Where is your best prompt right now
Probably in message 27 of some chat thread, sandwiched between a dozen failed attempts. You know it's there, and next time you need it, rewriting will be faster than finding it.
That's discarding an asset. A tuned prompt encodes your trial-and-error time, and trial-and-error time is the expensive part of this whole activity.
1. Break it into composable modules
Don't store prompts as monoliths. Split along the six parts and store each separately:
/prompts
/base
design-tokens-linear.md # design spec, reused across projects
seo-constraints.md # needed almost every time
a11y-constraints.md
guardrails.md
/pages
landing-saas.md
landing-local-business.md
pricing-page.md
A new project = one page template + four shared constraint blocks + a new palette.
The payoff is that one edit propagates. Find a gap in your SEO constraints, fix seo-constraints.md, and every future project benefits.
2. Give each prompt metadata
---
tools: v0 / Lovable
last verified: 2026-07
known issues: on Lovable, mobile breakpoints need one extra follow-up
sample output: ./screenshots/landing-saas-v3.png
---
"Last verified" is the critical field. Models keep changing; a prompt that worked six months ago may produce something quite different now. Without a date, you can't tell whether to trust it.
3. Version, don't overwrite
Never edit a good prompt in place. Keep v1, v2, v3, because:
- After a model update, an older version is sometimes better
- You'll want to know which edit made things worse
- Different versions suit different tools
Git is easiest; a dated folder is enough.
4. What makes a prompt worth selling
If you plan to list prompts on a marketplace, buyers aren't paying for the text. They're paying for the mistakes you already made on their behalf.
| Property | Why it's worth money |
|---|---|
| Real output preview | Lets the buyer judge — the single biggest conversion factor |
| Named target tool and version | Removes their adaptation cost |
| Built-in constraints (SEO / a11y / guardrails) | You only know to write these after being burned |
| Covers a complete scenario | "A full pricing page" beats "make buttons prettier" |
| Parameterised | Marked variables the buyer can swap for their own |
What isn't worth money: one-line generic templates, prompts freely available elsewhere, anything with no preview.
5. How to think about price
Don't price by word count. Price by the hours you save the buyer.
A prompt that lets someone skip three hours of iteration is worth a fraction of those three hours — not a fraction of its 500 words.
In practice, single prompts land somewhere between a few dollars and a few tens of dollars depending on completeness and reusability. Bundles covering a whole scenario — say eight pages of a complete SaaS site — command a clear premium over the sum of their parts.
6. Don't sell what isn't yours
One hard line: the prompt must be yours, or you must hold the right to sell it.
- Copying a public prompt and changing a few words doesn't make it original
- Prompts invoking a real brand, a public figure, or a specific company's visual identity may infringe
- Preview screenshots must be genuine output of that prompt, not someone else's work
On MotionSites these are in the Terms of Service — not advice, but a listing requirement.
The smallest first step
Open the last project you were happy with. Copy the prompt you used into a file. Add the date and the tool.
That's it. A month from now you'll have a dozen reusable assets instead of a dozen chat logs.
Want to see what curated looks like? Every prompt in the MotionSites library carries a preview, its target tool and a verification date.