从 JD 到定制简历，一套流程走完求职材料
装好之后怎么用
什么时候会用上它：Complete resume and CV builder toolkit for tailoring applications to job descriptions. Use this skill when the user wants to build, edit, or critique a resume or CV, generate a cover letter, optimize a LinkedIn profile, extract achievements from research papers, or set up a knowledge base from their work history. Triggers on: /make-resume, /edit-resume, /critique-resume, /make-cl, /new-session, /career-workbook, /optimize-linkedin, /setup-extract, /setup-build-kb, or any request to tailor a resume to a job posting, write a cover letter, critique a resume, optimize a LinkedIn profile, extract paper achievements, or build a resume knowledge base. This skill bundles nine sub-skills: career-workbook, new-session, make-resume, edit-resume, critique-resume, make-coverletter, optimize-linkedin, setup-extract, and setup-build-knowledgebase.
怎么用：在 AI 工具里直接描述任务即可自动触发。显式调用的写法各端不同：Claude Code / Cursor 等用 /resume-toolkit，Codex 用 $resume-toolkit，Kimi Code 用 /skill:resume-toolkit，ZCODE 用 $resume-toolkit，Gemini CLI 以自动激活为主。
适合需要针对不同岗位持续定制简历的人，尤其适合经历较多、项目较多，容易在不同简历版本之间失控的求职者。
它把事实核验、JD 拆解、项目筛选、简历生成、五视角审查和修改复审串成一条完整流程，重点避免虚构数据、遗漏岗位硬性要求，以及“只改措辞却没有提高匹配度”。
我的用法是：先用 career-workbook 核实经历，再用 make-resume 生成初稿，最后通过 critique-resume 和 edit-resume 完成审查与修改闭环。
技能内容
SKILL.md版本与文件 ›Resume Toolkit
Based on ARPeeketi/claude-resume-kit (MIT) — the original anti-fabrication, knowledge-base-first resume system. This fork extends it as a Claude Code skill for PM/TPM/PO applications.
A complete resume and CV builder. This meta-skill bundles nine sub-skills — read the relevant one for your task:
| Command | Sub-skill file | What it does |
|---|---|---|
/new-session | skills/new-session/SKILL.md | Two-step entry point: confirm Career Workbook (Step 1), then display session prompt with 4 JD input options (Step 2) |
/career-workbook | skills/career-workbook/SKILL.md | Display, review, and update the Career Workbook standalone |
/make-resume | skills/make-resume/SKILL.md | Generate a tailored resume or CV from a job description |
/edit-resume | skills/edit-resume/SKILL.md | Edit resume/CV or cover letter from critique feedback |
/critique-resume | skills/critique-resume/SKILL.md | Score and critique a resume/CV against a JD |
/make-cl | skills/make-coverletter/SKILL.md | Generate a tailored cover letter |
/optimize-linkedin | skills/optimize-linkedin/SKILL.md | Suggest copy-paste LinkedIn profile changes (goal-first, then JD) |
/setup-extract | skills/setup-extract/SKILL.md | Extract achievements from research papers into KB |
/setup-build-kb | skills/setup-build-knowledgebase/SKILL.md | Synthesize extractions into resume-ready knowledge base |
Standard Workflow (Job Application)
/new-session
│
├── STEP 1: Career Workbook Review (skills/career-workbook/SKILL.md)
│ Display timeline, certifications, verified metrics
│ ├── "yes / looks good" → proceed to Step 2
│ └── corrections provided → apply edits, loop back to confirm
│
└── STEP 2: Session Prompt (skills/new-session/SKILL.md)
Display template + 4 JD input options:
2a Copy-paste JD inline
2b File path or folder (JDs/*.txt)
2c Upload .txt / .pdf / image
2d URL (Claude fetches + extracts JD)
↓ (user fills template + provides JD)
/make-resume [JD] ← Phase 0: JD analysis, comp, strategy, framing — STOP
← Phase 1: Bullet plan — STOP
← Phase 2: Generate resume markdown + bullet audit — STOP
↓
┌─────────────────────────────┬──────────────────────────────┐
│ resume markdown done │ /make-cl │
│ (already complete) │ (generate CL markdown now) │
