reduce latency and costs while maintaining clear directions and constraints.