The Prompt Engineer's Library

OverviewIf your prompts don’t sell, it’s rarely because the model is “bad.” It’s because your prompt never locks the outcome: what you’re offering, who it’s for, what rules it must follow, how you’ll judge success, and what the final response must look like. This chapter builds that baseline prompting system - then gives you 50 plug-in frameworks you can reuse for high-conversion outputs.

You’ll get a repeatable structure for positioning and conversion (offer + audience + constraints + success metrics + response format), plus quick checks you can run before you hit “send.” Differentiator: every framework below includes an explicit response format and at least one concrete “constraint” you can copy-paste (price, word count, tone, compliance rules, or scoring rubric).

Takeaway prompt: After you finish the chapter, pick one real offer you’re working on (landing page, ad, outreach, proposal) and write your baseline prompt using the structure from the first few frameworks - then run it through the checklist in each item.

The Breakdown#1: Outcome-First Prompt Skeleton (OFS)Problem: Most prompts fail because they ask for “content” without pinning the outcome. You end up with good writing that doesn’t convert, because the model never knows what counts as a win (click, booking, signup) or how to measure it.

Solution: Start every prompt with 5 lines: Offer, Audience, Goal metric, Constraints, Response format. Use this exact order, and include numbers (e.g., “Goal: book 1 call per 50 messages” or “Target: 120-160 words”). End with a strict format block like: “Return: Subject | Hook | Proof | CTA | 1-sentence close.”

Result: You get outputs that match your funnel, not just your topic. Conversion improves because the model is optimizing for your metric and your layout.

#2: Offer Definition Card (ODC)Problem: If your offer is fuzzy (“help with marketing”), the model can’t position it sharply. Fuzzy offers cause generic benefits, weak specificity, and CTAs that don’t land.

Solution: Fill a card before prompting: Who gets results, What changes, Time to value, What’s included, Price anchor (even if it’s “from $X”), Risk reducer (guarantee, trial, refund). Paste the card into the prompt under “Offer Definition Card.” Ask the model to reuse your exact phrases where possible.

Result: Your output sounds like you, because the model is working from a fixed offer blueprint.

#3: Audience Slice + Objection MapProblem: Prompts that target “business owners” miss the specific trigger points that drive action. Without objections mapped, the model writes benefits but not the rebuttals that remove friction.

Solution: Provide 3 audience slices (e.g., “dentists with 2-5 locations,” “gym owners with declining memberships,” “local SaaS with long sales cycles”) and list the top 3 objections for each slice (time, trust, budget). Instruct: “Write for Audience Slice #1 only; address Objection #1 and #3 explicitly in the copy.”

Result: The copy becomes targeted and pre-sells the rebuttals, which typically lifts response rate.

#4: Constraint Stack (CS) for High-Converting CopyProblem: When you don’t set constraints, the model defaults to safe, broad language. That kills conversions because it avoids the exact emotional and logical moves your audience expects.

Solution: Add a “Constraint Stack” section with at least 6 constraints: word limit, reading level, forbidden phrases, claim rules (no guarantees), CTA style (direct booking link vs “reach out”), format. Example constraints: “Max 150 words,” “No buzzwords,” “No ‘guaranteed results’,” “CTA: ‘Book a 15-min fit check: [link]’.”

Result: You reduce variance and get sharper, more consistent outputs.

#5: Success Metrics Contract (SMC)Problem: “Make it persuasive” is not a metric. Without a success definition, the model can’t prioritize clarity, proof, or urgency in the right proportions.

Solution: Define a scoring rubric for the model to follow. Use 5 criteria with 0-2 points each, like: Hook clarity, Proof specificity, Objection handling, CTA directness, Constraint compliance. Ask: “After drafting, output a score table and revise once if any criterion < 1.”

Result: You get measurable persuasion instead of vibes.

#6: Response Format Lock (RFL)Problem: Even strong prompts drift in structure, and your sales funnel needs consistent blocks. Inconsistent formatting makes it harder to test and reuse copy across channels.

Solution: Provide a response template with exact headings and line counts. Example for email: “Subject (<= 60 chars) / Opening (1 line) / Problem (2 lines) / Proof (2 lines) / Offer (2 lines) / CTA (1 line) / P.S. (1 line).” Tell the model: “If you can’t fill a field, write ‘N/A’.”

Result: Faster iteration and cleaner A/B testing because outputs match your system.

...