ai-fluency-promptinglisted

# Prompting — Description made concrete A better prompt is almost always faster than a retry. Description has three layers, and a strong prompt covers all three: **Product** (what output), **Process** (how to approach it), **Performance** (how to behave). ## The six techniques 1. **Provide context** — who you are, why you're asking, the situation. Shapes tone, depth, and format more than anything else. 2. **Show an example of "good"** — a sample output communicates a pattern more precisely than describing it. 3. **Specify output constraints** — format, length, sections, what to include/exclude. 4. **Break into ordered steps** — when the task has phases or order matters, or you want to review intermediate results. 5. **Ask it to think first** — request reasoning before the answer on complex or high-stakes work. 6. **Define the role** — role + audience + tone in one line calibrates vocabulary, depth, and structure. **Secret weapon:** when you're unsure how to phrase a request, ask the agent to help you write it — "I'm trying to get you to help me with X; help me craft an effective prompt for it." ## Pattern library **Role + Context + Constraint opener** (start of any substantial task) ``` You are working on [project/context]. I am [role/background]. The goal is [specific output]. Constraints: [format, length, tone, what to avoid]. Do NOT [the specific thing the agent tends to get wrong here]. ``` The **"Do NOT"** line is the single highest-leverage addition —