Building with TypeSafe Jev plugin
An unofficial skill teaching agents to design Jev calls: typed questions, calibrated confidence and examples with code sketches
- 155 ★ on the repository
- Last commit
- MIT
Minimal access
Why this level
- The skill is reference instructions and runs no commands itself
- Examples push you to send text to a third-party model with your key, so watch what data leaves
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/building-with-typesafe-jevDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- The skill is reference instructions and runs no commands itself
- Examples push you to send text to a third-party model with your key, so watch what data leaves
This is third-party code. Review the repository files before installing.
What it does
Building with TypeSafe Jev is an unofficial community skill about working with the Jev model from TypeSafe AI. Jev belongs to the System One class: it does not write text. You send it content and a set of typed questions, and it answers each with a value and a calibrated probability, usually in 100–200 ms. There are three question kinds: Choice picks one option from a list (route a ticket to billing, shipping or support), Score places content on a scale you describe (how closely a pull request meets the spec) and Noul gives the probability that a yes/no statement is true. The problem the skill solves: agents treat Jev as one more chat model. The skill teaches them to design for it: where rules and regex are not enough and a judgment about text is needed (classify, route, moderate, score, rank, dedupe, extract, verify, gate an action). It collects links to 150-plus community projects sorted by how they work, with a code sketch for each pattern. The skill installs in Claude Code, Codex, Antigravity CLI, Muse and Muse Code. For Claude Code: `claude plugin marketplace add aaddrick/building-with-typesafe-jev` and `claude plugin install building-with-typesafe-jev@building-with-typesafe-jev`.
Who it is for. Developers who embed fast text decisions in an app: ticket routing, moderation, scoring and gates before an agent action.
Good fit when
- The code has an LLM call only to produce a label, score or yes/no
- Support tickets must be sorted by topic in fractions of a second instead of seconds of generation
- Before a dangerous agent command you need a fast check on whether it deletes files outside the project
- You want to find where in the app an AI judgment really improves the result
Not a fit when
- You need the official vendor skill: this one is community-made and the official one is in the typesafe-ai/skills repository
- You need text or code generation: the Jev model answers with values and probabilities
- You need a decision that plain rules or regex handle: an AI call is redundant there
Example request
Split incoming tickets into billing, shipping and support with calibrated confidence and show how to do it through JevLimitations
The skill is unofficial: it was not created, reviewed or endorsed by TypeSafe AI. It teaches how to design calls and does not itself grant access to Jev, so you need your own key and a service account. The model and prices are set by the vendor, so check them in the docs. Links to 150-plus community projects lead to third-party repositories, and the skill author is not responsible for their content.
The skill installs as a plugin and fires on the described signals: a judgment about text, brainstorming where AI fits in an app or mentions of TypeSafe AI, Jev, System One or the SDK. The repository has a set of checks in `evals/` and a results ledger.
How to disable. In Claude Code remove the plugin with `claude plugin uninstall building-with-typesafe-jev` and the marketplace entry. For other agents delete the skill folder from their directory.
FAQ
What is Jev?
A System One class model from TypeSafe AI. It answers typed questions with a value and a calibrated probability, usually in 100–200 ms, and generates no text.
Which question types exist?
Choice picks an option from a list, Score places content on a described scale and Noul gives the probability of a yes/no statement.
How do I install the skill in Claude Code?
Run `claude plugin marketplace add aaddrick/building-with-typesafe-jev`, then `claude plugin install building-with-typesafe-jev@building-with-typesafe-jev`.
How does it differ from the official skill?
TypeSafe AI publishes the official skill in the typesafe-ai/skills repository. This one is community-made and adds a project catalog and design patterns.
Which agents does it work with?
Claude Code, Codex, Antigravity CLI, Muse and Muse Code. Muse has an install script and Antigravity has the `agy plugin install` command.
Discussion
No questions or reviews yet.
Related
PluginsA skill for building with TypeSafe: typed questions to System One models with calibrated probabilities, not prompt parsing.
A plugin with a local MCP server: the agent scores its own code on a set of criteria through Jev and improves it by the numbers
A Claude Code plugin that trims noisy Bash output with the Jev model before the result reaches the main model, without a summary
Jevgrep
jevgrep
A CLI for coding agents: finds code by a task description rather than by a string. A key from one of the model providers is needed
Jev-Code-Reviewer ranks changes in an agent PR as P0, P1 or P2 and retells them in plain language on the GitHub page.
JevGrep finds code from a description of what it does and returns lines with paths, as a CLI and as an agent MCP server.
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| Building with TypeSafe Jev (this card) | no | 155 | ||
| TypeSafe Agent Skills | no | 2,895 | the official skill set from TypeSafe AI, with no community project catalog | |
| awesome-jev | no | not required | 602 | a link list of Jev projects without a skill for the agent |
README in short
The repository contains the skill and plugin manifests for Claude Code and Codex. The author places a notice of unofficial status and a link to the official skill before the install section. The README is translated into several languages.
README badge
Original SKILL.md
---
name: building-with-typesafe-jev
description: Use when code needs a judgment about natural language that rules or regex cannot make (classify, route, triage, moderate, score, rank, match, dedupe, filter, extract, verify, or gate an action), when brainstorming where AI could fit in an app, when an LLM call exists only to pick a label, score, or yes/no, or when code uses TypeSafe AI, Jev, System One, typesafe-sdk, or @typesafe-ai/sdk.
---
# Building with TypeSafe Jev
## Overview
Jev is a **System One model**: it takes a `state` plus named, typed questions and returns typed answers with calibrated probabilities. It never generates text. **Code owns the workflow. Jev supplies narrow snap judgments.** It is not a chat or coding LLM and cannot power a coding agent.
**Before you design anything new, check prior art:** open `prior-art/INDEX.md`. It maps intents (control loop, gate, rerank, stream filter, incremental, agent memory, LLM pairing, and more) to shape files. Each shape file has a code sketch, field lessons, and linked community projects. Or grep `prior-art/` for a domain word. It also lists the known bad fits.
Exact request/response shapes, SDK signatures, limits, and errors: see `api-reference.md`. It ends with a map of which live docs page to read for which task. Patterns and the cookbook index: see `patterns.md`. The live docs win on conflict: fetch `https://docs.typesafe.ai/llms.txt`, and append `.md` to any page path. Do not guess field names. Check them there. If the docs cannot be reached, read the installed SDK's types, tell the user you did, and do not invent details that depend on the version.
**Open-ended request** ("where could AI help in this app?"): work backward from what the app should show, select, change, or hand off. Offer two or three directions from `prior-art/INDEX.md` and recommend one. **Concrete request:** pick the shape and build. Either way, keep the user's stack and scope, and add Jev only where code needs a judgment.
## Test against the live API when you can
You get better results when you check a design against the real endpoint. A live call catches a wrong field name, and it shows when Jev reads a question differently than you meant. If `TYPESAFE_API_KEY` is set in your shell, run a small probe before the code reaches the project. One call costs a fraction of a cent (input tokens only, 4.2 cents per million). If the key is not set, design from `api-reference.md` and say that you did. Do not go looking for the k