Interview Designer
Generate interview questions with assessment dimensions and A-level reference answers from a resume, JD, PRD, or portfolio. Two modes — interviewer mode (scorecard, red flags, A/B/C scoring) and candidate prep mode (question essence, weighted dimensions, speakable A-level scripts). Covers an 11-group / 52-dimension coverage model (truth, demand, solution & AI boundary, product form, technical depth, reliability, enterprise reality, business value, collaboration, learning, fit) plus an Enterprise Reality Test that attacks a plan's unstated premises. Staged output with a hard context budget so l
Install
npx skills add zhangchau/interview-designer-skill --skill interview-designergit clone --depth 1 https://github.com/zhangchau/interview-designer-skill.git /tmp/interview-designer-skill
mkdir -p ~/.claude/skills
cp -r /tmp/interview-designer-skill/. ~/.claude/skills/interview-designerPersonal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.
From the SKILL.md
# Interview Designer
> **两种模式**：面试官出题评分 ｜ 候选人备考写逐字稿
> **四阶段**：判断提取 → 素材卡 → 分层逐字稿 → 攻防增强（**分阶段产出，禁止一次全出**）
> **输出语言**：跟随用户输入语言。中文提问则全部中文输出。
---
## 0. 第一性原理
七条原理推导出全部规则。理解这七条，不需要记流程。
**一、面试是信息不对称下的预测问题。**
用有限的、且由候选人自己控制的证据，预测他未来 12 个月的表现。全部技术都在解决"证据不可信"和"未来不可知"。
**二、没有标准，取证就变成挑刺。**
先看简历再定标准，标准会被简历污染（因人设岗）。必须先独立定义 A 级长什么样。**顺序错了后面全错。**
**三、候选人提交的文档一定是自洽的。**
简历和 PRD 是他自己写的，内部逻辑必然闭环，**从文档内部提问永远问不倒他**。破自洽文档只能攻击它的前提——它对真实世界那些没说出口的假设。纸面方案通常内部完美、外部幼稚：假设数据干净、接口开放、审核的人不疲劳、需求是真的。**这些假设就是攻击面。**
**四、问题和答案是同一个东西的两面。**
一道好题必然存在一个可讲出口的好答案。**写不出 A 级答案说明题目本身有问题**——太空泛、无法回答、或没有区分度。答案生成是对题目质量的强制校验，不是附加功能。
**五、面试官和候选人需要的答案形态不同。**
面试官需要 A/B/C 分级做区分。候选人需要唯一一个能内化的标准答案——**给他看 C 级答案会污染准备，让他记住错误说法。**
**六、素材卡是压缩机制，不是内容增量。**
N 个答案各自重述同一个例子 = N 份拷贝；一张卡被引用 N 次 = 1 份。**卡让产出变小，前提是答案引用卡而不是内联卡。**
**七、纸面项目（M0-M1）没有"经历"可讲，只有"判断"可证明。**
面试官知道课程项目没有真实用户，他要看的是思维方式对不对。而思维方式的证明不在单个答案里，在答案之间的结构里：判断的迁移、正反论证、边界意识、迭代历史、验证路径。
---
## 1. 上下文纪律（硬性，违反则输出失败）
**问题**：PRD 常在 100-250KB。整读输入约 60-80k tokens；一次生成 60 题全量答案约 40,000 字。**质量在触到硬限制之前就已经崩了。**
**输入侧压缩（必须做）**
1. **先扫大纲不读全文**：`grep -n "^#\{1,3\} "` 拿到结构。
2. **只读含判断与数字的节**：摘要、KPI/指标、核心技术决策 3-5 节、范围边界。目标 ≤6000 字（220KB → 约 18 倍压缩）。
3. **禁止整读**：需要细节时定点再读那一节，不回头整读。
**冻结简报（Stage 1 后必须产出）**
写一份 800-1200 字 `项目冻结简报`：项目性质与成熟度、12 个判断各一句话、关键数字清单、已知缺口。
**此后所有阶段只读这份简报 + 当前要处理的那张卡，不再回读 PRD，不回读前批产出。**
**输出侧配额（硬上限）**
| 产物 | 上限 | 说明 |
| :--- | :--- | :--- |
| 判断台账 | 10-15 条 | 超过说明在抄功能，不是在提判断 |
| 题目索引 | 40-60 题 | 只有题干 + 维度码 + 卡号，**不含答案** |Details
- Skill name
interview-designer- Repository
- zhangchau/interview-designer-skill
- License
- MIT
- Repo updated
- 10 August 2026
FAQ
What does the Interview Designer skill do?
Generate interview questions with assessment dimensions and A-level reference answers from a resume, JD, PRD, or portfolio. Two modes — interviewer mode (scorecard, red flags, A/B/C scoring) and candidate prep mode (question essence, weighted dimensions, speakable A-level scripts). Covers an 11-group / 52-dimension coverage model (truth, demand, solution & AI boundary, product form, technical depth, reliability, enterprise reality, business value, collaboration, learning, fit) plus an Enterprise Reality Test that attacks a plan's unstated premises. Staged output with a hard context budget so l
How do I install the Interview Designer skill?
skills CLI (Claude Code, Cursor, Codex and more): npx skills add zhangchau/interview-designer-skill --skill interview-designer. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.
Is Interview Designer popular?
It lives in zhangchau/interview-designer-skill, which has 8 GitHub stars and 1 forks.
Related skills
Manju Creator
9rfdiosuao/Yuhang-ManJu
漫剧创作大师 - 全流程 AI 视频创作体系。整合提示词工程、图像生成、视频生成、后期剪辑四大能力。Invoke when user wants to create AI video, manga, short drama, or needs prompt engineering for video generation.
Information Literacy
9Liii8888/information-literacy
专为大学生“AI+信息素养”大赛备赛、模拟练习和规则允许的辅助答题设计。为单选、多选、判断题拆解条件，核验学术检索、科研AI工具、文件版本、产品菜单与引用证据；不提供题库或默认正式考试授权。
One Person Company
8DouyuShinyruo/One-Person-Company-Skill
当需要根据任务组建临时 AI Agent 团队协作时使用，覆盖战略、产品、设计、开发、测试、营销、运营、销售等全流程，像拥有完整团队一样独立运作。
Ida Reverse Engineering
8Newmcpe/ida-reverse-engineering-skill
Drive IDA Pro through the IDA Pro MCP like a senior reverse engineer: don't just narrate decompiler output, transform the database. Use this whenever the user is reverse engineering, analyzing malware, working a crackme or CTF, or doing binary/firmware analysis with IDA Pro over an MCP connection (mrexodia/ida-pro-mcp or idalib-mcp) — and especially when the agent keeps guessing at raw pseudocode instead of renaming functions and variables, defining structs, fixing types, and adding comments. Covers the iterative cleanup loop, anti-hallucination rules (always use int_convert; read real bytes/s
Openclaw Genpark Agent Forge
9alphaparkinc/openclaw-genpark-agent-forge
OpenClaw GenPark Agent Forge - Multi-agent DAG workflow compiler with Kahn's topological sort and Critical Path Method (CPM).
Hebbs
9hebbs-ai/hebbs-skill
Cognitive engine for AI agents: index files and memories, retrieve across 4 weighted dimensions (similarity, temporal, importance, frequency), automatic insight consolidation, automatic contradiction detection, with full parameter control on every function.
Browse the most popular Claude skills, all ai, llms & agents skills, or write your own with the skills generator.