Jev Skills
Five skills for the Jev model: sorting records, source checks, rubric scoring and choosing an agent's next action.
- 668 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Real mode sends data to an external API and costs money; keys are stored locally.
Install
- Works in
Manual install
Install Jev Skills for this coding agent, including all scenario skills: https://raw.githubusercontent.com/wuyoscar/jev-skill/main/docs/install.md Check my environment and handle the installation. Use jev for setup to confirm with me: A: real Jev via my OpenRouter or official TypeSafe account; B: simulation with this agent. Wait for my choice. Guide any key entry through local secret settings, never this chat. Verify offline first; ask before sending data or making a paid call.
The first README option: a prompt for Codex, Claude Code or OpenCode. The agent checks the environment and installs the skills.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Real mode sends data to an external API and costs money; keys are stored locally.
This is third-party code. Review the repository files before installing.
What it does
Jev Skills is a collection of five skills that teach an agent to call the Jev model for judgments: label tickets, check a claim against a source, score an answer against a rubric, choose a legal next action. The agent gathers evidence and acts, Jev returns a label or a score. The repository holds 108 scenarios with request templates and 14 recorded input/output pairs. Real API mode uses an OpenRouter or TypeSafe key, simulation mode needs no key: the agent itself plays Jev.
Who it is for. Builders of agent pipelines who need bulk sorting, scoring and checking handled by a separate cheap model call.
Good fit when
- A support queue is split by urgency and topic.
- An agent checks claims in a report against source documents.
- Outputs of several models are scored with one rubric.
- A browser agent picks the next legal step on a page.
- A plain-language idea becomes ready-made questions for Jev.
Not a fit when
- The task is solved by plain code without a model: a field filter or a regex.
- There is no OpenRouter or TypeSafe key and no wish to use simulation mode.
- Company policy forbids sending data to an external service.
Example request
Triage this support queue by urgency with Jev, list borderline cases separately and do not send data without my confirmation.Limitations
Real mode costs money: Jev API calls are paid, and the agent must ask before sending data. Python 3.10+ and network access are required. The projects section of the README describes third-party work that is neither installed nor tested here. Jev labels remain model judgments, so confident and agreeing answers also deserve spot checks. The five-skill set is described as a source preview, while the published v0.2.0 release has 11 entry points.
How to disable. Delete the `jev*` skill folders from the project or agent skills directory. Remove OpenRouter and TypeSafe keys from local settings.
FAQ
Which skills are included?
Five: `jev` for questions and batch calls, `jev-triage` for sorting, `jev-documents` for source checks, `jev-eval` for rubric scoring and `jev-act` for action choice.
Is an API key required?
Real Jev needs `OPENROUTER_API_KEY` or `TYPESAFE_API_KEY`.
Is Node.js required?
No. According to the README the default install route needs neither Node nor npm, and Python 3.10 or newer is required for API mode.
Where are the examples?
In the catalog of 108 scenarios: each has a request template, and some are shown with a recorded result.
What if there is no key?
Choose simulation mode: the agent plays Jev itself. The result is approximate but free and sends no data out.
Discussion
No questions or reviews yet.
Related
SkillsThirteen skills hand an agent's small decisions to the cheap Jev model: model, skill, memory passages and next step.
The Jev classifier as an agent tool: labels items, answers yes or no and ranks candidates with probabilities within milliseconds.
Jev-Style runs small decision models on your machine: a probability per option, six skills, MCP and a Claude Code guard hook
A local router merges models, subagents, skills and MCP tools into one candidate set and picks who handles the next step.
A plugin with a catalogue of public Jev examples: a skill and an MCP server show how others solved the same decision task.
jeo-skills is a skill catalog with a router that searches, previews and installs only your picks into Claude Code and Codex.
README in short
Jev chooses, classifies and scores while the agent supplies evidence and acts. The collection gathers scenarios, templates and a list of third-party projects around the model.