Repolode
Submit
wuyoscar

wuyoscar/jev-skill

A curated collection of demos, workflows and five installable agent skills built around the Jev decision model for coding agents.

View on GitHub
Stars
638
Forks
48
Open issues
2
Last commit
3 days ago

What is it?

Awesome Jev Skills is a collection of demos, workflows and agent skills for Jev, a model that chooses among candidate options, classifies inputs and scores them. The repository packages five skills (jev, jev-triage, jev-documents, jev-eval and jev-act) that coding agents such as Codex, Claude Code or OpenCode can install and use for tasks like stuck-loop recovery, support ticket routing, document evidence checks and browser action selection.

The repo also catalogs roughly 67 community projects and resources, from browser and game demos to local-model alternatives and evaluation reports, along with 108 scenario recipes. Each scenario gives an input shape, expected output and an editable template, and a set of recorded input/output pairs from real API calls on synthetic examples is included for reference.

The skills separate the decision step from execution: Jev returns a selection or score, while the agent or host tool performs any actual action. Setup supports either an OpenRouter key or the official TypeSafe service, with an agent-simulation mode that is clearly labeled as not calling Jev.

Key features

  • 01Five installable skills for coding agents covering triage, document evidence, evaluation and action choice.
  • 02About 108 scenario recipes with input shapes, outputs and editable templates.
  • 03Recorded input/output pairs from real Jev API calls on synthetic examples.
  • 04Setup through OpenRouter or the official TypeSafe service, with an offline dry-run check.
  • 05Catalog of community demos, local-model alternatives and evaluation reports.

Use cases

  • Recovering a coding agent that is repeating the same failing command.
  • Routing or labeling batches of support tickets with a small sample reviewed first.
  • Checking whether a claim is supported by specific source passages in documents.
  • Prototyping a rule-based decision step inside an agent workflow.

Quick start

  1. 1

    Give the install prompt to your agent

  2. 2

    Ask the agent to update later

Things to know

  • Jev is a decision model, not a standalone app, so it needs an agent to supply evidence and take action.
  • Many listed projects are community work and are not tested or official.

Topics

#ai-agents#llm#coding-agents#decision-making#agent-skills#workflow#classification

FAQ

What is jev-skill?
A curated collection of demos, workflows and five installable agent skills built around the Jev decision model for coding agents. Awesome Jev Skills is a collection of demos, workflows and agent skills for Jev, a model that chooses among candidate options, classifies inputs and scores them. The repository packages five skills (jev, jev-triage, jev-documents, jev-eval and jev-act) that coding agents such as Codex, Claude Code or OpenCode can install and use for tasks like stuck-loop recovery, support ticket routing, document evidence checks and browser action selection.
Is jev-skill free to use? What license does it have?
jev-skill is open source under the MIT license. Check the license terms before using it in commercial products.
Is jev-skill still maintained?
Actively maintained. The last commit to jev-skill was on Oct 17, 2026.
What can I use jev-skill for?
Recovering a coding agent that is repeating the same failing command. Routing or labeling batches of support tickets with a small sample reviewed first. Checking whether a claim is supported by specific source passages in documents. Prototyping a rule-based decision step inside an agent workflow.
How do I install jev-skill?
Give the install prompt to your agent Ask the agent to update later
Where can I get jev-skill?
The source code and installation instructions are on GitHub: https://github.com/wuyoscar/jev-skill

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects