The single highest-leverage skill in this course: a prompt structure that works in Claude, ChatGPT, Gemini and Grok identically. Learn it once and you never relearn "how to prompt tool X".
The four-part prompt skeleton
- 1Role — who the model should act as ("You are a careful financial analyst").
- 2Task — the single concrete thing you want, with the output format.
- 3Context — the facts, constraints, and examples it needs (the part people skip).
- 4Iterate — react to the first draft instead of expecting perfection.
Role: You are an experienced B2B copywriter for Indian SaaS.
Task: Write 3 cold-email subject lines (max 7 words each) for the
product below. Output as a numbered list, no preamble.
Context: Product = invoicing tool for freelancers. Audience = solo
designers earning Rs.50k-1L/mo. Tone = direct, not salesy.
Avoid the words "revolutionise" and "game-changer".Notice there is nothing tool-specific here. Paste it into any chat tool and you get comparable results. The skill is structure and context, not magic words.
Context is the lever most people never pull
One or two examples of the output you want (a good past email, a sample row) moves quality more than any clever phrasing. This is "few-shot" prompting — it works on every model.
Iterate like an editor, not a gambler
- Bad: re-run the same prompt hoping for luck.
- Good: "The 2nd line is closest. Make all three that punchy and cut adjectives."
- Good: "You assumed B2C — this is B2B. Redo with that constraint."
Claude rewards long, structured context. Grok rewards asking for live info explicitly. Gemini rewards pointing it at a Doc. The skeleton stays the same; you just lean into each tool's strength (lessons 3–7).
Role · Task · Context · Iterate. Master this once and every tool gets better — context and examples do the heavy lifting.