Prompts that survive

Reusable prompt design

May 06, 2026

Most prompts die after one use

You write a prompt, get an output, copy it, move on. Next week the same task comes around and you start over from scratch. Slightly different wording, slightly different output, none of them quite as good as the last one.

The problem is not the content. The problem is that the prompt was never designed to be reused or to produce a structured, reliable output in the first place. So every output is a one-off, every output is a little weaker than it could be, and the time you thought AI was saving you quietly leaks back out through rewrites.

A developer on Reddit shared four prompts from their free library at PromptCreek that break this pattern. They get used week after week without rewriting. The reason is not clever roleplay or a magic phrase. It is forced output structure.

In partnership with

Your prompts are leaving out 80% of what you're thinking.

When you type a prompt, you summarize. When you speak one, you explain. Wispr Flow captures your full reasoning — constraints, edge cases, examples, tone — and turns it into clean, structured text you paste into ChatGPT, Claude, or any AI tool. The difference shows up immediately. More context in, fewer follow-ups out.

89% of messages sent with zero edits. Used by teams at OpenAI, Vercel, and Clay. Try Wispr Flow free — works on Mac, Windows, and iPhone.

The four prompts (and why each one works)

SOP Writer

Use case: internal documentation precise enough for a new hire.

What makes it work: the prompt forces IF/THEN branches inline, WARNING/CAUTION/NOTE callouts at safety-critical steps, and version-control metadata: author, approver, review date. That’s the stuff everyone forgets in DIY SOPs. Without it, you get a checklist. With it, you get a working document. The IF/THEN branching alone is worth it. Real workflows have decision points. Flat checklists pretend they don’t. When a new hire hits step 7 and something doesn’t look right, they need to know what to do next. A branching SOP tells them. A flat checklist sends them to Slack.

Variables: industry, compliance framework (SOC 2, ISO 27001), complexity level. Same prompt, very different outputs depending on inputs.

Research Paper TL;DR Generator

Use case: turning dense papers into summaries you can actually internalize in two minutes.

What makes it work: most summary prompts overstate findings. The model takes “suggests” and upgrades it to “proves.” This one is structurally built to resist that. It explicitly tells the model to hedge the way the original authors hedged. It also forces a Limitations section, so you get what the paper cannot tell you, not just what it claims. If you’re making decisions based on research, that distinction matters enormously. A study with 40 participants run over two weeks does not prove the same thing as a 10-year longitudinal study with 10,000 subjects. The Limitations section is where that context lives.

Variables: summary depth (TL;DR to lit-review entry), jargon level, target reader.

Competitor Briefing Generator

Use case: turning scattered competitor data into something a stakeholder can act on.

What makes it work: every claim in the output must be either grounded in evidence you provided OR explicitly flagged as inference. That’s the line between useful briefing and ChatGPT confabulating about competitors. The output structure: Positioning Map, Profiles, Comparative Table, Strategic Implications. That produces a document, not a wall of paragraphs. The Comparative Table alone saves hours. Instead of toggling between tabs or trying to hold five competitor positioning statements in your head, you have one view that shows exactly where the gaps and overlaps are.

Text & Instant Messaging

Variables: industry, briefing depth, analysis framework, strategic focus, audience.

Pain-Point Amplifier (Sales Copy)

Use case: writing the problem section of a sales page or email.

What makes it work: the output is forced across five pain dimensions: daily impact, hidden cost, emotional toll, social dimension, future trajectory. Forcing the model along multiple axes produces copy that feels understood instead of attacked. There’s also an “Internal Monologue” section capturing what the reader says to themselves about the problem. Professional copywriters charge serious money for that level of specificity. Most sales copy stops at the surface problem. This prompt digs into the social dimension (how the problem affects how others perceive you) and future trajectory (what happens if nothing changes in six months). Those two dimensions are where buying decisions actually live.

The prompt also explicitly instructs the model to write from empathy, not exploitation. That single instruction changes the quality of what comes out. The difference is copy that makes someone feel seen versus copy that makes someone feel cornered.

GTM Atlas, by Attio

GTM Atlas is a free resource every operator should read. Curated by Attio, the AI CRM, and written by GTM leaders from Lovable, Granola, and Vercel, you'll get:

  • ICP, outbound, and retention frameworks from operators who've built them

  • The qualification signals that actually predict conversion

  • Conversion plays that don't rely on a pitch deck

Mapped by operators. Curated by Attio.

*Ad

The pattern worth stealing

None of these prompts are clever. They are structural. Labeled sections. Decision trees. Evidence requirements. Explicit hedging rules. Dimensional breakdowns. If you strip out the specific use case, you are left with a template that could be adapted to a dozen other situations. That's the actual value here.

The other thing they all share is variables. Every prompt asks what kind of output you want before generating. Hard-coded prompts are single-use. Templated prompts with variables are reusable indefinitely. The SOP prompt with "healthcare, HIPAA" produces something completely different from the same prompt with "software onboarding, SOC 2." Same structure. Different output. Zero rewriting.

That's the whole shift. Stop writing prompts as one-off requests. Start writing them as templates with slots. The first version takes longer. Every version after that takes seconds.

My honest take

I keep seeing people grade their prompts by output quality on a single run. That's the wrong scoreboard. The right question is not "did this prompt work once" but "would I happily run this prompt every week with different inputs and trust the output every time?"

If the answer is no, the prompt is single-use, even if the one output looked good. Reusable prompts are an asset. Single-use prompts are a tax you pay every time the same task comes back around.

The full text of any of these four is free at PromptCreek with no paywall. The library has 1,200+ prompts, which is worth a browse if you need something beyond these.

Open one of your most-repeated prompts. Add the structure. Run it once. See what shifts.

How Jennifer Aniston’s LolaVie brand grew sales 40% with CTV ads

The DTC beauty category is crowded. To break through, Jennifer Aniston’s brand LolaVie, worked with Roku Ads Manager to easily set up, test, and optimize CTV ad creatives. The campaign helped drive a big lift in sales and customer growth, helping LolaVie break through in the crowded beauty category.

*Ad