resume

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Resume: Universal ATS Resume Optimization Skill

简历：通用ATS简历优化技能

Invocation:

```
/resume
```
— Interview the user, then build a new resume from scratch.
```
/resume <file>
```
— Parse an existing resume, interview the user to confirm/fill gaps, then diagnose and rewrite to fix every ATS flag.
```
/resume-score <file>
```
— Brief interview for context (target role + JD), then estimate the ATS score with per-category breakdown.
```
/resume-jd-tune <file> <jd>
```
— Brief interview to confirm skill claims, then keyword-tune the resume to the JD.

Interview is mandatory on every command — see Behavior below.

Scripts: Located at the plugin root

scripts/

directory —

build_ats_resume.js

(DOCX builder),

score_resume.py

(heuristic scorer).

Orchestrates a 5-phase workflow (interview → diagnose → rewrite → build → score) across any field. Detects career stage (entry / mid / senior / executive) and industry from user input, then applies the matching ruleset. Produces a single ATS-safe DOCX that scores in the low-to-mid 90s on any major ATS scanner.

调用方式：

```
/resume
```
— 与用户进行访谈，然后从头创建一份新简历。
```
/resume <file>
```
— 解析现有简历，与用户访谈以确认/补充信息缺口，然后诊断并重写以修复所有ATS识别问题。
```
/resume-score <file>
```
— 进行简短访谈获取背景信息（目标岗位+JD），然后估算ATS评分并提供分类明细。
```
/resume-jd-tune <file> <jd>
```
— 进行简短访谈确认技能资质，然后针对JD对简历进行关键词适配。

所有命令均需强制进行访谈 — 详见下方「行为规范」。

脚本文件： 位于插件根目录

scripts/

下 —

build_ats_resume.js

（DOCX生成器）、

score_resume.py

（启发式评分工具）。

该技能可跨领域协调5阶段工作流（访谈→诊断→重写→生成→评分）。能从用户输入中识别职业阶段（入门/中级/高级/高管）及行业，然后应用匹配规则集。生成符合ATS标准的单一DOCX文件，在主流ATS扫描工具中评分可达90分左右。

Behavior

行为规范

ALWAYS interview the user — every invocation, no exceptions.

No arguments (
/resume
): Run the full interview, then build resume from scratch.
File argument (
/resume path/to/resume.docx
): Parse the file, then still run the interview to confirm extracted facts and fill any gaps before diagnosing or rewriting.
Score (
/resume score <file>
): Run a brief interview (target role + JD if available) so the score is contextualized to what the user is targeting, then score.
Tune (
/resume tune <file> <jd>
): Run a brief interview (gaps in JD, skills user actually has) before tuning.
Diagnose (
/resume diagnose <file>
): Same as score — brief interview for context, then diagnose.

The interview always uses

mcp__conductor__AskUserQuestion

— one question per tool call, never batched. Numbered options where known; the tool auto-adds an "Other" free-text input on every prompt.

必须与用户进行访谈 — 所有调用均需执行，无例外。

无参数（
/resume
）：执行完整访谈流程，然后从头创建简历。
带文件参数（
/resume path/to/resume.docx
）：解析文件后，仍需执行访谈，确认提取的信息并补充缺口，再进行诊断或重写。
评分命令（
/resume score <file>
）：执行简短访谈（目标岗位+可用JD），使评分与用户目标岗位匹配，然后进行评分。
适配命令（
/resume tune <file> <jd>
）：执行简短访谈（JD信息缺口、用户实际掌握的技能）后再进行适配。
诊断命令（
/resume diagnose <file>
）：与评分命令相同 — 先进行简短访谈获取背景信息，再进行诊断。

访谈需始终使用

mcp__conductor__AskUserQuestion

工具 — 每次工具调用仅提问一个问题，不得批量提问。已知选项需编号；工具会自动在每个提示中添加「其他」自由输入项。

Orchestration Logic

编排逻辑

The skill runs 5 phases. Don't skip phases. Phase 1 (Interview) is mandatory on every invocation.

