PROMPTS
The Anatomy Of A Perfect Prompt ⚡
Five parts. Most people use one, and that's why their AI sounds like a robot. The full skeleton plus 3 finished examples are on this page, ready to copy.
Why Your Outputs Feel Generic
Every disappointing AI answer has the same autopsy: the prompt was missing layers. The formula is five parts, and each one closes a specific failure mode:
01 · ROLE
Who should Claude BE right now? Without it you get the generic assistant. "Act as a [type of expert] who [their specialty]."
02 · CONTEXT
What can it not know without you? This is the #1 fix for generic answers. The output can only be as specific as the input.
03 · EXAMPLE
Show it, don't describe it. "Professional but friendly" means nothing. One real sample of your writing beats ten adjectives.
04 · RULES
Block the mistakes before they happen. You already know how this output usually goes wrong. Say so upfront instead of editing it out after.
05 · FORMAT
Say exactly what to hand you, including what you'll need NEXT. Don't make it guess what done looks like.
The Fill-In-The-Blank Template
This is the master skeleton with guidance written inside every bracket, so you know exactly what good looks like as you fill it in. It takes about 90 extra seconds versus a one-line prompt, and it changes everything.
Copy it, fill the brackets, delete the bracket guidance, send.
The Raise Email
All five layers on the highest-stakes email most people ever send. Notice how the rules block the classic failure modes (apologizing, over-explaining) before they can happen.
Copy this prompt
The Content Engine
Same skeleton, completely different job. The context layer carries the audience and the recent miss, the format layer asks for the diagnosis you'd never think to request.
Copy this prompt
The Big Decision
This one teaches the advanced lesson: the example layer is deliberately SKIPPED, because a decision doesn't need your writing voice. The skeleton is a checklist, not a cage. Drop layers that don't serve the task, never the ones that do.
Copy this prompt
One more for the road: when an answer comes back mediocre, don't patch it. Say "knowing everything you know now, scrap this and do the elegant version." First attempts carry first-attempt baggage.
The real unlock of AI agents is not full autonomy, but human-in-the-loop systems design. Structure your agent inputs with explicit constraints, negative prompts, and automated test checkpoints. When building tools, keep token consumption lean by caching system prompts and isolating tasks into specialized sub-agents.