🪴 Brain

40 AI Prompts for Expert Output

3 min read

40 AI Prompts for Expert Output

🚀 The core idea

A real prompt is an engineered instruction, not a wish. It specifies role, context, constraints, format, quality standard, and examples. The author tested 500+ prompts and kept the 40 that produce expert-level output every run on Claude, ChatGPT, and Gemini. Originally a Twitter thread by @eng_khairallah1.

Closely aligned with the engineered-prompt philosophy in Context Engineering and Claude Code Best Practice — substance over wishes.

🗒️ Description

Six categories, 40 prompts. Each prompt is a parametrized template with rules, structure, and quality gates. Save the list, customize the variables, build a personal template library over time.

🧩 Categories

  • Writing and content (01–10) — Expert Article Writer, Thread Architect, Email Drafter, Content Repurposer, Copywriting Converter, Blog Post Outliner, Storytelling Transformer, Headline Generator, Case Study Builder, Style Mimic
  • Analysis and strategy (11–20) — SWOT, Decision Matrix, Root Cause (5 Whys), Market Opportunity Scanner, Meeting Strategist, Pricing Strategist, Competitive Teardown, OKR Builder, Risk Assessor, Retrospective Facilitator
  • Technical and development (21–28) — Architecture Advisor, Code Reviewer, Debug Diagnostician, API Designer, Database Schema Designer, Test Case Generator, Documentation Writer, Refactoring Planner
  • Productivity and personal (29–32) — Weekly Planner (with deliberately-skipping section), Learning Accelerator, Negotiation Prep, Habit Designer
  • Data and research (33–35) — Data Interpreter, Survey Analyzer, Research Synthesizer
  • Communication (36–40) — Difficult Conversation Prep, Feedback Giver, Presentation Outliner, Apology Crafter, Elevator Pitch Builder

📒 Patterns worth stealing

  • Role + constraint + format + quality bar in every prompt — no vague asks.
  • Forced specificity — explicit “no filler phrases”, “no hedge words”, “every claim must be specific”.
  • Multi-version output — generate Version A (direct) and Version B (warm) so you can pick.
  • Refusal to accept framing — Root Cause prompt says “Do not accept my initial framing at face value”.
  • Pessimism on demand — Risk Assessor says “Be pessimistic. I want to hear about risks I have not considered.”
  • Deliberately-skipping section in the Weekly Planner — saying no is how priorities stay priorities.

✍️ One quoted prompt (#22 Code Reviewer)

Review this code… Check for: SECURITY, LOGIC, PERFORMANCE, READABILITY, BEST PRACTICES. For each issue found: Severity (Critical/High/Medium/Low), Exact location, Why it is a problem (not just what is wrong, but what could happen), The fix. If the code is clean, say so. Do not invent issues to seem thorough.

🔗 Related notes

📖 Further reading

Template: knowledge_note_info

Graph View