shenron0101/interview-questions

Interview Questions Skill

You generate structured interview questions from vault notes, saved as Obsidian-compatible markdown with collapsible callout blocks.

---

Step 0 - Gather parameters

Ask for missing parameters in a single message:

| Parameter | Default | Notes | |-----------|---------|-------| | topic | (required) | e.g. "OLS regression", "Dynamic Programming", "Options pricing" | | raw_dir | raw/ | Source folder | | count | 15 | Number of questions | | difficulty | Mixed | Easy / Medium / Hard / Mixed | | types | Mixed | See type table below | | role | General | Quant Researcher / SWE / ML Engineer / Trading | | output_dir | outputs/interview-questions/ | Where to save the .md |

Question types (comma-separated, or "Mixed"):

| Code | Type | Best for | |------|------|----------| | concept | Definition + intuition | All roles | | scenario | "What would you do if..." | Trading / Quant | | gotcha | Common misconception traps | All roles | | explain | "Explain to a non-expert" | All roles |

---

Step 1 - Read the vault

python3 "${CLAUDE_PLUGIN_ROOT}/scripts/read_vault.py" "<raw_dir>"

• Summarize: "Found N files. Using: [list]"
• Filter content to the requested topic. If you find nothing relevant, say so and offer to generate questions from your own knowledge instead (ask user to confirm).

---

Step 2 - Plan question distribution

Before writing, plan the spread:

• If difficulty = Mixed: roughly 30% Easy, 50% Medium, 20% Hard
• If types = Mixed: spread across concept, derive/code, scenario, gotcha
• Ensure questions are not redundant - each should test a distinct idea
• Weight toward what's commonly tested in real interviews for role

For Quant / Trading roles, prioritise:

• Derivations (OVB proof, Wald statistic, GMM consistency, etc.)
• Numerical intuition ("What happens to beta-hat if you add a correlated regressor?")
• Market microstructure / options pricing edge cases

For ML Engineer, prioritise:

• Bias-variance, regularisation, overfitting
• Implementation traps (data leakage, class imbalance)
• System design components

---

Step 3 - Generate the markdown

Output format

Every question uses Obsidian's collapsible callout syntax:

> [!question]- Q{n} [{difficulty}] [{type}]: {Question text}
>
> **Answer:**
> {Core answer - clear and complete. Use $LaTeX$ for math.}
>
> **Key Points:**
> - {Point 1}
> - {Point 2}
> - {Point 3}
>
> **Common Mistake:**
> {One specific wrong answer / confusion to avoid}
>
> **Follow-up:** {One harder follow-up question}

The - after [!question] makes it collapsed by default in Obsidian.

Difficulty badges

Use these exactly:

• [Easy] - can be answered in < 60 seconds
• [Medium] - requires 1-3 minutes of reasoning
• [Hard] - derivation or deep systems thinking required

Type badges

Use: [Concept], [Derive], [Code], [Scenario], [Gotcha], [Explain]

Special formatting rules

Math: Use $...$ for inline and $$...$$ for display math. Example: $\hat{\beta} = (X'X)^{-1}X'Y$

Code blocks: Use fenced code blocks with language tag:

> ```python
> def binary_search(arr, target):
>     lo, hi = 0, len(arr) - 1
>     ...
> ```

Note: inside a callout, indent code blocks with > prefix.

Tables: Use markdown tables inside the callout:

> | Model | ACF | PACF |
> |-------|-----|------|
> | AR(p) | Decay | Cuts off |

Derivation questions: Structure the answer as numbered steps, each <= 2 lines. Show the key algebra, don't skip steps that would trip up an interviewer.

---

Step 4 - File structure

The complete output file:

---
topic: {topic}
date: {YYYY-MM-DD}
role: {role}
difficulty: {difficulty}
count: {count}
sources: [{list of source filenames}]
tags: [interview-prep, {topic-slug}]
---

# Interview Questions: {topic}

> **Role:** {role} | **Difficulty:** {difficulty} | **Generated:** {date}

---

{Q1 callout}

{Q2 callout}

...

{QN callout}

---
*Sources: {comma-separated filenames} | vault-skills v1.0*

---

Step 5 - Save

Filename: <output_dir>/<topic-slug>_interviews_<YYYY-MM-DD>.md

Create <output_dir> if it doesn't exist. Write the file.

Tell the user: "Saved {count} questions to <path>."

---

Step 6 - Report

Interview questions generated
  Topic:      {topic}
  Role:       {role}
  Count:      {count} questions
  Difficulty: {difficulty distribution, e.g. "5 Easy / 8 Medium / 2 Hard"}
  Types:      {type distribution}
  Sources:    {N} files
  Output:     {path}

Tip: In Obsidian, click the arrow on each callout to expand.
     Use Ctrl+Click to open in a new pane for side-by-side review.

---

Quality Checklist

• [ ] Every callout has Answer, Key Points, Common Mistake, Follow-up
• [ ] No two questions test the same concept
• [ ] Math is in LaTeX math mode (not Unicode)
• [ ] Code questions have correct syntax
• [ ] Difficulty labels match actual difficulty
• [ ] Frontmatter YAML is valid
• [ ] Callout syntax uses > [!question]- (with dash for collapsed)