A structured, repeatable workflow for designing, scoring, and iterating on LLM prompts. Use this skill whenever a user wants to build, improve, or formalize a prompt — for any or any agent. Trigger on phrases like "help me write a prompt", "build a system prompt", "design a prompt for", "how should I prompt Claude to", "let's engineer a prompt", "make a prompt for", or any time a user is about to hardcode instructions into an AI feature without a structured process. Also trigger when a user says "my prompt isn't working" or asks to improve an existing prompt.
This skill includes a SKILL.md with agent instructions.