# Resume Crafter Skill
<div align="center">
## 简历.skill
---
> 简历? 易如反掌。
<img src="assets/cv-skill-readme.jpg" alt="简历? 易如反掌。" width="220" />
[](LICENSE)
[](https://claude.ai/code)
[](https://agentskills.io)
[](#中文)
[中文](#中文) | [English](#english)
</div>
---
## 中文
还在依赖网页版AI写简历?是不是经常遇到这些糟心问题:AI生成的内容复制回Word,格式变形、排版错乱,得花大量时间手动微调;AI从不主动向你确认信息,凭着脑补瞎编乱造,写出来的经历脱离实际、全是幻觉,根本没法直接投递;生成的简历格式全看天意,字体、间距、布局完全不受控制,想调整却无从下手?
简历.skill 正是为解决这些痛点而生——它是一套面向 OpenCode / Claude Code / Codex 等 agent 的简历制作全自动化工作流,用更专业、更可控的方式,帮你轻松制作高质量简历。
它的核心优势就是精准解决网页版AI的弊端:用LaTeX编写简历,直接交付精致、美观、紧凑的PDF文件,无需手动调整格式;遇到任何不确定的信息点,都会主动停下来逐一向你确认,彻底告别简历“幻觉”;内置多套精选模板,涵盖中文/英文、求职/学术、有/无头像照片等多种场景,格式完全可控,根据你的需求自动选择。
简单说:你负责提供真实经历,它负责把材料读清楚、问清楚、写清楚,最终直接交给你可以直接投递的 pdf 简历文件。
### 安装
复制这段话给 OpenCode / Claude Code / Codex 等 agent,一键配置:
```text
Fetch and follow instructions from:
https://raw.githubusercontent.com/Li-Baichuan-James/cv-skill/main/INSTALL.md
```
### 快速开始
1. 让 agent 按 `INSTALL.md` 完成安装与验证。
```text
https://raw.githubusercontent.com/Li-Baichuan-James/cv-skill/main/INSTALL.md
```
2. 在会话里自然描述需求即可。例如:
```text
帮我把这份旧简历改成 1-2 页的英文学术简历,用于phd申请。
```
3. 也可以直接说明语言、用途、风格、重点 等等。
```text
这份简历使用 中文/英文,用于求职/学术申请, 要/不要头像照片,重点突出科研项目和工程经历...
```
4. 正式动笔前,它会主动询问所有不确定的点,不会瞎猜/乱编。
5. 经你确认后,直接交给你可以直接投递的 pdf 简历文件,以及对应的latex源代码(方便后续修改更新)。
### 用户侧流程(你会看到什么)
1. Agent 先梳理信息:读取聊天、Word、PDF、截图或其他材料
2. 信息不清楚时会停下来问,不会硬猜
3. 起草前先整理需求、来源和约束
4. 简历中的最终表述会进入 `work/claim-source-map.md`
5. 交付时输出 `work/` 和 `output/`,方便复查和继续编译
6. 你会得到可以直接投递的 pdf 简历文件,以及对应的latex源代码
### 核心特性
- **事实优先**:先提取,后写作;缺信息就问,不靠脑补补齐
- **可追踪**:关键 claim 对应来源、位置、原始表述或用户确认
- **可复现**:`output/resume.tex` 和 `output/common/resume.cls` 放在一起,方便离线重编译
- **多模板**:支持 industry ATS、带照片版、research resume、中文标准模板
- **边界清楚**:只使用四个简历 skills,必要时配合 `docx` / `pdf` 读取源材料
- **可验证**:仓库提供 `tools/verify.ps1` 检查技能、模板、测试场景和示例输出
### 仓库内容
- `skills/`:四个可安装的 skill 入口
- `templates/`:精选 LaTeX 简历模板
- `examples/`:示例输入输出
- `tests/`:场景测试
- `docs/`:架构、贡献、发布和验证说明
- `tools/verify.ps1`:本地验证脚本
### 模板类型
| 模板 | 适用场景 |
| --- | --- |
| `templates/zh/standard` | 中文标准简历 |
| `templates/research/ats` | 面向学术申请的英文学术简历 |
| `templates/industry/ats` | 面向企业投递的求职英文简历 |
| `templates/industry/photo` | 用户明确要求需要包含头像照片的简历 |
### 它适合做什么
- 从聊天记录、项目说明、论文经历、实习经历中整理第一版简历
- 从 `.docx`、`.pdf`、截图或混合材料中提取信息并重写
- 生成可继续编辑、可本地编译的 LaTeX 简历
- 在本地 `xelatex` 可用时生成最终 PDF
### 不适合做什么
- 不凭空捏造实习、论文、奖项、指标或学校信息
- 不把聊天里的灵感当成事实,除非你确认它可以写进简历
### 与 `docx` / `pdf` companion skills 的关系
- `docx` 负责读取或提取 Word 简历材料
- `pdf` 负责读取、提取、检查 PDF 简历材料
- `resume-crafter` 负责简历 workflow:材料归一化、事实核查、写作、审阅、打包和交付
- 简单说:`docx` / `pdf` 帮 agent 读清楚文件,`resume-crafter` 负责把简历做完整
### 许可说明
本仓库的分发、复用与镜像以仓库内现有的 `LICENSE` 文件为准。
---
## English
`resume-crafter` is a four-skill package for producing factual 1-2 page resumes from chat notes, existing resumes, Word documents, PDFs, screenshots, and mixed source material.