└─────────────────────────────┴──────────────────────────────┘
↓ (both markdowns ready)
/critique-resume ← Score resume + CL together
↓
┌─────────────────────────────┬──────────────────────────────┐
│ /edit-resume (resume) │ /edit-resume (CL) │ ← apply tier-1 fixes
└─────────────────────────────┴──────────────────────────────┘
↓ (edits approved — CLI work ends here)
┌─────────────────────────────────────────────────────────────┐
│ DOCX + PDF → use Claude Web (claude.ai) │
│ Paste the final markdown → ask Claude to generate DOCX/PDF │
│ Claude Web produces better-formatted output than CLI │
└─────────────────────────────────────────────────────────────┘Key principle: CLI generates all markdown (resume + CL), critiques, and edits. DOCX and PDF are always produced in Claude Web (claude.ai) — paste the approved markdown and ask Claude to generate the DOCX/PDF. Claude Web renders significantly better output than CLI for document formatting.
Effort Defaults (embedded — no /effort command required)
| Sub-skill | Effort | Reason |
|---|---|---|
/new-session | low | Orchestration only; no generation |
/career-workbook | low | Display + optional edit; no resume generation |
/make-resume | medium | Structured phases; high effort adds no bullet quality |
/make-cl | medium | Fixed structure; flows from Phase 0 framing |
/edit-resume | medium | Targeted edits only |
/critique-resume | high | Scoring depth benefits from thorough treatment |
How to Invoke a Sub-skill
When the user invokes a command, read the corresponding sub-skill file from the skills/ folder and follow its instructions exactly.
For example, if the user types /make-resume JDs/Acme.txt:
- Read
skills/make-resume/SKILL.md - Follow its startup, phase, and stop instructions
All sub-skills share the same project workspace conventions (CLAUDE.md, session_learnings.md, Career_Workbook_Master_Lokapati.md, output/, JDs/).
Knowledge Base
- Career Workbook:
../career-progression/Career_Workbook_Master_Lokapati.md— read ONCE at startup; never re-read mid-session - Session Learnings:
session_learnings.md— file existence map, token patterns, DOCX build workflow - Job Tracker:
job_tracker.md— application status; append a row after every new session
Verified Workbook Metrics (canonical list — do not use metrics outside this list without explicit Workbook citation)
| Workbook | Verified Metrics |
|---|---|
| W1 (Alcon) | Metrics marked TBD — use directional language only ("improved", "reduced") |
| W2 (Baxter Sr 2022) | 30% product release delay reduction; 25% audit readiness improvement; 30% manual process reduction; 40% audit findings reduction; 500+ LoVs consolidated |
| W3 (GCHP) | 98% data accuracy; 15% audit trail improvement; 12% reconciliation error reduction; zero operational disruption |
| W4 (Baxter Jr 2019) | 30% faster product release; 100% supply chain traceability; 30% faster recall decision cycles; 75% manual reconciliation reduction; 30% onboarding time reduction; zero data loss |
| W5 (Kiddie Commute) | 20% CSAT improvement; 30% onboarding reduction; $5M Series A; 12% operational cost reduction |
| W6 (USD Recycling) | 25% drop-off time reduction; 75% laptop recovery rate; $50K revenue |
| W7 (Nike) | Supply chain KPI framework — 6 composite metrics; no quantified company outcomes |
| W8 (Infosys) | Merger context (TWC–Charter $78.7B) — no personal metrics confirmed |
| W9 (TCS) | 1M+ agent platform scale; specific % TBD |
Fabrication rule: If a metric is not in the table above, do NOT use it. Write directional language instead.
评论 0
更多
登录后可点赞、收藏、评论和举报。
还没有评论，先发起一个具体问题。