My Interview

A bilingual Codex skill for company research, resume-to-JD matching, interview question prediction, mock interviews, STAR evaluation, and natural HR outreach scripts.

中英双语面试准备 Skill：从公司调研、简历匹配、面试题预测，到模拟面试、STAR 评估、HR 打招呼话术，一次生成可复用的面试准备包。

my-interview turns a company name, role, JD, and resume into a structured interview preparation package. It is designed for candidates who want practical, evidence-based prep instead of generic question lists.

一句话：它不是“给你一堆通用面试题”，而是基于你的真实简历、目标 JD、公司情报和面经线索，推演面试官最可能追问什么，以及你应该如何诚实、有结构地回答。

It works especially well for:

• Preparing for a specific company and role with fresh company research.
• Matching a resume against a JD and finding real gaps before the interviewer does.
• Predicting interview questions from company intel, JD requirements, resume weak points, and known interview experience.
• Practicing with company-specific mock interviews and STAR-based scoring.
• Writing natural HR greeting, referral, and follow-up scripts without inflated AI-sounding language.

Why This Skill Is Different

• It starts from evidence. Resume facts, JD requirements, company research, and interview reports are kept separate and labeled.
• It predicts pressure points. The strongest questions often come from mismatched titles, missing metrics, vague ownership, career transitions, or JD gaps.
• It refuses fake polish. When a metric, motivation, or responsibility is missing, the skill marks the gap instead of inventing a better story.
• It produces reusable assets. Candidate knowledge base, STAR storybank, prep packages, and debrief notes can compound across applications.
• It speaks like a person. Outreach scripts are checked against common AI-writing patterns and business-jargon filler.

What It Produces

By default, the skill writes the full preparation package to:

preps/{company-role-date}/full-prep.md

The generated prep package includes:

1. Intake summary and reusable candidate knowledge base.
2. Company research with source confidence labels.
3. Product deep-dive when the role depends on a specific product.
4. Resume x JD match analysis with concrete gap notes.
5. Predicted interview questions ranked by priority.
6. Draft answers grounded only in real resume facts.
7. HR outreach scripts, referral blurbs, and follow-up messages.
8. Mock interview flow and STAR evaluation rubric.

The conversation stays short while the detailed output is saved to files, so the result can be reviewed in VS Code, copied to Notion, or shared as a single document.

Core Features

Feature	What it does
Company research	Searches business model, recent news, workplace reputation, interview experiences, and role-specific signals.
Resume-JD matching	Maps each JD requirement to evidence in the resume, then marks match strength and gaps.
Question prediction	Generates P0-P4 questions from real interview intel, JD requirements, resume contradictions, and company context.
Candidate knowledge base	Builds reusable candidate profile and STAR story inventory across multiple applications.
Mock interview	Runs a company-specific mock interview with role-appropriate follow-up style.
STAR evaluation	Scores behavioral answers with weighted S/T/A/R criteria and concrete improvement actions.
HR scripts	Generates direct, natural outreach copy for BOSS, LinkedIn, email, referral, and follow-up scenarios.
Bilingual mode	Uses Chinese, English, or mixed bilingual output based on user language, JD language, and interview context.

Command Reference

Command	Purpose
/interview or natural language	Start the full interview prep pipeline.
/mock {slug}	Run a mock interview from an existing prep package.
/storybank	Manage reusable STAR stories.
/storybank add	Add a new STAR story.
/storybank list	List saved stories by competency.
/storybank gaps	Find missing competency coverage.
/hype {slug}	Generate a pre-interview confidence brief.
/debrief {slug}	Record post-interview notes and improve future predictions.
/update-prep {slug}	Merge new interview intel into an existing prep package.
/greet	Generate natural HR greeting scripts.
/list-preps	List generated prep sessions.

Natural language also works. Examples:

Use $my-interview to prepare for a Google SWE L4 interview. Here is the JD: ...

用 $my-interview 帮我准备字节跳动后端开发面试，这是 JD 和我的简历：...

/greet 根据这份 JD 和我的简历，写一段 BOSS 直聘打招呼话术，不要 AI 味。

Installation

Copy this folder into your Codex skills directory:

~/.codex/skills/my-interview

or, on Windows:

%USERPROFILE%\.codex\skills\my-interview

Then invoke it explicitly:

Use $my-interview to prepare for ...

The skill can use web search for company research. If web search is unavailable, it will clearly mark research gaps instead of inventing information.

Workflow

flowchart TD
    A["Company + Role + JD + Resume"] --> B["Candidate Knowledge Base"]
    B --> C["Company Research"]
    C --> D["Product Deep-Dive, if relevant"]
    D --> E["Resume x JD Match"]
    E --> F["Predicted Questions"]
    F --> G["Draft Answers + HR Scripts"]
    G --> H["Mock Interview"]
    H --> I["STAR Evaluation"]
    I --> J["Debrief + Update Prep"]

Loading

Design Principles

• No fabricated experience. Answers are grounded in the resume, candidate knowledge base, portfolio, or facts supplied by the user.
• Confidence labels everywhere. Research findings are marked as HIGH, MEDIUM, LOW, or GAP.
• Interviewer realism over generic prep. Questions are generated from JD requirements, real interview signals, company context, and resume weak points.
• File-first output. Long-form prep is written to files; the chat only shows progress and the most important reminders.
• Natural language for outreach. HR scripts avoid empty praise, inflated business jargon, and generic AI-style endings.

Repository Structure

my-interview/
├── SKILL.md
├── README.md
├── LICENSE
├── agents/
│   └── openai.yaml
└── references/
    ├── company-culture-tags.md
    ├── competency-taxonomy.md
    └── star-framework.md

Generated user data is intentionally ignored by git:

preps/
storybank/
candidate-knowledge-base-*.md

References Included

• references/star-framework.md - STAR scoring rubric, high-score answer patterns, and common mistakes.
• references/competency-taxonomy.md - 15 behavioral competency categories used for storybank tagging.
• references/company-culture-tags.md - Interview style signals for FAANG, startups, domestic big tech, banks, consulting, and more.

Limitations

• Company research quality depends on available web search and source freshness.
• PDF resume extraction may fail for scanned or encrypted files; plain text is the most reliable input.
• The skill will not invent metrics, titles, responsibilities, or motivations. Missing facts are marked for the candidate to fill in.

License

MIT License. See LICENSE.