dbreunig/building-with-jev-skill
User repository on GitHub
A skill for writing and improving programs that call Jev, TypeSafe's System One model
- Forks
- 12
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Practices & Patterns
Building with Jev
An agent skill for writing and improving programs that call Jev: question design, state structure, answer composition, confidence thresholds, and how to diagnose a question that keeps coming back wrong.
- Stars
- 155
- Reading time
- 3 min
- Published by
- Community
- Author
- dbreunig
- Last verified
- Sep 2026
About this project
This repo holds an agent skill for writing and improving programs that call Jev, TypeSafe's System One model. The skill covers question design (Choice, Score, Noul), state structure, answer composition in code, confidence thresholds, and diagnosis of questions that answer wrong or with low confidence. It targets jev-1.13. Run these two commands inside Claude Code: The skills CLI installs the skill into Claude Code, Codex, Cursor, and other agents: Copy the skill folder into your personal skills directory:
Quoted from the project's README. Read the full README.
Highlights
- The skill file lives at skills/jev/SKILL.md and targets jev-1.13.
- Covers question design, state structure, answer composition, and confidence thresholds.
- Explains how to diagnose questions that answer wrong or with low confidence.
- Installs through the Claude Code plugin marketplace, npx skills, or a manual copy.
Questions about this entry
Short answers drawn from the linked project or page.
Just instructions. The whole payload is one file, skills/jev/SKILL.md, and its guidance is drawn from the TypeSafe documentation with the specific pages listed. There is no runtime, no client code and no key handling in the repo.
The skills CLI (npx skills add dbreunig/building-with-jev-skill) installs it into Claude Code, Codex, Cursor and other agents. There is also a Claude Code plugin route through its marketplace and a manual copy into ~/.claude/skills/jev.
That is one of the things it covers: diagnosing questions that answer wrong or return low confidence, alongside question design for Choice, Score and Noul, state structure, composing answers in code, and choosing confidence thresholds.
It targets jev-1.13 specifically and teaches TypeSafe's primitives only, so behaviour on newer model versions is not covered and none of it transfers to other System One models without rework. It also loads on its own whenever a task mentions Jev or TypeSafe, which can be more help than you want mid-conversation.
One file: skills/jev/SKILL.md. The repository is described as holding an agent skill, and every route to it points at that same path, whether through the plugin marketplace, the skills CLI or a manual copy. There is no package to build, no runtime dependency, no CLI and no key handling anywhere in the repo.
Trace it. The README says the guidance draws on the TypeSafe docs and that SKILL.md lists the specific pages, so any piece of advice can be compared against the source it came from. Since the skill targets jev-1.13, a newer model version is the first thing to re-check, and none of it transfers to another System One model without rework.
Yes. It is one Markdown file, so you can open skills/jev/SKILL.md on GitHub, or clone the repository and read it locally. Installing is only needed if you want an agent to load it: the manual route copies the folder into ~/.claude/skills/jev, while the skills CLI installs it into Claude Code, Codex, Cursor and other agents at once.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: Copy and paste, with a decision in between
X: Jev Calc: a notebook that calculates anything
X: dmx.to reshapes the timeline in plain language
X: A canvas you control by pointing and speaking
More Jev resources like this
Claude Code, Codex, and pi plugin plus MCP server and npm library that hands steps needing no text output to Jev: one batched call answers typed questions about a state, and an opt-in PreToolUse gate can only deny or ask.
Jev Tools & CLIs#community#javascript#claude-code
Community
shitianfang/jev-useJavaScript
shitianfang/jev-useJavaScriptAn open lab for using Jev correctly next to Claude Code: a patched jev-router with per-turn model routing, a health-gated fallback with alerts, and write-ups of two silent bugs and of where tokens actually go.
Jev Practices & Patterns#community#catalog-import#logicrw
Community
pasblinn/jev-labTypeScript
pasblinn/jev-labTypeScriptAn MIT-licensed case study with two working Python programs pairing Jev and Claude Code: a support-triage app with Jev inside it, and a Claude Code safety hook where Jev judges the agent's own shell commands. Tests and evals run offline, free.
Jev Practices & Patterns#community#github#python
Vercel's Pranit reports benchmarking the fx auto-mode safety classifier with Jev: about 5 to 18 times faster and more accurate than GPT-5.6 Luna.
Jev Practices & PatternsPost#community#x#benchmark
Community
A widely read counterpoint to Jev-based context compaction: compaction is reconstruction, not filtering, and scoring tool calls to drop history misunderstands context management.
Jev Practices & PatternsPost#community#x#compaction
Community
A teams-first multi-agent orchestration plugin for Claude Code with opt-in Jev hooks for sub-millisecond judgment points, decision caching and per-point egress controls. TypeScript, MIT, with a staged team pipeline and tmux CLI workers.
Jev Tools & CLIs#community#claude-code#orchestration
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.