prompt2jev
Agent skill and CLI that converts plain requirements, LLM prompts or prompt-calling code into a typed Jev request and runnable script.
Overview
prompt2jev is a skill for coding agents plus a dependency-free CLI. It takes one of three inputs, a requirement in plain words, an existing system prompt or template, or the code that calls an LLM and parses its output, and rebuilds that judgment as a TypeSafe Jev request: a state plus atomic choice, score or noul questions.
The skill tells the agent to move anything code can compute (arithmetic, dates, exact matching) into constants and code, drop formatting instructions, and keep only real judgments as questions. The CLI validates the request against the API contract, lints it against TypeSafe's documented practices, and generates a runnable program in Python or JavaScript with the official SDK, dependency-free Python, or raw curl. Five archetype requests cover routing, guardrails, rubrics, extraction and claim checks. It is aimed at developers migrating prompt-and-parse LLM steps to a decision model.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Other | Which parts of this prompt or code are judgments, and should each become a choice, score or noul question? | choice · score · noul |
| Choice | Archetype questions such as routing a request to one of several labels(inferred) | — |
How it works
- Give the agent a requirement, a prompt, or the file that calls an LLM
- The skill splits each judgment into one atomic typed question and moves computable rules into code
- The CLI validates and lints request.json against the API contract offline
- The CLI generates decide.py (or JS/curl) that sends the request and branches on probabilities
- Review the listed assumptions before automating
Strengths
- Offline validation and dry run before any paid call
- Generated code branches on returned probabilities instead of parsing generated text
- Works across several coding agents that read SKILL.md files
Caveats
- Requires a TypeSafe API key for real calls
- Converted questions still need human review of the stated assumptions
Getting started
claude plugin marketplace add sumleo/prompt2jev && claude plugin install prompt2jev@prompt2jev, or paste the install prompt from docs/install.md into your agent.At a glance
- Type
- Agent skill
- Status
- beta
- Language
- Python
- Runtime
- Agent skill + CLI
- Has code
- yes
- Has tests
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
Agent skill and CLI that turn natural language, an LLM prompt, or the code that runs one into a TypeSafe Jev decision: typed state, Choice/Score/Noul questions, and a runnable script
logicrw/awesome-jev-projectsAgent skill and CLI that turn natural language, an LLM prompt, or the code that…
wh000wh000/awesome-jev-liveAgent skill and CLI that turn natural language, an LLM prompt, or the code that runs one into a TypeSafe Jev decision: typed state, Choice/Score/Noul questions, and a runnable script.
hellogumbo/awesome-jev
Frequently asked questions
What does prompt2jev produce?
A validated request.json with state and typed questions, a script that sends it and branches on the answers, and a list of assumptions to confirm.
Do I need to write JSON myself?
No. The README says you paste a usage prompt into your agent and the skill and CLI build the request.
Does it need a TypeSafe key?
Only for real calls; installation and the dry run are offline and make no paid call.
Related projects
Unofficial hosted or local Jev MCP server giving agents a quick second opinion: it picks yes/no, scale or choice and returns probabilities.
MCP serverbetaTypeScriptSwift library and CLI for Jev with enum-typed Choice, Noul and Score questions, plus an agent skill for coding agents.
CLIproductionSwiftPython MCP server wrapping Jev as classify, score, check, match and screen tools with confidence on every answer.
MCP serverbetaPythonPython decision gateway and stdio MCP server exposing Jev via OpenRouter's alpha Decisions endpoint to Claude Code, Codex and Cursor.
MCP serverbetaPythonGo MCP server giving any client typed Jev decisions via OpenRouter or TypeSafe, with validated answers and no paid retries.
MCP serverproductionGoPython FastMCP server wrapping Jev's noul, choice and score judgments, with configs for many coding agents and an OpenCode Zen route.
MCP serverbetaPython