Yaxin9Luo/spending-effort-with-jev/tree/main/plugins/spending-effort-with-jev
spending-effort-with-jev
在发送每条消息的瞬间显示它需要哪个 /effort 级别,由 TypeSafe 的 Jev 判断,并在你同意后切换。基于《Using Claude Code: Spending your effort》。非官方。
关于这个 mod
spending-effort-with-jev
在发送每条消息的瞬间显示它需要哪个 /effort 级别,并在你同意后切换。TypeSafe 的 Jev 模型会读取你的消息,并针对 Claude 即将运行的级别给出一行判定(⬆ needs high、⬇ low is enough、✓ medium fits)。如果级别看起来不对,无论过高还是过低,对话框(ask_first)或提示列上方的横条(1:切换,2:保留)都会为该消息的工作阶段切换级别。它不会自行切换。
需要 Claude Code 2.1.287 或更高版本,以及 TypeSafe API key;启用插件时输入,并保存在安全存储中。
选项:language(en / zh)、quiet、ask_first、log_decisions。
完整文档、截图与评估:https://github.com/Yaxin9Luo/spending-effort-with-jev 隐私:https://github.com/Yaxin9Luo/spending-effort-with-jev/blob/main/PRIVACY.md
MIT 许可。独立项目,与 Anthropic 或 TypeSafe 无关联。
安装
请先查看作者 README 确认 marketplace 和插件名称;命令可能随仓库结构改变。
claude plugin marketplace add Yaxin9Luo/spending-effort-with-jev claude plugin install spending-effort-with-jev
原文 / README
spending-effort-with-jev
Shows which /effort level each message needs, the moment you send it, and switches on your OK. TypeSafe's Jev model reads your message and gives a one-line verdict against the level Claude is about to run on (⬆ needs high, ⬇ low is enough, ✓ medium fits). When the level looks wrong, in either direction, a dialog (ask_first) or a band above the prompt (1: switch, 2: keep) switches it for that message's turn. It never switches on its own.
Needs Claude Code 2.1.287 or later and a TypeSafe API key, entered when you enable the plugin and kept in secure storage.
Options: language (en / zh), quiet, ask_first, log_decisions.
Full documentation, screenshots and evaluation: https://github.com/Yaxin9Luo/spending-effort-with-jev Privacy: https://github.com/Yaxin9Luo/spending-effort-with-jev/blob/main/PRIVACY.md
MIT licensed. Independent project, not affiliated with Anthropic or TypeSafe.

