Markdown Output Prompt — Lock the Document Structure
The contract that stops AI documents from restructuring themselves: a pinned section skeleton, forced tables, and strict consistency rules.
Turn product positioning into above-the-fold hero copy: a headline, subheadline, primary and secondary CTA, a trust cue, an objection-handling line, and three alternative headline angles to test.
A hero section fails when it describes the company instead of promising the visitor an outcome — and when its claims are invented to sound impressive. This prompt does neither. It takes the product, the audience, the core problem, the positioning, and whatever real proof exists, and returns above-the-fold hero copy: an outcome-led headline, a concrete subheadline, a primary and a secondary CTA, a trust cue, an objection-handling line, and three labelled alternative headline angles ready to test. Every claim traces to the positioning or proof you pasted; anything unsupported is marked [needs evidence] rather than invented — no fabricated testimonials, customer counts, awards, or guarantees. It produces ready-to-place hero copy, not a wireframe, a brand strategy, or a long-form sales page.
Lead with the outcome
The headline names what the visitor gets, not what the company is — the subhead makes it concrete.
Earn the click twice
A primary CTA for the ready and a secondary for the hesitant, each an action plus its value.
Never invent proof
Trust cues use only real proof; missing proof becomes a [needs evidence] placeholder, not a fabrication.
The [needs evidence] marker appears in item 5, the TRUST CUE, when your 'known proof or constraints' block was left blank or set to 'none'. Rather than fabricate a testimonial, customer count, or award, the prompt names the proof that belongs there and flags it. Paste a real metric, integration, or named guarantee to fill it.
Item 7 returns three headlines, each tagged by angle — outcome, pain-relief, speed, or cost-of-status-quo — so every variant tests a different hook instead of a reworded near-duplicate. You paste the labelled candidates into your split-testing tool as competing hero headlines; the prompt supplies them, but running the test and picking a winner stays with you.
The 'current positioning / differentiator' block matters most: it shapes the PRIMARY CTA (item 3) into an action-plus-value label like 'Start free', and the SECONDARY CTA (item 4) into a lower-commitment option such as 'See how it works'. Paste a thin positioning line and both buttons read generic; a sharp differentiator gives them something specific to promise.
The contract that stops AI documents from restructuring themselves: a pinned section skeleton, forced tables, and strict consistency rules.
Overview, Installation, Usage, Examples, Configuration — the README skeleton with required, runnable code examples.
Overview, Authentication, Endpoints, Error Handling, Rate Limits — endpoint docs in an identical structure, with parameter tables and runnable examples forced.
Summary, Why It Matters, What Happens Next — the executive summary contract for readers who will never open the source.
Stop getting 'Sure, here is the JSON…' — the output-contract pattern that forces models to return only parseable JSON: schema, example, and a strict rule block.
The JSON won't parse and you can't see why. Deterministic cause-sniffing — trailing commas, single quotes, unclosed brackets — and the repair prompt that fixes it.
Build prompts that produce documents in a fixed structure — headings, sections, and tables.
The full path to a business website that holds together — plan the content, structure the site, design the components, write the page copy, then ship it as one coherent whole.
Write a landing page that converts, not one that just describes — sharpen the value proposition, draft the hero and benefits, answer objections at the CTA, then A/B the variants to pick the stronger.