agentsclimarketplace

Greenroom

Skill YunyueLi/greenroom/skills/greenroom

Open-source AI workbench for interview prep: resume + JD to role intel, story bank, scripts, mock interviews, and live copilot.

Install
npx -y skills add YunyueLi/greenroom --skill greenroom

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 4 stars4 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Interview preparation workbench entry point. Runs the full pipeline from resume + target company + JD to a complete prep workspace (research, experience bank, verbatim scripts, industry brief), or routes to a single step. Use when the user says they have an interview coming up, wants interview prep, uploads a resume with a target job, mentions 面试准备 / 我要面试了 / 帮我准备面试 / 帮我把面试材料全做出来 / 一条龙, or asks "where do I start". Do NOT use for writing resumes from scratch, salary negotiation after offer, or live in-interview assistance.

The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

5.1 KB, as published. Nobody here has run it

Greenroom · 候场

入口 skill。两种用法:全流程(给齐简历 + 目标公司/岗位/JD,一次生成全套备战材料)或单步路由(识别用户处在哪一步,转给专项 skill)。

全流程(推荐的默认路径)

用户给出(缺什么问什么,一次问全):

  1. 简历(文本或文件路径)
  2. 目标公司 + 岗位名 + JD 原文(或链接)
  3. 表达风格偏好(可选:自己的说话样本/过往面试转写,或一句描述,如"沉稳、少修辞";写进 profile.md 的「风格偏好」节,逐字稿按它调)

然后按序执行,每步落盘后只汇报一行进度,不中途长篇输出:

调用产出
1初始化工作台(见下)profile.md 骨架
2job-inteljobs/<slug>/intel.md(公司与面试官调研、JD 逐条对照、考题预测)
3story-bankstory-bank.md(经历卡,含本岗讲法)
4industry-brieflibrary/<行业-岗位>-通识.md(行业与岗位参考阅读)
5interview-scriptjobs/<slug>/script.md(可朗读逐字稿)

收尾:汇报生成清单 + 提示打开控制台 app/greenroom.html 阅读;建议面试前跑一次 mock-interview、面试后 24 小时内跑 debrief。这套工作台同时是实时提词后端的取数源(逐字稿 + 调研,见 docs/realtime-bridge.md)。

六步方法(单步路由用)

#步骤做什么专项 skill产出文件
1岗位调研JD 拆解、公司/面试官调研、考题预测job-inteljobs/<slug>/intel.md
2经历库把真实经历整理成经历卡,按岗选讲法story-bankstory-bank.md
3逐字稿口语化、可直接朗读的答题稿interview-scriptjobs/<slug>/script.md
4行业通识行业格局、岗位画像、高频概念参考阅读industry-brieflibrary/*.md
5模拟面试按面试官背景出题、连续追问、评分mock-interviewjobs/<slug>/rounds/mock-N.md
6面后复盘复原问答、沉淀本轮要点debriefjobs/<slug>/rounds/rN-debrief.md

贯穿所有步骤的底线:生成内容只用候选人真实给出的经历和数字,数字要带得出出处,给不出准确数字就用定性说法,绝不编。

找到或创建工作台

工作台是一个本地文件夹,格式契约见仓库 docs/workspace-spec.md。判定顺序:

  1. 当前目录或其子目录里找 profile.md + jobs/(或 frontmatter 带 type: profile 的文件)→ 找到即用,向用户确认一句。
  2. 找不到 → 问用户工作台在哪;用户说没有 → 走初始化。

初始化流程

  1. 问清两件事:工作台建在哪个路径;候选人简历(贴文本或给文件路径)。
  2. 创建目录骨架:
<workspace>/
├── profile.md
├── story-bank.md
├── library/
└── jobs/
  1. 从简历提取生成 profile.md(frontmatter type: profile):履历线(时间倒序)、教育背景、关键数字;用户给了风格偏好就加「风格偏好」节。提取数字时把出处一并记在 profile.md 里(哪份报表/哪次复盘),后续逐字稿要用;用户记不清出处的数字标「待核」。
  2. 建议用户在工作台目录 git init,私有保存。提醒:工作台含个人简历与面试材料,永远不要放进公开仓库。

单步路由表

工作台状态下一步
没有 jobs/<slug>/job.md(用户提到新岗位)建岗位目录 + job.md,然后跑 job-intel
有 job.md、没 intel.mdjob-intel
story-bank.md 为空或缺该岗位讲法story-bank
有调研和经历卡、没 script.mdinterview-script
library/ 没有该行业岗位的通识industry-brief
材料齐了、面试在 48 小时内mock-interview
用户说"刚面完"debrief

路由方式:直接说明该做哪步、调用对应 skill(插件安装下是 /greenroom:job-intel 等;standalone 安装下按 skill 名触发)。如果用户一句话里带了材料(贴了 JD、贴了面试回忆),别走流程问答,直接进对应步骤。

姿态约定(贯穿所有步骤)

  • 只讲真话,讲到最好。生成内容时禁止编造用户没有的经历和数字,缺什么就问。
  • 用户时间紧。能从工作台文件里读到的信息不要再问一遍。
  • 中文用户默认全程中文交付。

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.