Interview (always) — Ask the questions in
```
references/interview-prompts.md
```
one at a time using the
```
mcp__conductor__AskUserQuestion
```
tool. Hard limit: ≤4 options per question (tool constraint). For wider question spaces (e.g., target role across all fields), split into a branching pair: ask field/category first, then specific role. For questions with known options (industry type, seniority, page count, tone, etc.), provide numbered choices. For open-ended questions (achievements, metrics, skills), pass
```
"options": []
```
and let the user type freely via the auto-provided "Other" input. Never include "Other" yourself — it's added automatically. Ask one question per tool call, wait for answer, then ask next. Adapt follow-ups based on previous answers — skip questions already answered or already extracted from an uploaded file (but still confirm). Minimum required before drafting: target role, industry, years of experience, current title, JD (if available), top achievements with metrics, core skills, prior ATS flags, page-count preference. For each role the user wants listed, loop through title / employer / dates / team / shipped / improved / led / metrics. Never draft, score, diagnose, or tune before completing the required-block interview.
Diagnose — If rewriting, map every issue to specific lines (vague bullet, passive voice, verb repetition, non-standard heading, missing metric). Share with the user before touching anything.
Rewrite — Apply the rules in
```
references/resume-best-practices.md
```
. Quantify every bullet, vary verbs (no verb >2x), active voice only, canonical job titles, standard section headings, target-role signal in two places.
Build — Populate
```
scripts/build_ats_resume.js
```
DATA object with the rewritten content, run with
```
node
```
. Validate page count; tighten margins / font / old roles if over target. Never cut metrics.
Score — Run
```
scripts/score_resume.py
```
on the output. Report per-category scores and the weighted overall. Honest caveat: ~94 is a realistic generic ceiling; per-JD tuning pushes to 96–98.

该技能执行5个阶段，不得跳过任何阶段。第1阶段（访谈）为所有调用的强制环节。

访谈（必选） — 使用
```
mcp__conductor__AskUserQuestion
```
工具，逐个提问
```
references/interview-prompts.md
```
中的问题。严格限制：每个问题选项≤4个（工具约束）。对于范围较广的问题（如全领域目标岗位），拆分为分支提问：先询问领域/类别，再询问具体岗位。对于有已知选项的问题（行业类型、职级、页数、语气等），提供编号选项。对于开放式问题（成就、数据、技能），传入
```
"options": []
```
，让用户通过自动提供的「其他」输入项自由填写。请勿自行添加「其他」选项 — 工具会自动添加。每次工具调用提问一个问题，等待回答后再提问下一个。根据之前的回答调整后续问题 — 跳过已回答或已从上传文件中提取的问题（但仍需确认）。开始撰写前的必填信息：目标岗位、行业、工作年限、当前职位、JD（如有）、带数据的核心成就、核心技能、过往ATS识别问题、简历页数偏好。对于用户希望列出的每个岗位，循环询问职位名称/雇主/任职时间/团队/交付成果/改进成果/领导项目/数据指标。未完成必填访谈环节前，不得进行撰写、评分、诊断或适配操作。
诊断 — 若进行重写，需将每个问题映射到具体内容行（模糊描述、被动语态、动词重复、非标准标题、缺失数据）。在修改前告知用户。
重写 — 应用
```
references/resume-best-practices.md
```
中的规则。所有项目描述均需量化，动词多样化（同一动词使用不超过2次），仅使用主动语态，采用标准职位名称、标准章节标题，在两处体现目标岗位信号。
生成 — 将重写后的内容填入
```
scripts/build_ats_resume.js
```
的DATA对象，使用
```
node
```
运行脚本。验证页数；若超出目标页数，调整边距/字体/早期岗位内容，但不得删除数据指标。
评分 — 对输出文件运行
```
scripts/score_resume.py
```
。提供分类评分及加权总分。需如实说明：通用简历的合理最高分为约94分；针对JD优化后可提升至96–98分。

Field Detection

领域识别

Detect the user's field from their target role and current title, then pick the skills-section structure:

Field	Skills categories
Software Engineering	Languages, Backend, Frontend, Data, Cloud & DevOps, Architecture & Leadership
Product Management	Methodologies, Tools, Domains, Analytics, Leadership
Marketing	Channels, Tools, Analytics, Content, Strategy
Design (UX/UI/Product)	Tools, Methods, Disciplines, Research, Leadership
Sales	Methodologies, Tools, Industries, Skills, Languages
Finance / Analytics	Tools, Modeling, Reporting, Compliance, Domain
Data / ML	Languages, ML/DL, Data Platforms, Cloud, Tools
Healthcare	Clinical Skills, EMR Systems, Certifications, Specialties
Legal	Practice Areas, Software, Languages, Bar Admissions
Operations / HR	Frameworks, Software, Domains, Compliance, Languages

