MCP.so
登录

lever官方

github / lever

基本信息

分类

开发与编程

代码仓库

github / lever

Skill

lever

关于 lever

GitHub's official Copilot plugin of feedback-loop agent skills: /lever turns a vague complaint into a measurable Loop Card, runs parallel experiment arms, writes unattended briefs, shepherds PRs to green, reviews with several models and sets how much autonomy agents get.

#github-copilot#feedback-loops#agents#automation

安装 lever

copilot plugin marketplace add github/lever && copilot plugin install lever@lever

如何使用 lever

  1. 1

    在 GitHub Copilot(CLI 或桌面应用)中安装插件:先添加 github/lever 插件市场,再安装 lever@lever。

    copilot plugin marketplace add github/lever
  2. 2

    然后安装插件本身。

    copilot plugin install lever@lever
  3. 3

    新开一个会话,先试一个只读命令,例如:/lever loop our test suite feels slow. just the card, don't change anything。你会得到一张 Loop Card:衡量指标、测量方法、噪声大小和怎样算有效改进;不会推送或合并任何东西。

  4. 4

    运行 /lever setup 回答三个问题并设置参数:自主程度(L0 只提建议,L2 把 PR 跑绿后由你合并,最高 L5)和投入档位(standard、high 或 max)。

  5. 5

    按目标使用命令:brief、deliver、refactor、shepherd、land 用于构建;loop、arms、review 用于改进;automate、health 用于让任务自动运行;mirror、steer、autonomy 用于按你的习惯调整 lever。也可以在 /lever 后直接描述问题,由它挑选命令。

功能与用途

  • 16 skills behind one /lever command: brief, deliver, refactor, backlog, shepherd, land, status, loop, arms, review, automate, health, mirror, steer, autonomy, spend

  • Read-only verbs such as loop return a Loop Card (metric, measurement, noise, win bar) without pushing anything

  • Autonomy dial L0-L5 decides who merges; spend standard/high/max sets model tier and parallel lanes

  • Install: copilot plugin marketplace add github/lever, then copilot plugin install lever@lever

  • Work is tracked as GitHub issues with the lever label; automations install paused and dry-run first

  • MIT licensed, from GitHub (created 2026-10-25)

要求与注意事项

  • 它是什么:16 个 skill(每个命令一个),来自 GitHub 内部最积极使用 AI 的工程师的工作方式,统一封装为 /lever 命令。像 loop 这样只读的 skill,也会在你的需求匹配时自动加载。
  • 要求:支持插件的 GitHub Copilot。若要让 Copilot 云端编码智能体也能使用,可按仓库安装:把 skills/ 以及 docs、templates、config、automations 复制到 .github/ 下。
  • 限制:任何推送、提交 issue、开 PR 或定时运行的操作都受你的自主程度设置和仓库分支保护约束;自动化在安装时默认暂停,并会先在历史事件上试运行,之后才开启。
  • lever 发起的工作会以带 lever 标签、分配给你的 GitHub issue 形式跟踪,因此 /lever status 可以列出所有进行中的任务。
  • MIT 许可;仓库创建于 2026-10-25。

参考来源

已于 2026-11-11 对照源仓库核实。

评论

开发与编程分类下的其他 Skills

vercel-react-best-practices

vercel-labs / agent-skills

React and Next.js performance optimization guidelines from Vercel Engineering. This skill should be used when writing, reviewing, or refactoring React/Next.js code to ensure optimal performance patterns. Triggers on tasks involving React components, Next.js pages, data fetching, bundle optimization, or performance improvements.

improve-codebase-architecture

mattpocock / skills

Scan a codebase for deepening opportunities, present them as a visual HTML report, then grill through whichever one you pick.

shadcn

shadcn / ui

Manages shadcn components and projects — adding, searching, fixing, debugging, styling, and composing UI, including chat interfaces. Provides project context, component docs, and usage examples. Applies when working with shadcn/ui, component registries, presets, --preset codes, or any project with a components.json file. Also triggers for "shadcn init", "create an app with --preset", or "switch to --preset".

caveman-compress

juliusbrussee / caveman

Compress natural language memory files (CLAUDE.md, todos, preferences) into caveman format to save input tokens. Preserves all technical substance, code, URLs, and structure. Compressed version overwrites the original file. Human-readable backup saved as FILE.original.md. Trigger: /caveman-compress FILEPATH or "compress memory file"

caveman-stats

juliusbrussee / caveman

Show real token usage and estimated savings for the current session. Reads directly from the Claude Code session log — no AI estimation. Triggers on /caveman-stats. Output is injected by the mode-tracker hook; the model itself does not compute the numbers.

full-output-enforcement

leonxlnx / taste-skill

Overrides default LLM truncation behavior. Enforces complete code generation, bans placeholder patterns, and handles token-limit splits cleanly. Apply to any task requiring exhaustive, unabridged output.