The package is designed around source discipline: extract first, clarify blockers, draft only supported claims, review factual risk, and deliver a directly submittable `output/resume.pdf` plus corresponding output-local LaTeX source that can be recompiled outside the working directory.
### Installation
Tell your LLM agent:
```text
Fetch and follow instructions from:
https://raw.githubusercontent.com/Li-Baichuan-James/cv-skill/main/INSTALL.md
```
### Primary Entrypoint
Invoke `resume-crafter` for user-facing resume work.
Bundled supporting skills:
- `resume-intake-and-extraction`
- `resume-authoring-and-assembly`
- `resume-review-and-delivery`
### Quick Start
1. Install and validate the package through `INSTALL.md`:
```text
https://raw.githubusercontent.com/Li-Baichuan-James/cv-skill/main/INSTALL.md
```
2. Keep the full repository checkout available as the asset root and provide it as `CV_SKILL_ROOT`:
```text
CV_SKILL_ROOT=<absolute path to the full cv-skill checkout>
```
3. Ask naturally for the resume deliverable you want. For example:
```text
Build a 1-2 page ATS resume from the attached resume and my target backend software engineer notes.
```
You can also specify language, template family, source scope, or review constraints directly:
```text
Create a concise research resume from these academic materials. Keep only source-backed claims and ask before using uncertain metrics.
```
### Intended Use
- Create a new 1-2 page resume from chat-provided requirements and source material
- Convert an existing resume into LaTeX
- Rewrite an industry resume for ATS-oriented review
- Adapt academic or research material into a concise research resume
- Produce Chinese-language resumes using the standard Chinese template
- Package a directly submittable PDF and reproducible LaTeX source
### Scope Boundaries
This package does not produce long academic CVs. If the input is a long academic CV, the workflow should offer a concise 1-2 page research resume instead.
The workflow must not invent unsupported employers, degrees, publications, awards, dates, metrics, or tools. Ambiguous or missing facts should be clarified or omitted rather than guessed.
### Workflow Guarantees
- Extraction happens before drafting.
- Blockers stop the flow instead of being guessed around.
- Every final claim is resolved in `work/claim-source-map.md`.
- Each run uses a fresh workspace.
- Final delivery requires `output/resume.pdf`.
- Corresponding reproducible source is delivered through `output/resume.tex` and `output/common/resume.cls`.
- If XeLaTeX is missing, the workflow attempts to install or activate a XeLaTeX-capable environment; if PDF generation still cannot complete, final delivery is blocked.
### Expected Workspace
```text
resume-workspace-YYYYMMDD-HHMMSS/
work/
requirements-summary.md
claim-source-map.md
review.md
common/
resume.cls
output/
resume.tex
resume.pdf
common/
resume.cls
```
Compile from the directory containing `resume.tex`:
```powershell
xelatex -interaction=nonstopmode -halt-on-error resume.tex
```
If `xelatex` is unavailable, the workflow attempts to install or activate a XeLaTeX-capable TeX distribution. If installation or compilation fails, the run reports a blocker instead of presenting source files alone as final delivery.
### Template Matrix
| Template | Intended use |
| --- | --- |
| `templates/industry/ats` | ATS-friendly industry resumes |
| `templates/industry/photo` | Industry resumes where a photo layout is explicitly appropriate |
| `templates/research/ats` | Concise research resumes adapted from academic material |
| `templates/zh/standard` | Chinese-language standard resume format |
### Dependencies And Companion Skills
- `resume-crafter` is the primary user-facing entrypoint.
- `resume-intake-and-extraction` normalizes chat, Word, PDF, screenshot, and mixed source material.
- `resume-authoring-and-assembly` drafts the 1-2 page LaTeX resume source.
- `resume-review-and-delivery` reviews factual safety and packages final outputs.
- Upstream `docx` is used when Word source material must be read or extracted.
- Upstream `pdf` is used when PDF source material must be read or inspected.
- Host image or OCR capability is required for screenshots or scanned sources.
- `xelatex` is required for final PDF delivery.
### Asset Root Contract
`CV_SKILL_ROOT` is the absolute path to the full repository checkout. Installed skill folders are runtime entrypoints only. Templates, examples, tests, docs, and verification tools remain under `CV_SKILL_ROOT`. If `CV_SKILL_ROOT` is unknown, the agent must ask for it before using repository assets.
### Verification
Run the repository verifier before publishing or after changing skills, templates, tests, examples, or installation docs:
```powershell
tools/verify.ps1
```
The verifier compiles template variants in temporary directories when `xelatex` is installed. Publishing a verified release requires running these compile checks in an environment with XeLaTeX.
### Repository Contents
- `skills/`: four installed skill entrypoints
- `templates/`: resume templates and shared LaTeX class assets
- `examples/`: example inputs and expected workflows
- `tests/`: skill and packaging test fixtures
- `docs/`: architecture, contribution, publishing, and verification docs
- `tools/verify.ps1`: repository verification helper
### License
Distribution, reuse, and mirroring of this repository are governed by the `LICENSE` file included in the repository.