根据用户的目标岗位和当前职位识别领域，然后选择技能板块结构：

领域	技能分类
软件工程	编程语言、后端、前端、数据、云与DevOps、架构与领导力
产品管理	方法论、工具、领域、分析、领导力
营销	渠道、工具、分析、内容、策略
设计（UX/UI/产品）	工具、方法、学科、研究、领导力
销售	方法论、工具、行业、技能、语言
金融/分析	工具、建模、报告、合规、领域
数据/机器学习	编程语言、机器学习/深度学习、数据平台、云、工具
医疗	临床技能、EMR系统、认证、专业方向
法律	执业领域、软件、语言、律师资格
运营/HR	框架、软件、领域、合规、语言

Career Stage Detection

职业阶段识别

Detect stage from stated years of experience and seniority, then pick length and structure:

Stage	Length	Order
Entry (0–2 yrs)	1 page	Contact → Summary (optional) → Education → Projects → Experience → Skills
Mid (3–9 yrs)	1 page preferred, 2 if dense	Contact → Summary → Experience → Skills → Education → Certifications
Senior / Staff / Principal (10+ yrs)	2 pages	Contact → Summary → Experience → Skills → Education
Executive (Director+)	2–3 pages	Contact → Executive Brief → Selected Achievements → Experience → Skills → Education → Board / Advisory

根据用户陈述的工作年限和职级识别职业阶段，然后选择简历长度和结构：

职业阶段	简历长度	内容顺序
入门级（0–2年）	1页	联系方式 → 个人简介（可选）→ 教育背景 → 项目经历 → 工作经验 → 技能
中级（3–9年）	优先1页，内容密集时可2页	联系方式 → 个人简介 → 工作经验 → 技能 → 教育背景 → 认证
高级/资深/首席（10年以上）	2页	联系方式 → 个人简介 → 工作经验 → 技能 → 教育背景
高管（总监及以上）	2–3页	联系方式 → 高管摘要 → 精选成就 → 工作经验 → 技能 → 教育背景 → 董事会/顾问经历

Quality Gates

质量校验标准

Before handing off the DOCX, verify:

Quantification ratio ≥ 60% of bullets carry a number. Target 90%+.
Verb variety — no verb appears more than 2x across the whole resume.
Active voice — 0 passive constructions ("was responsible for," "tasks included," etc.).
Section headings — exactly 4 standard: PROFESSIONAL SUMMARY, SKILLS (or TECHNICAL SKILLS), PROFESSIONAL EXPERIENCE, EDUCATION.
Page count matches user's target (1 or 2).
Job titles — every title is canonical; compound titles split.
No vague filler — "hard worker," "passionate," "team player," "results-driven" are banned.
No fabricated metrics — every number traces back to user input.
Contact info in body, not header / footer.

If any gate fails, re-run the rewrite phase on the failing section before building.

交付DOCX文件前，需验证以下内容：

量化比例 ≥60%的项目描述包含数据指标，目标为90%以上。
动词多样性 — 整个简历中同一动词使用不超过2次。
主动语态 — 无被动结构（如“负责”“包含任务”等表述）。
章节标题 — 严格使用4个标准标题：PROFESSIONAL SUMMARY、SKILLS（或TECHNICAL SKILLS）、PROFESSIONAL EXPERIENCE、EDUCATION。
页数符合用户目标（1页或2页）。
职位名称 — 所有职位名称均为标准名称；复合职位需拆分。
无模糊套话 — 禁止使用“勤奋”“热情”“团队合作者”“结果导向”等表述。
无虚构数据 — 所有数据均来自用户输入。
联系方式 位于正文，而非页眉/页脚。

若任何校验项未通过，需对不合格部分重新执行重写阶段，再生成文件。

Community Footer

社区信息页脚

When to show

展示时机

After producing a final DOCX if this is the user's first interaction with the skill.
After
```
/resume score
```
when the score is in the 90s.
When the user asks "who built this" or "how do I contribute."

首次使用该技能并生成最终DOCX文件后。
使用
```
/resume score
```
命令且评分达到90分以上时。
用户询问“谁开发了这个工具”或“如何贡献”时。

When to skip

跳过时机

During the interview phase.
During iteration (user asking for changes).
When the user explicitly asks for brief / no-frills output.

Footer text: "Built with the

resume

skill for Claude Code. MIT-licensed, contributions welcome at https://github.com/kunalrawat425/claude-resume."

访谈阶段。
迭代过程中（用户要求修改时）。
用户明确要求简洁输出时。

页脚文本："Built with the

resume

skill for Claude Code. MIT-licensed, contributions welcome at https://github.com/kunalrawat425/claude-resume."

