Resume — One Source, Many Tailored Resumes

A complete resume management system: structured experience data → AI-tailored LaTeX → PDF.

experiences/  →  /resume generate  →  resume.tex  →  make en  →  output/Name-Title.pdf
   (素材)         (AI 定制化)          (LaTeX源)      (编译)        (输出)

Subcommand Routing

Parse $ARGUMENTS to determine which workflow to run:

init [text] → Init Workflow
generate [job title or JD file] or gen [...] → Generate Workflow
add [description] → Add Experience Workflow
Empty or unrecognized → Show available subcommands and ask user what they want to do

Prerequisites Check

Before any workflow, verify the environment:

Check if resume.cls exists in the current working directory. If not, copy from ${CLAUDE_SKILL_DIR}/assets/resume.cls
Check if Makefile exists. If not, copy from ${CLAUDE_SKILL_DIR}/assets/Makefile
Create directories if missing: experiences/work/, jobs/, output/, .history/
Check if xelatex is available: which xelatex. If not, warn the user: "XeLaTeX is required. Install with: brew install --cask mactex"

Init Workflow (`/resume init [text]`)

Initialize the experience data library from scratch.

Mode Detection

Based on $ARGUMENTS (after stripping "init"):

User pasted large text (resume, work history, LinkedIn export) → Import Mode
Empty or short description → Q&A Mode
File path provided (PDF/Markdown) → Read file, then Import Mode

Import Mode

Parse the provided content, extracting: personal info, education, work history (company, period, department, role, projects, metrics), personal projects, skills, honors
Confirm with user — show structured result, ask about missing info or inaccurate metrics
Generate files into experiences/:
- profile.md — personal info + education
- work/<company>.md — one file per company
- projects.md — personal projects
- skills.md — skill inventory
- honors.md — honors + quantified metrics

Q&A Mode

Walk through one question at a time. Never ask multiple questions at once.

Basic info → name, email, phone, wechat (optional), blog/github (optional) → write experiences/profile.md
Education → school, degree, major, period, highlights → append to experiences/profile.md
Work history (loop) → company, department, role, period, projects, metrics, tech stack → write experiences/work/<company>.md. After each: "Any more work experience?" Continue or move on.
Personal projects → name, link, description, tech stack, highlights → write experiences/projects.md
Skills → auto-generate from collected tech stacks, show for confirmation → write experiences/skills.md
Honors → extract quantified metrics from work history + user additions → write experiences/honors.md

File Format Reference

profile.md:

---
name: Name / English Name
email: email@example.com
phone: "+86 xxx"
wechat: xxx
blog: https://...
github: https://...
---

## Education

### School · College | Degree (Period)
- Major, highlights

work/<company>.md:

---
company: Company Name
company_en: English Name
url: https://...
location: City
period: YYYY/MM -- YYYY/MM
department: Department
role: Role
tags: [tag1, tag2, ...]
---

# Project Name
> Timeline | Role

## Background
...

## Key Work
...

## Metrics
...

## Tech Stack
...

Key principles:

Keep raw details in experiences — trimming happens at generation time
Tags should cover core tech and business domain keywords
Metrics must include specific numbers
One file per company, filename in lowercase English (e.g., bytedance.md)

Generate Workflow (`/resume generate [target]`)

Generate a tailored resume for a specific job target.

Step 1: Read All Materials

Read all files:

experiences/profile.md
All .md files under experiences/work/
experiences/projects.md
experiences/skills.md
experiences/honors.md

Read ${CLAUDE_SKILL_DIR}/assets/resume-example.tex to understand the LaTeX structure and available commands.

If the argument is a file path, also read that JD file.

Step 2: Analyze Job Requirements

Based on JD or job title, determine:

Core technical requirements — what skills matter most
Experience priority — which work experiences are most relevant
Framing angle — how to present the same experience differently for this role
Content selection — include/omit personal projects section, which experiences to trim

Step 3: Present the Plan

Before generating .tex, show the user:

Target positioning (one sentence)
Content selection (which experiences to include/omit)
Key differentiation (how this resume differs from a generic one)

Wait for user confirmation before proceeding.

Step 4: Generate LaTeX

Generate resume.tex following these rules:

Use resume.cls commands strictly:
- \name{}, \basicInfo{}, \email{}, \phone{}, \wechat{}, \homepage[]{}
- \section{}, \datedsubsection{}{}, \role{}{}, \rolewithdate{}{}{}
- itemize with [parsep=0.25ex]
Keep to 1 page (critical for Chinese version):
- 5-6 sections: positioning, education, work, (projects), skills, honors
- 3-5 bullet points per project
- Trim early internships to 1-2 lines each
- If too long, cut early internships and low-relevance bullets first

LaTeX header for Chinese:

\documentclass{resume}
\usepackage{xeCJK}
\setCJKmainfont{SourceHanSerifSC-Medium}
\setCJKsansfont{SourceHanSerifSC-Medium}
\setCJKmonofont{SourceHanSerifSC-Medium}

For English version: remove the three xeCJK lines.

Formatting:
- Use \textbf{【keyword】} to mark each bullet's keyword
- Escape: & → \&, % → \%
- Use \href{url}{text} for hyperlinks

Step 5: Build

Backup current resume.tex to .history/resume_$(date +%Y%m%d%H%M%S).tex
Write generated content to resume.tex
Run make en to build PDF
Copy PDF to output/ with job-specific name

Step 6: English Version (if requested)

Remove xeCJK lines
Translate all content to English
Build PDF

Add Experience Workflow (`/resume add [description]`)

Add or update experience data incrementally.

Step 1: Classify Input

Determine what the user is adding:

New work experience → create/update experiences/work/*.md
New personal project → update experiences/projects.md
New skill → update experiences/skills.md
New honor/certification → update experiences/honors.md
Personal info change → update experiences/profile.md

Step 2: Read Existing File

Read the target file to understand current content and format.

Step 3: Update

Apply updates following the file format conventions above. Keep raw details — don't trim prematurely.

Step 4: Confirm

Show the diff to the user, confirm before writing.

Step 5: Remind

After updating: "Experience updated. Run /resume generate [target] to regenerate your resume."

resume