building-with-jev-skill
Agent skill teaching Choice, Score, Noul question design, state shape, and thresholds for jev-1.13 integrations.
Overview
building-with-jev-skill (github.com/dbreunig/building-with-jev-skill) is Drew Breunig's agent skill for writing programs that call Jev well: state shape, single-factor questions, composing answers in TypeScript or Python, confidence thresholds, and debugging low-confidence failures. Content lives in skills/jev/SKILL.md and targets jev-1.13. Install via Claude Code marketplace plugin jev@building-with-jev, npx skills add dbreunig/building-with-jev-skill, or manual copy into ~/.claude/skills/jev.
Problem: Teams new to System One keep stuffing multi-factor questions into one prompt and then wonder why thresholds feel random in production.
Built for: Claude Code, Codex, and Cursor users who want a maintained skill that teaches Choice, Score, and Noul design for jev-1.13.
First indexed on Jev Directory: 2026-09-23
Creator and team
- Name
- Drew Breunig
- Handle
- @dbreunig
How Jev is used
- Role in the product flow
- Pedagogical skill invoked on Jev integration tasks; not a runtime gate itself
- Primitives
- ChoiceScoreNoul
- State in
- SKILL.md patterns for structuring state text and splitting questions documented against TypeSafe docs links in the repo.
- Decision out
- Teaches how to read probability vectors and set thresholds in application code rather than in model prose.
- Install skill via plugin, skills CLI, or manual copy
- Claude loads SKILL.md when tasks mention Jev or TypeSafe questions
- Invoke explicitly with /jev when you want the rubric on demand
- SKILL.md references specific docs.typesafe.ai pages for primitives
This repo is meta tooling: it does not call Jev on your traffic by itself. Instead it trains whichever agent is coding your integration to stop asking should we proceed style chat prompts and start asking decomposed nouls and Choices with testable thresholds. Pair it with kerpopule-hermes-jev-skills when you want runtime routing skills, or devagrawal09-jev-review when you want a staged review dashboard. The value is consistent question grammar across teammates who might otherwise copy brittle examples from random gists.
Sourced performance claims
- README states the skill targets jev-1.13 and covers question design, state structure, thresholds, and diagnosis.Source: github.com/dbreunig/building-with-jev-skill README
- Public GitHub repo dbreunig/building-with-jev-skill had 131 stars and 5 forks when this listing was drafted.Source: GitHub API September 2026
- skills/jev/SKILL.md is the canonical skill body with install paths for Claude Code plugin and vercel-labs skills CLI.Source: github.com/dbreunig/building-with-jev-skill README
Features and stack
Features
- SKILL.md focused on System One question design
- Claude Code plugin marketplace install path
- npx skills add for Codex, Cursor, and other agents
- Manual copy instructions for ~/.claude/skills/jev
- Sources cite docs.typesafe.ai pages inside SKILL.md
Stack
- Agent skills format
- Claude Code plugins
- vercel-labs skills CLI
Pricing: Open source skill; TypeSafe usage depends on the apps you build after reading it.
Links
FAQ
- Does this skill call Jev for me?
- No. It teaches how to write calls. Your application or agent runtime still needs a TypeSafe key for live inference.
- Which Jev version does it target?
- README and SKILL.md target jev-1.13. Check SKILL.md when TypeSafe ships newer model notes.
- How do I invoke it in Claude Code?
- Install the jev@building-with-jev plugin or copy skills/jev manually, then use /jev or let Claude auto-load on Jev tasks per README.
Related learn guides
Original Jev guidance that pairs with this product pattern.
- Jev & TypeSafe
What Jev is, how System One fits your stack, and where this directory ends and TypeSafe docs begin.
Related products
Hand-picked neighbors with rich profiles or overlapping tags.