Article
CommunityJev Practices & Patterns
Using Jev in Domain Abuse Review: A Workflow Design
ABTdomain's workflow design for brand-protection review: three layers of typed Jev questions turn each new domain that carries a brand word into Release, Watch or Flag, with 0.3 and 0.7 probability lines deciding what happens next.
- Format
- Article
- Reading time
- 3 min
- Published by
- Community
- Author
- Xiaolei Wang
- Last verified
- Oct 2026
Highlights
- Three layers: the name alone, a data and status pass, then use under UDRP; each layer asks Jev one question.
- Bins come from fixed lines: below 0.3 Release, 0.7 or above Flag, and anything between goes on to the next layer.
- On 100 apple-containing domains, Layer 1 released 18, flagged 24 and passed 58 on; applerepairzones.com scored 0.69.
- Layer 3 saw the 26 domains found in use and released 13, put 12 on Watch and flagged 1; greenapplemenu.com read 0.26.
- The 100-domain test list is on Hugging Face, and a follow-up run swapped apple for google on three names near 0.5.
Watch out
Grounded on the article text captured on 2026-10-15; the capture is truncated inside Layer 3's example table, so the Watch example and later sections are not covered. The 18/58/24 and 13/12/1 counts come from the article's own tests, and it presents the workflow as a design for one typical case where the brand keyword appears in the domain name, not a general abuse detector. No pricing or latency figures appear in the captured text.
Questions about this entry
Short answers drawn from the linked project or page.
A three-layer design from the ABTdomain team for judging new domains that carry a brand word. Where a layer uses Jev it asks one question and gives Jev only the facts that question needs; what a layer cannot settle goes on to the next, and the reviewer makes the final call.
Layer 1 asks whether the name refers to the brand, and Jev returns a probability: for applerepairzones.com the three runs gave 0.70, 0.70 and 0.68, a mean of 0.69. Layer 3 asks, drawing on the UDRP, whether the name was registered to target the brand and is used to profit from its reputation in the brand's field or a related field.
Below 0.3 a domain is released, at 0.7 or above it is flagged, and the middle band goes on to the next layer; the lines were set after several runs on different groups of names. Flagged cases are handed to a follow-up step, and Watch means the domain is looked at again when it changes.
All 100 domains contained apple and carried 95 SLDs; Jev released 18, flagged 24 and passed 58 on to Layer 2. Examples: greenapplefruitmart.com scored 0.15 as the fruit, applerepairzones.com at 0.69 did not settle it, and applewatchzone.com at 0.89 contains the Apple Watch mark.
The workflow scores the SLD, the part left of the dot, so the TLD cannot affect the score, and a name registered on several TLDs gets one score. Layer 1 also sees a short description of the brand with its marks and official domains rather than the full case file.
Of the 58 domains Layer 1 passed on, Layer 2 found 26 in use; Layer 3 released 13, put 12 on Watch and flagged 1, with greenapplemenu.com at 0.26 and aionapple.com at 0.41. The UDRP-based question is narrower than full UDRP: cases such as registering a name to sell it to the brand owner fall outside it.
Xiaolei Wang, who runs model evaluation on EuroHPC supercomputers for ABTdomain; the article is dated October 9, 2026 and updated October 10. The 100-domain test list is published on Hugging Face, and a previous article tested Jev against a small selector inside dksplit, their domain-name word splitter.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
YouTube: Kubernetes admission control with a decision model, fail-closed.
Hacker News: SafeDep Gryph plus Jev checks for compromised coding-agent behavior.
2 pointsDetecting Compromised AI Coding Agents with Jev and GryphRead the thread on Hacker NewsX: Jev and Laya on security queries
X: Testing Jev for prompt injection
More Jev resources like this
A pre-registered evaluation of Jev as a one-second, training-free screen for SCADA alarm triage in water distribution networks. On a four-class EPANET C-Town benchmark it matched a hand-written rule tree, and a gated cascade spared an LLM reviewer 35-38% of windows.
Jev Practices & PatternsArticle#community#research#arxiv
Community
Blind security benchmarks for Jev on public corpora: 662 prompt-injection messages and 200 matched vulnerable-code pairs, with raw per-sample results and a TUI dashboard.
Jev Practices & Patterns#community#go#benchmarks
Community
Gaurav-Gosain/jev-sec-benchGo
Gaurav-Gosain/jev-sec-benchGoZero-dependency reverse-proxy traffic classifier and bot mitigator shipped as Node.js and Python packages. The in-tree engine decides in tens of microseconds, and setting TYPESAFE_API_KEY elevates the gate to jev-latest for five parallel typed questions.
Jev Repos & SDKs#community#catalog-import#logicrw
Community
hemanth/traffic-guardPython
hemanth/traffic-guardPythonA GitHub Action that reviews every pull request with TypeSafe Jev and fails the check when it finds a problem. It asks four yes/no questions about each part of the diff and compares files that look duplicated, catching secrets, vulnerabilities, needless complexity and prompt injection.
Jev Integrations & Adapters#community#catalog-import#yibie
Community
turenlabs/lisaPython
turenlabs/lisaPythonA local-first Chrome extension that inventories a page's browser-visible attack surface, runs deterministic checks for mixed content, insecure forms and missing headers, and can add typed Laya or official Jev decisions through an optional gateway. Findings carry severity, confidence and evidence.
Jev Tools & CLIs#community#catalog-import#logicrw
Kitsuno's engineering write-up on training an open-weight Laya model to take over parts of its Jev usage: eight days of work, run in shadow beside Jev on every job-side request, without training on Jev's outputs.
Jev Practices & PatternsPost#community#catalog-import#madewithjev
Community
From the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.