Skip to content
Metatext Decide — typed decisions for your agents, with a probability on every answerGet early access
Metatext
Sign Up

jev-agent-skill

Chinese-language agent skill that offloads classify, screen, score and compliance checks to Jev via OpenCode Zen's free tier.

PluginbetaMIT
View on GitHub ↗

Overview

jev-agent-skill is an agent skill, documented in Chinese, that moves high-frequency small judgments (classification and routing, batch screening, scoring, and compliance checks) from the main coding model to Jev, accessed through OpenCode Zen's free tier. It works with Claude Code, ZCode and any agent that has a skills directory, and the agent calls it automatically based on trigger rules in SKILL.md.

It bundles a zero-dependency jev.py caller hardened against problems the author hit in practice, such as transient HTTP 500s from the gateway, WAF-safe user agents and GBK-safe stdin piping, and a production pipeline for triaging comments from a Taobao shop that keeps raw items out of the agent's context. The README reports 10+ real calls on 20 September 2026 with 283-694 input tokens, zero cost and about 2 s end to end, four parallel questions working, and Chinese state and criteria handled correctly, with a benchmark report against the main model. It is adapted from TypeSafe's official skills (MIT) and is MIT licensed.

What the decision model decides

TypeQuestionAnswers
ChoiceWhich category or recipient does this ticket, comment or log line belong to?—
Yes / noIs this item worth a closer look?yes · no
Yes / noIs this statement supported or this output compliant?yes · no

How it works

  1. Install the skill in the agent's skills directory with a Zen key
  2. Agent recognises a classify/screen/score/check task
  3. jev.py sends the state and questions to Jev
  4. Only conclusions return to the agent's context

Strengths

  • Raw batch items stay out of the main model's context
  • Free-tier access per the README
  • Retry and encoding fixes from real use

Caveats

  • Depends on OpenCode Zen's free tier remaining available
  • Documentation in Chinese

Getting started

Put a Zen key in ~/.jev/zen.key, `git clone https://github.com/yuyang2230/jev-agent-skill.git ~/.claude/skills/jev`, then pipe a test request into jev.py.

At a glance

Type
Plugin
Status
beta
Language
Python
Runtime
agent skill
Has code
yes
License
MIT

Use cases and integrations

How curated lists describe it

  • Developer tooling: Claude Code/ZCode skill that offloads classify/route, batch-screen, score, and compliance-check judgments to Jev via OpenCode Zen's free tier, bundling a zero-dependency jev.py caller (transient-500 retry, WAF-safe UA, GBK-pipe-safe stdin) and a production Taobao-shop comment-triage pipeline that keeps raw items out of the agent context.

    yibie/awesome-jev

  • Claude Code/ZCode skill that offloads classify, screen, score, and compliance-check judgments to Jev via OpenCode Zen's free tier; ships a retry-hardened zero-dependency caller and a shop comment-triage pipeline.

    cobanov/awesome-jev

Frequently asked questions

What does the skill offload to Jev?

Classification, batch screening, scoring and compliance checks.

Why does it save tokens?

Batch items stay in the script and only short conclusions enter the agent's context.

How is Jev accessed?

Through OpenCode Zen's free tier, per the README.

Related projects

  • jev-architect (BenjaminPolge)

    Skill that makes Claude Code and Codex split AI features into Jev, LLM or plain-code steps before writing code.

    Projectbeta
  • jev-skill-gate

    Scores installed Claude Code skills against the current project with Jev and hides irrelevant ones via skillOverrides.

    Agent skillbetaJavaScript
  • jevskill

    Agent skill that has Jev pick the few relevant lines from big logs before your coding agent reads them, with a reversible reduce and ledger.

    ProjectbetaPython
  • codex-gemini-jev-bridge

    Local MCP bridge letting Codex hand drafting to Gemini via a Google AI Pro quota, with Jev checking drafts against explicit criteria.

    MCP serverbeta
  • jev-browser-wingman

    MCP wingman that hands an agent's browser clicks and typing to Jev on a real logged-in Chromium, alongside Playwright MCP.

    MCP serverbetaTypeScript
  • Oko

    Local code search for coding agents: ripgrep/BM25 shortlist, then Jev judges each chunk's relevance, served over MCP.

    MCP serverbetaRust