Mastering the Command Line of AI
Most people type a vague request into ChatGPT, Claude, or Gemini and accept whatever comes back. The difference between a mediocre answer and a genuinely useful one is rarely the model — it is the instruction. Treating prompts like shorthand "commands" lets you control the style, depth, and tone of any response in a single word, so you stop re-explaining yourself every time.
The shortcuts below are custom triggers you define once. Paste the legend into your AI's custom-instructions or memory setting (in ChatGPT: Settings → Personalization → Custom Instructions), tell it "when I use these codes, respond accordingly," and from then on a single token reshapes the entire reply.
Explanation Shortcuts
- /eli5: Extreme simplicity (Explain Like I'm Five) — strips jargon and uses plain language.
- /human: Removes the robotic, list-heavy tone for natural, conversational language.
- /deep: Comprehensive, technical breakdowns with edge cases and trade-offs.
- /analogy: Explains an unfamiliar concept via a familiar real-world comparison.
Writing & Creation
- /genz: Trendy, social-media-ready phrasing for captions and short-form.
- /hook: Focuses output on attention-grabbing first lines that stop the scroll.
- /script: Produces a full YouTube/Reel video architecture — hook, body, CTA.
- /rewrite: Keeps your meaning but tightens, clarifies, and varies the rhythm.
Productivity
- /check: Advanced proofreading — hunts grammar, logic, and consistency mistakes.
- /roadmap: Converts a fuzzy idea into a step-by-step, time-boxed timeline.
- /tldr: Compresses a long document into the three points that actually matter.
The Three Rules That Make Any Prompt Better
Shortcuts are leverage, but they sit on top of fundamentals. Whatever you ask, apply these:
- Give it a role. "Act as a senior copywriter reviewing a draft" beats "fix my text" — a role loads relevant context automatically.
- Show the format you want. Ask for a table, five bullet points, or a 100-word paragraph. Ambiguity is where bad output comes from.
- Give one example. A single sample of the tone or structure you're after ("like this: …") is worth a paragraph of description.
Worked example: Instead of "write an Instagram caption," try: "/hook /genz — Act as a creator in the tech niche. Write three caption options for a Reel about saving time with AI tools. Each under 120 characters, ending with a question." The same model, a far better result.
Save this legend once and your AI stops guessing what you meant. The goal of prompt engineering isn't memorizing tricks — it's removing ambiguity so the model spends its effort on your problem instead of decoding your request.