Reference Files

参考文件

```
references/resume-best-practices.md
```
— full 2026 do's/don'ts across 10 sections.
```
references/interview-prompts.md
```
— exact questions the skill asks the user.
```
references/scoring-rubric.md
```
— per-category weighting for the score estimator.

Load these only when needed — don't preload into context during the interview phase.

```
references/resume-best-practices.md
```
— 涵盖10个板块的2026年完整规范。
```
references/interview-prompts.md
```
— 技能向用户提问的具体问题。
```
references/scoring-rubric.md
```
— 评分估算的分类权重标准。

仅在需要时加载这些文件 — 访谈阶段请勿预加载到上下文。

Scoring Methodology

评分方法

ATS Score (0–100)

ATS评分（0–100分）

Weighted across 8 categories. See

references/scoring-rubric.md

for full per-band criteria.

Category	Weight
Parsability / format	15%
Section headings	10%
Quantified impact	18%
Verb variety	12%
Active voice	10%
Vague-filler absence	10%
Length appropriateness	10%
Visual simplicity (no images / tables / columns)	15%

基于8个分类进行加权计算。完整分级标准详见

references/scoring-rubric.md

。

分类	权重
可解析性/格式	15%
章节标题	10%
量化成果	18%
动词多样性	12%
主动语态	10%
无模糊套话	10%
长度合理性	10%
视觉简洁性（无图片/表格/分栏）	15%

Priority Levels

优先级等级

When diagnosing, rank issues for the user:

Critical — prevents the resume from parsing at all (photo, table-based layout, missing sections).
High — drops the score 5+ points (vague bullets, passive voice, verb repetition 3+, non-standard job titles).
Medium — drops the score 2–5 points (redundant phrasing, long bullets, irrelevant skills).
Low — polish (en-dash vs hyphen, skill ordering, tense consistency within a single role).

诊断时需为用户划分问题优先级：

严重 — 导致简历完全无法被解析（照片、表格布局、缺失章节）。
高 — 导致评分下降5分以上（模糊描述、被动语态、动词重复3次以上、非标准职位名称）。
中 — 导致评分下降2–5分（冗余表述、过长描述、无关技能）。
低 — 优化项（短横线与连字符区别、技能排序、单一岗位内时态一致性）。

Sub-Skills

子技能

resume-score — Standalone heuristic scorer. Invoke with
```
/resume-score <file>
```
. Has its own slash command.
resume-jd-tune — Standalone JD keyword tuner. Invoke with
```
/resume-jd-tune <file> <jd>
```
. Has its own slash command.

resume-score — 独立启发式评分工具。调用命令为
```
/resume-score <file>
```
，拥有独立斜杠命令。
resume-jd-tune — 独立JD关键词适配工具。调用命令为
```
/resume-jd-tune <file> <jd>
```
，拥有独立斜杠命令。

Subagents

子代理

None currently. The skill runs inline.

目前无任何子代理，该技能为内联运行。

Error Handling

错误处理

Can't parse the uploaded file — fall back to asking the user to paste the content as text. Don't guess.
Missing metric — ask the user for the number; never invent.
Verb repetition unfixable — if the user has 8 bullets all about "launching," diversify the verbs to Released / Shipped / Rolled out / Built / Deployed / Developed / Delivered / Created. Don't reuse a verb more than 2x.
Content overflows 2 pages — tighten in this order: margins (0.5–0.6"), font (body to 10pt), oldest roles to 1–2 bullets. Never cut metrics.
User requests designer variant (sidebar, photo, pills) — this skill deliberately doesn't produce those because they score lower on ATS. Recommend a separate tool for recruiter-facing pretty versions.
User provides fake or unverifiable metrics — include them but flag in the response; final responsibility is the user's.

无法解析上传文件 — 转而请求用户粘贴文本内容，请勿猜测。
缺失数据指标 — 向用户询问具体数据，不得虚构。
动词重复无法避免 — 若用户有8条关于“launching”的描述，替换为Released / Shipped / Rolled out / Built / Deployed / Developed / Delivered / Created等动词，确保同一动词使用不超过2次。
内容超出2页 — 按以下顺序调整：边距（0.5–0.6英寸）、字体（正文字号改为10pt）、最早的岗位保留1–2条描述。不得删除数据指标。
用户要求设计师版本（侧边栏、照片、胶囊样式） — 该技能刻意不生成此类版本，因为它们在ATS中的评分较低。建议用户使用其他工具生成面向招聘者的美观版本。
用户提供虚假或无法验证的数据 — 可包含这些数据，但需在回复中注明；最终责任由用户承担。