The Prompt Library, online.

What's new in this edition

• Material first, instructions last. Pasted documents now go at the top and the task at the end, the order today's AI tools handle best with long inputs.
• Clearly labeled inputs. Each piece of pasted material sits inside a simple label, like <transcript>, so the AI never confuses your notes with your instructions.
• The goal, not just the task. Prompts now say who the output is for and why, which helps the AI make better judgment calls.
• Evidence over guesswork. Prompts ask the AI to quote its sources, mark unknowns and ask questions instead of filling gaps.
• Checkable sources. Because many tools now search the web, the fact-check asks for links rather than the book's FROM MEMORY label.
• Plainer, leaner instructions. Fewer "never"s and all-caps warnings, which modern models tend to over-apply, and more positive direction with reasons.
• Fairness and privacy built in. The hiring, stakeholder and buying-committee prompts now keep to job-relevant, professional, public information.
• Corrected math. The value-model example now counts no savings before the time-to-value point.

Before you paste anything

AI tools process exactly what you give them. Before pasting, ask: would I be comfortable if my customer saw exactly what I just typed?

• Use your company's approved AI tool. Enterprise plans keep your inputs out of model training and limit retention.
• Leave out anything a customer would expect to stay private: internal financials, roadmaps, personnel matters.
• Never paste regulated data (health, financial-account or similar) into a general-purpose tool.
• When in doubt, swap names for placeholders. The output is just as useful.