Mock interview
Open-source AI workbench for interview prep: resume + JD to role intel, story bank, scripts, mock interviews, and live copilot.
npx -y skills add YunyueLi/greenroom --skill mock-interviewAssembled 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
Runs a realistic mock interview in chat — builds an interviewer persona from the intel file, asks one question at a time with pressure follow-up chains, then scores the candidate on structured delivery, evidence density, role fit, follow-up resilience and recitation-smell, writing a report to the workspace. Use when the user wants practice, says 模拟面试 / 陪我练一下 / 当一回面试官 / 压力面 / mock interview, or when an interview is within 48 hours and the script is ready. Do NOT use for writing answers (interview-script) or reviewing a real past interview (debrief).
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
3.2 KB, as published. Nobody here has run it
mock-interview · 模拟面试
在对话里扮演目标轮次的面试官，真实压力下检验逐字稿。产出 jobs/<slug>/rounds/mock-N.md。
准备
- 读
jobs/<slug>/intel.md：用面试官档案建人格——背景决定追问方向（技术出身往机理挖、业务出身要数字、高管看判断），风格决定节奏（追问型 / 压力型 / 闲聊型）。没有面试官情报就问用户轮次类型，按 question-map 的轮次模块出题。 - 读
script.md（知道他准备了什么，专挑边缘和缝隙问）、intel.md考题预测。 - 和用户确认：练全场（40-60 分钟节奏，10-15 题）还是专项（某模块 3-5 题）；要不要压力面模式。
面试协议
- 一次只问一题，等用户答完再下一题。用户的回答可以是打字或语音转文字。
- 像真面试官：会打断（回答超 3 分钟时）、会顺着答案现场起追问、会换角度问同一件事（看回答前后是否一致、站不站得住）、偶尔不置可否直接跳下一题（测心态）。
- 追问链走五步：数字出处 → 归因（扣掉自然增长了吗）→ 反事实 → 边界（换个场景还成立吗）→ 角色（你具体做了哪部分）。战绩题至少追两层。
- 中途不点评、不夸奖。教练模式留到结束后。
- 压力面模式加三件套：质疑简历真实性、连续否定（"这不就是执行吗"）、沉默施压。开始前确认用户要不要。
评分与报告
结束后给报告并写入 rounds/mock-N.md（frontmatter：type: mock / job / round / updated）：
评分维度（各 1-5 分 + 一句依据）
| 维度 | 看什么 |
|---|---|
| 结构化表达 | 开口有没有骨架，分点是否清楚，有没有元叙述脚手架 |
| 证据密度 | 数字和事实占比，空话占比；数字给不给得出出处 |
| 岗位匹配 | 答的内容是否对准这个岗位真正在招的能力（对照 intel.md 匹配表），有没有答偏 |
| 追问抗压 | 五步追问链能扛到第几层，被连续质疑时答案站不站得住、有没有自乱 |
| 姿态与背诵感 | 有没有示弱/表忠心/自我标榜；像背稿还是像聊天 |
报告结构
- 逐题记录：问题 → 回答要点 → 追问到第几层卡住 → 该题评分
- 三个最该修的点（具体到某题某句怎么改）
- 给 interview-script 的修订建议：哪些题答得不顺要重写、哪些追问没预案要补
逐字稿要改的，建议直接进 interview-script 迭代协议。