yanliudesign/offer-toolkit-skill / job-description-skill
job-description-skill
VerifiedJob Description 解码器 + Offer 策略系统。把任何 JD 翻译成「这家公司到底在招什么人 / 我的匹配度 / Gap 在哪 / 为什么投/不投 / 面试会问什么 / 下一步怎么走」。只有 3 步:贴 JD → 给简历 → 自动生成并打开一份 HTML Offer Strategy Report。是整个求职链路的入口,向下衔接 Resume Skill 和 BQ Skill。关键词:job description, JD, 职位描述, 匹配度, ATS, resume tailor, 简历定制, 招聘经理, 投不投, should I apply, 面试预测, offer strategy。
prompts/interview-predictor.md
View on GitHub# Interview Predictor
**目标:** 基于 JD(已解码)输出 Top 20 道大概率会被问的题,分四类,每道题标"为什么会被问"。Behavior 类导流到 BQ Skill。
**铁律**
- 必须先有 Decode 产出。预测题目本质上是把 Must Have / Hidden Signal / 团队画像翻译成可能的提问。
- 不要装作知道某公司"实际面试题题库"。所有预测都说"基于 JD 推断 + 行业惯例",不要造谣公司具体面试流程。
- 一次给完 20 道,分清楚类别和优先级,不要让用户面对一份平铺的题单。
---
## Step 1 · 收齐输入
1. **JD Decode 产出**(必需)。
2. **公司近期产品 / 业务方向**(可选)—— 用户知道就告诉,不知道就标"行业惯例推断"。
3. **用户对哪个环节最虚**(可选)—— 影响题目优先级排序。
---
## Step 2 · 按四类生成
每类 5 道左右,**每道题必须标来源**(对应 JD 哪条 / 哪个 Hidden Signal)。
### 类别 1 · Behavior(行为题)
由 Hidden Signals 和 Must Have 里的"软能力"推断。常见模板:
- Hidden Signal 含 `ambiguity` → "Tell me about a time you had to make a decision with incomplete information."
- Hidden Signal 含 `ownership` → "Tell me about a time you took ownership of something outside your role."
- Hidden Signal 含 `influence` → "Tell me about a time you influenced a decision without authority."
- Must Have 含 `stakeholder management` → "Walk me through a project with significant cross-functional friction."
- Must Have 含 `end-to-end` → "Tell me about a project you owned from 0 to ship."
每道题写成:
```
B1. "Tell me about a time you had to drive a decision with incomplete information."
来源:Hidden Signal #2(ambiguity / unstructured)
考察:Dealing with Ambiguity
→ 用 BQ Skill 找 / 挖一个 ambiguity 类故事
```
**Behavior 部分结尾统一一句**:
> "上面 5 道 Behavior 题,建议用 **BQ Skill** 挖故事或查故事库。如果还没有故事库,BQ Skill 可以从这几道题反向帮你挖故事。"
### 类别 2 · Product Thinking / Domain(专业题)
根据 JD 涉及的产品领域出题。Designer / PM 岗位常见:
- "How would you design [JD 里提到的产品场景]?"
- "Critique [该公司现有产品 / 同类产品] — what would you change?"
- "How would you measure success for [JD 里强调的功能 / 方向]?"
- "Walk us through a design process from research to ship."
设计师岗加 portfolio 题:
- "Pick your strongest case study and walk us through it."
- "Show us a project where you changed direction based on user feedback."
### 类别 3 · Technical / Craft(技能题)
根据 JD 列的硬技能出。比如:
- JD 提到 AI / LLM → "Walk me through how you'd design for hallucinations / model failure modes."
- JD 提到 enterprise / B2B → "How do you design for admin vs end-user personas?"
- JD 提到 0→1 → "How do you decide what to ship in v1 vs defer?"
- JD 提到 data-driven → "Tell me about a metric you defined and how it changed product decisions."
非设计岗(PM / Eng)换对应技能题。
### 类别 4 · Company-specific(针对该公司 / 团队)
3-5 道根据这家公司具体推断的题。例如:
- 公司在做 AI agent → "What's your point of view on the right UX for autonomous agents?"
- 公司是 startup → "Why startup over big tech? Why this size of company?"
- 公司是 mission-driven → "What about our mission resonates with you?"
- 用户从大厂跳过去 → "How will you adjust to a less structured environment?"
- 用户跨行业 → "Why this domain after [previous domain]?"
每道题标"基于 JD / 公司情况推断",**不要装作内部消息**。
---
## Step 3 · 排个优先级
20 道题平铺信息量太大。最后给一个 **Top 5 必练**清单:
```
🎯 优先级最高的 5 道(按风险×概率排):
1. B3 — 因为 Hidden Signal 反复出现 ambiguity
2. P2 — 因为是 JD 核心场景
3. C1 — 因为用户简历有大厂跳创业的画像风险
4. ...
```
排序依据:**出现概率 × 用户准备的薄弱度**。
---
## Step 4 · 存档 + 导流
1. **更新 JD 文件**:加 predicted_questions section,记录 20 道题。
2. **导流**:
> "面试题预测完成 ✅。建议:
> - Behavior 部分 → 去 **BQ Skill** 挖故事 / 模拟面试
> - Product / Craft 部分 → 自己整理 talking points,重点准备 Top 5
> - 想模拟面试 → BQ Skill 有轻量模拟流程
> - 还想看其他职位 → 继续给我新 JD"
---
## 常见卡点
| 症状 | 处理 |
|---|---|
| 用户问"这家公司具体面试流程是什么" | 老实说:「我没有内部信息,建议查 Glassdoor / Blind / Levels.fyi / 联系前员工」 |
| JD 信息太少推不出 20 道 | 给能推出的(10-15 道)+ 1 个清单:"剩下的题需要这些信息:① ② ③" |
| 用户希望具体答案 | 说"这里只出题,准备答案:Behavior 去 BQ Skill;Product 题建议自己准备 talking points(我不替你编案例)" |
| 用户面试已经很近 | 直接给 Top 5 必练 + 每道题的"准备方向",跳过完整 20 道 |
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |