What problem does it solve? Writing prompts for production LLM agents often results in vague instructions, leaked system/user boundaries, missing guardrails, and brittle behavior. This Skill enforces a structured, research-backed workflow so prompts are clear, safe, and correctly layered before they ship. ## Core Features & Use Cases - Mandatory clarification gate: Confirms structure (XML default vs Markdown), system/user split, and output target before any drafting begins. - Layered prompt architecture: Orders role, instructions, tools, guardrails, examples, context, task, and output format for clarity and prompt-caching efficiency. - Review and audit tooling: Ships a review checklist plus references on techniques (few-shot, CoT, ReAct, Reflexion), reliability, injection defense, and context engineering. - Ready-to-adapt templates: Includes XML and Markdown templates for system prompts, user prompts, and combined prompts. - Use Case: You need a system prompt for a contract-analysis agent with tool access and strict grounding rules. The Skill walks you through the gate questions, then produces a separate system prompt (role, tools, guardrails, examples) and user prompt (context, task, reminder) in XML. ## Quick Start Ask the agent to write a production system prompt for your LLM agent and answer its clarification questions about structure, system/user split, and output format.
