Write resale listings that claim only what photos show
A developer built Honest Lister to write resale listings from photos; on 7 test items, he reports unsupported claims fell from 1.14 per listing to 0.
Done with Custom agentmoondream2
What they did
A developer built Honest Lister after a reseller had to refund an item an AI tool wrongly called leather. For each detail, the tool asks a freely available image-reading AI what it is, how sure it is and what in the photo shows it. A small checker then labels each claim as seen, guessed, unconfirmed or contradicted. It prices from similar sales or declines, and won't publish while a risky claim is unresolved.
How it went
The author reports unsupported claims fell from 1.14 per listing to 0 on 7 labeled test items. Correctly declining to guess rose from 47% to 100%. It kept 96% of the details that could truly be known from the photo.
Worth knowing
Running the small AI model on a laptop's regular processor took 285 seconds for one question, so real use needs a graphics card. The online demo uses a simpler fixed setup instead.
Try it yourself
Look at this photo of a second-hand item I want to sell: [photo]. For each attribute (category, colour, condition, brand, size, material), tell me the value, your confidence, and what in the photo shows it. Treat brand, size and material as unverified unless a label or logo is clearly visible. Then do two things. First, draft a listing using only the well-supported attributes, hedging the uncertain ones. Second, list what I must confirm manually and which extra photo would settle each point. Suggest a price range only if you can find enough comparable sold listings, and tell me the sample size. Do not post anything.Swap the highlighted [parts] for your own details.
Source: DEV Community · Rudratosh Shastri · Oct 14, 2026
Written with AI assistance from the original post and checked against our editorial guidelines. Results are as reported by the source, not independently verified.
