Twelve curated topics covering the day-to-day craft of working with LLMs: zero- vs few-shot prompting, system prompts and personas, temperature/top_p sampling, structured JSON outputs, function/tool-calling, chain-of-thought reasoning, ReAct, token budgeting, prompt caching, prompt-as-code templating, prompt-injection defenses, and LLM-as-judge evaluation. Three worked examples per topic plus common mistakes.