Job Search Kit

An open-source toolkit for fixing your résumé and LinkedIn with an AI agent. Clone it, point your agent at it, and invoke each capability as a skill (auto-triggered by intent) or a slash command (explicit /command). Works with Claude Code and Cowork.

The philosophy throughout: diagnose before rewriting, anchor everything to your real target jobs, and never fabricate a number. Wherever the agent lacks your data it emits [METRIC NEEDED] instead of inventing one.

Capabilities

Each is a standalone skill and a slash command.

Capability	Skill / Command	What it does
Job Analyzer	job-analyzer	Parse a JD into requirements + literal ATS keywords. Feeds the résumé tools.
Resume Review	resume-review	Brutally honest diagnostic of an existing résumé. Read-only — finds problems, doesn't rewrite.
Resume Improve	resume-improve	Strengthen an existing résumé overall (bullets, summary, skills, ordering).
Resume Build	resume-build	Build a résumé from scratch via guided intake.
Resume Tailor	resume-tailor	Re-aim a résumé at one specific job, matching its exact keywords.
Resume Interview	resume-interview	Interactive Q&A that fills [METRIC NEEDED] gaps and probes for improvements.
Cover Letter	cover-letter	Tailored one-page cover letter (hook → value-match → proof → close) for a specific job.
LinkedIn Rewrite	linkedin-rewrite	The 5-step sequenced LinkedIn profile workflow.

All résumé outputs render to a single-column, ATS-safe PDF (the only output format).

Requirements

• An agent with a filesystem and shell — Claude Code or Cowork. The résumé skills run a script to produce the PDF, so a plain browser chat (no shell) can do the writing but cannot render the PDF.
• Python 3 with the rendering deps: pip install -r requirements.txt (WeasyPrint; or the Playwright fallback — see requirements.txt).

The LinkedIn prompts and all the text work need none of the above; only PDF rendering does.

Install

This repo is a self-contained Claude Code plugin and a single-plugin marketplace, so it installs as one unit (don't cherry-pick individual skill folders — internal paths resolve via ${CLAUDE_PLUGIN_ROOT} and only work when installed whole).

Option A — install as a plugin (recommended):

/plugin marketplace add mvineethr/job-search-kit
/plugin install job-search-kit@job-search-kit

Skills are then namespaced, e.g. /job-search-kit:resume-review, /job-search-kit:resume-tailor.

Option B — load locally for development:

git clone https://github.com/mvineethr/job-search-kit
claude --plugin-dir ./job-search-kit

Option C — no install: keep the repo in your working folder and tell the agent "use the job-search-kit skills in this repo" (works in Cowork, paths resolve against the repo root).

How to use it

New here? Read INSTRUCTIONS.md for the full step-by-step guide (install, inputs, every capability, the end-to-end workflow, theming, troubleshooting). The quick version:

1. Gather inputs (see templates/inputs-checklist.md): your résumé, your LinkedIn PDF, and 3–5 target job descriptions.
2. Analyze the target: /job-analyzer on your JDs.
3. Pick your move:
   • No résumé yet → /resume-build
   • Have one, want the truth → /resume-review
   • Want it stronger generally → /resume-improve
   • Targeting one posting → /resume-tailor
   • Fixing LinkedIn → /linkedin-rewrite
4. Fill every [METRIC NEEDED] with a real, defensible number.
5. Render the PDF (résumé tools do this for you):

   pip install -r requirements.txt
   python scripts/render_pdf.py shared/resume-html/resume-template.html FirstLast_Resume.pdf
   

6. ATS self-check: open the PDF, select-all → copy → paste into a text editor. If the text comes out in order with nothing missing, it'll parse.

Strategy: optimize once vs. tailor per job

The kit is built around a two-tier model — get this right and you stop wasting effort:

Stable layer (set once, refresh occasionally). Your LinkedIn profile and your master resume. LinkedIn is a single public page every recruiter sees, and you don't control when they look — so you can't and shouldn't re-tailor it per application. Optimize both once for your target role family, using the keywords that recur across several JDs (not one posting), then refresh quarterly or when your target shifts. → linkedin-rewrite (profile) and resume-improve (master resume).

Per-application layer (do every time). Your submitted resume — this is one file per application, so per-JD tailoring belongs here and only here. Tailor a copy of the master resume to each posting's exact requirements and keywords. → resume-tailor.

Rule of thumb: LinkedIn + master resume = role-level and stable; the submitted resume = job-level and tailored. Keep all three factually consistent (same titles, dates, employers) — only emphasis and keywords shift per job, never the facts.

ATS rules

The single source of truth is shared/ats-rules.md — single column, standard headings, literal JD keywords, real selectable text, no tables/columns/ images, PDF only. Every résumé skill follows it.

Themes: the HTML template has a theme layer (CSS variables for accent color, font, and density) with presets — Classic, Modern (navy/teal), and Compact — so two people using this kit don't produce an identical-looking résumé. Themes change only the look; the ATS-safe structure never changes.

Repo layout

job-search-kit/
├── .claude-plugin/        # plugin.json manifest + marketplace.json
├── requirements.txt       # PDF rendering deps (WeasyPrint, pypdf)
├── README.md  ·  LICENSE (MIT)  ·  CONTRIBUTING.md  ·  .gitignore
├── skills/                 # auto-triggered skills (one folder each)
│   ├── job-analyzer/        resume-review/      resume-improve/
│   ├── resume-build/        resume-tailor/      linkedin-rewrite/
├── commands/              # /slash-command wrappers (one .md each)
├── prompts/linkedin/      # the 5 sequenced LinkedIn prompts
├── INSTRUCTIONS.md        # detailed step-by-step usage guide
├── PLAN.md                # vision + roadmap (planning doc)
├── shared/
│   ├── ats-rules.md        # shared ATS reference
│   ├── resume-html/        # ATS-safe single-column HTML template (themeable)
│   └── cover-letter-html/  # matching cover-letter template
├── scripts/render_pdf.py  # HTML → PDF (WeasyPrint, Playwright fallback)
├── templates/             # inputs checklist + voice doc
└── examples/              # sample output

What's original here

This kit is not a copy of someone's prompt list. The bulk of it is original work:

• A full résumé toolkit — resume-review (diagnostic), resume-improve (general strengthening), resume-build (from scratch via guided intake), resume-tailor (targeted to one JD), and resume-interview (interactive gap-filling) — none of which come from any external source.
• A job-analyzer that turns job descriptions into structured ATS keyword profiles.
• A shared ATS ruleset, an ATS-safe single-column HTML résumé template, and an HTML→PDF rendering pipeline (render_pdf.py).
• Skill + slash-command packaging and a Claude Code plugin so it installs and runs as one unit.
• A design philosophy enforced across every capability: diagnose before rewriting, anchor to real job descriptions, and never fabricate a metric ([METRIC NEEDED]).

Acknowledgment

One component — the LinkedIn profile workflow (linkedin-rewrite, in prompts/linkedin/) — adapts a 5-prompt sequencing idea from Abhijay Arora Vuyyuru's AI Action Letter #29, "I rewrote my LinkedIn with 5 Claude prompts". That method (an uncopyrightable idea) informed one of the seven capabilities; the prompts were rewritten and the rest of the kit is independent. Credited here out of good practice.

License

MIT — see LICENSE. Clone it, fork it, share it.