prompt-engineeringlisted

# Prompt Engineering Skill ## When to activate - Writing a system prompt for a Claude-powered application or agent - Improving the reliability or quality of LLM outputs - Designing few-shot examples to steer model behaviour - Debugging why a model isn't following instructions - Setting up a prompt evaluation framework - Formatting prompts for structured outputs (JSON, markdown tables, etc.) ## When NOT to use - RAG system design — use the rag-architect skill - Agent orchestration — use the agent-construction or langgraph skill - Fine-tuning models — different process entirely - Prompt injection defence — that's a security topic for the security-reviewer agent ## Instructions ### System prompt design ``` Write a production system prompt for [use case]. Use case: [what the AI assistant does — customer support / coding assistant / data extractor / etc.] User: [who sends messages — end users / developers / internal staff] Output format: [free text / JSON / markdown / structured] Constraints: [what the assistant must never do] Tone: [professional / friendly / concise / technical] System prompt structure: ## Role and purpose [Define exactly who the assistant is and what it does] [Be specific — "You are a customer support assistant for Acme Inc., helping users troubleshoot billing issues" is better than "You are a helpful assistant"] ## What you do [List 3-5 specific things the assistant handles] [Use imperative: "Answer questions about...", "Help users...", "Extract..."]