Screen user-submitted listings before they go live
Ask a decision model a set of questions about each new listing's text, links and images, and send anything above a threshold to review.
Author's figure: Author's figure: 9 text and link questions, 8 image questions, review threshold 0.45 (EmDash CMS, using Cloudflare Clef). Reported by the builder, not measured by us.
The use case
Marketplaces, app stores and plugin directories accept listings from strangers, and some are scams, impersonations or offensive. This CMS asks a decision model a set of questions about the words and URLs in each new listing, and a second set about every image attached, looking for fraud, brand copying and scams. Listings that score above a fixed review threshold wait for a moderator, and the rest publish. The same design works with Jev, using its image preview for the visual checks. Moderators make every rejection, and sellers should have a way to appeal.
Example workflow
- Collect each new listing's text, links and images on submission
- Ask 8-10 Noul questions on text and links
- Ask image questions on icons and screenshots with Jev image preview or Clef
- Hold listings above the review threshold for a moderator
- Publish the rest and log scores for audits
Tools
- Jev (TypeSafe AI)
- Cloudflare Clef
- Cloudflare Workers
Risks and guardrails
- Jev image input is still a preview
- Scammers adapt wording to pass checks
- Wrong holds delay honest sellers
Sources
Written by Agents Lookup. Credits and official pages for the tools this use case names:
- EmDash moderates its plugin registry with Clef by EmDash CMS on Made with Jev: madewithjev.com/builds/emdash-clef-plugin-moderation
- EmDash moderates its plugin registry with Clef: original post or repo by EmDash CMS: emdashcms.com/blog/how-emdash-uses-clef-to-moderate-the-plugin-registry