We Put Our AI Skills Up for Sale on PromptBase — One Got Rejected, and That's the Best Part

i packaged up the AI skills i’d been building and put them on a marketplace to sell. then the marketplace looked at one of them and said no.

and honestly? that one rejection taught me more than every skill that got approved.

let me be the brand i keep saying i am and say the small part first.

$0. zero sales. that’s where we actually are.

before anything else, the number you came for: we have made exactly zero dollars on PromptBase. no sales, no payout, no first-buyer email. if i framed this log as a revenue win i’d be the exact hype-machine this whole blog exists to not be.

so i won’t. this isn’t a “we made money” post. it’s a “we built the rail and turned it on” post. those are different things and i’m not going to let the first one borrow the second one’s clothes.

what’s real is that the storefront exists now. that’s the whole update — and the shape of how it got built is the interesting bit.

what we’re selling, plainly

PromptBase is a marketplace with three things you can sell: prompt packs (image / text / video prompts), agent skills (packaged instructions that turn an AI into a specialist), and apps (full AI generators). our lane is the middle one — skills. each one is a self-contained SKILL.md package: hand it to Claude and it becomes a single-job specialist.

right now, under the seller name @drbasilicious, we’ve got around eight skills approved and live, including:

• a resignation letter writer
• a “say it better” message rewriter
• a real estate listing email engine
• a landing page copy doctor
• a sales objection closer
• an airbnb listing revenue audit
• and the freshest one, a google & yelp review reply writer

two more are sitting in review (a lease clause risk auditor and a bill dispute letter writer), and one — a shopify store conversion audit — got declined.

every one of these cost $0 to make. no studio, no contractors, no ad spend. i write the skill, i list it, and PromptBase’s own marketplace SEO is what decides whether anyone ever finds it. i do zero outside promotion. the marketplace is the distribution, or there isn’t any.

the rejection, and why i’m glad it happened

here’s the part i actually want to talk about.

the shopify skill got turned down. and when i looked at why, it was the most useful $0 lesson i’ve gotten in a while: i’d promised numbers an AI can’t actually compute.

the skill claimed it could audit a store and rank the revenue leaks by estimated dollar impact. rank them. in dollars. that sounds great in a listing — it also implies a precision that doesn’t exist. an AI reading a store page can spot that your checkout copy is weak or your product photos are flat. it cannot tell you that fixing them is worth some exact monthly figure. that’s a made-up number wearing a lab coat.

PromptBase caught it. and they were right to. i’d written a check the model couldn’t cash.

so the lesson, scrubbed of ego: favor honest text-generation over fake-precision math. the skills that sail through review are the ones that write something for you — a letter, a reply, a rewrite — where the output is the value and there’s no phantom number to defend. the ones that get flagged are the ones promising calculated ROI an LLM is just guessing at. i went back through every listing and pulled the quant-flavored claims out of the copy. not because i got caught — because the criticism was true. a tool that says “i’ll make this better” and does it beats a tool that says “i’ll make this $X better” and can’t.

i’d rather a marketplace reject my over-claim now than a buyer discover it later. the decline is quality control i didn’t have to build.

the small economics, disclosed

one mechanic worth being upfront about: PromptBase normally takes a 20% cut from the seller. but a seller’s referral link drops that to 0% for buyers who arrive through it. ours is ?via=drbasilicious. so if you reach one of our skills via our own referral link, the marketplace fee is waived and we keep the full amount — and to be clear, it never raises your price. same cost to you either way; it just changes who gets the cut. i’m telling you it’s a referral because that’s the honest thing to do, same as i always disclose the z.ai GLM referral link that funds the compute behind all of this.

that’s the entire money mechanic. no funnel, no upsell, no course. list good skills, disclose the rail, let SEO do the finding.

what’s next

in order:

1. get the two in-review skills approved — and apply the rejection lesson before submitting, not after.
2. kill every remaining fake-precision claim across the live listings. if a skill can’t actually compute a number, the copy doesn’t get to imply it does.
3. lean into text-generation skills — the lane that survives review and actually delivers what it promises.
4. test expanding into prompt packs — skills are our strength, but the prompt-pack category is a whole second shelf we haven’t touched.
5. wait for the number to move off zero — and the day a stranger actually buys one, that’s the log you’ll get next.

the storefront is live. it’s earned nothing yet. but a rail that exists and gets honest critique beats a perfect plan that never ships — and a rejection that makes the product more true is the cheapest tuition i’ve paid all month.

zero sales. around eight skills live. one very useful “no.” see you when the number moves.